FR2604814A1 - LANGUAGE ANALYZER - Google Patents

Abstract

LANGUAGE ANALYZER ACCORDING TO THE INVENTION INCLUDES A 1018 DICTIONARY IN WHICH DICTIONARY DATA COMPRISING MORPHEM DATA FOR WORDS, COMPOSED WORDS AND LOCATIONS, AND A MEANS OF EFFECT 1016 MORPHOLOGICAL ANALYSIS OF A PHRASE INTRODUCED WITH REFERENCE TO THE DICTIONARY, OR THE DICTIONARY CONTAINS DATA OF DEGREE OF COUPLING WHICH INDICATES THE DEGREE OF COUPLING BETWEEN EACH OF THE WORDS CONSTITUTING THE WORDS COMPOSED AND THE LOCATIONS, AND THE MEANS OF ANALYSIS DICTIONARY FOR THE RESPECTIVE WORDS CONTAINED IN THE ENTRY PHRASE AND, WHEN SEVERAL DICTIONARY DATA ARE FOUND FOR A SINGLE WORD IN COMBINATION WITH OTHER WORDS, IT CHOOSES THE HIGHEST COUPLING OF WORDS BASED ON THE DATA OF THE DEGREE OF COUPLING.

Description

REQU TE (continued) 2 6 0 4 81 4 st use this package continued "in case

  one of the locations in the sayrpj-rait request bundle, insufficient to contain the information to be provided. Each of the sections thus completed on this Tla's'e will be identified by its

  néro and its title l p. ex. * 2 Applicant (continued)).

  7) DECLARATION OF PRIORITY DATE OF DEPOSIT NUMBER

  Country of origin JP 24/10/86 n 251916/86 JP 27/10/86 n 253763/86 JP 27/10/86 n 255512/86 October 3, 1987 being a Saturday, we validly deposit this day Monday October 5 1987; MV Paris 208.79.35 DB0162 1060, 06-79 continued The present invention relates to a language analyzer and, in particular, it relates to a language analyzer intended to analyze natural languages for use in communication devices.

automatic translation.

  The prior art language analyzers make

  intervene the problems described below.

  When morphing an input string into a sentence, for example of words, it is important, and this gives a significant effect to the result of the analysis, to determine whether

  a certain word is used alone or if it is used in a common mnt

  posed or a phrase, coupled with one or more words.

  Conventional analyzers employ a system which

  is to conduct the analysis for the case where successive words are supposed to constitute a sentence and for the case where they are independent words and finally to choose an appropriate translation on the basis of its result, or else a system consisting of preferentially determine these successive words as constituting a sentence. However, the former system requires an extended processing time, while the latter system involves risks

  high to produce a wrong translation.

  During the morphological analysis carried out for the production of a translation, a "part of speech" or some other information about the morpheme, for example a word, is obtained by means of a search in a dictionary. In the

  common words such as nouns and verbs, since most

  part of them can be stored in the dictionary, we

  can easily find them for information.

  However, since expressions such as those indicating, for example, length (m), speed (m / s), acceleration (m / s2) and the like include very diverse genres, it is not effective to store them all in the dictionary, since this would increase so

  useless memory capacity relating to diction information-

nary.

2 6 0 4 8 1

  In the case of morphological analysis of a sentence, the numerical expression in a certain language cannot always agree exactly with those of other languages. By

  example, the fundamental concept of counting numbers, that is

  say different positional notation between Japanese language and

  European languages like English for example.

  So, for example, if the English expression for the numeric value "a hundred and two thousand two hundred and four

  (one hundred and two thousand two hundred and four) "breaks down simple-

  its constituents and is easily replaced by the expression

  corresponding Japanese sion, it can be analyzed simply in the form "100 and 2200 and 4". This would ultimately translate correctly to "102,204", that is, "10-man 2-sen 3-yaku 4 (10-ten

  mille 2-mille 2-cent 4) "in Japanese. In the moral analysis

  phological carried out at the time of the formation of the translation, an introduced sentence is divided into reference units of the dictionary by means of delimiters such as spaces, periods and commas, and a search is made in the dictionary by means of units dictionary reference to produce part of speech or other information. In this case, a proper name for example can be used in two or more different meanings, depending on the context. For example, "Osaka City (the City of Osaka)" is an expression used to indicate a group acting as subject, for example "the City of Osaka decided ...", and in the sense indicating the place and acting as an object, such as "... in the City of Osaka". To this extent, since only one meaning is stored for each proper name, it cannot ensure different uses which are function of the meaning, this

  which reduces the accuracy of the analysis.

  For example, in the case of the English sentence "... in the Central Park John Willson had a ... (in Central Park John Willson has a ...)", the sentence cannot be grammatically analyzed unless its recognition comes only after the separation between the proper name "Central Park" and the other proper name "John Willson" in this context. Similarly, in the case of the English phrase "... in Boston Mr. Baker was ... (... in Boston Mr. Baker was ...)", this must be recognized after separation between the proper name "Boston" and the proper name "M. Baker". However, in the classical system, a series of similar proper names is erroneously recognized as constituting as a whole a

unique proper name.

  For example, in the morphological analysis of a sentence

  English, a succession of words starting with a capital letter-

  cule is generally grammatically analyzed as constituting a proper noun. However, if words starting with an uppercase letter continue like this, it is not always appropriate to recognize them as a single proper name as a whole. There is a case where they are actually several proper names which appear successively by an occasional fact. In the morphological analysis carried out at the time of production of the translation, part of the speech or other information is obtained by searching in a dictionary. In this case, since most common nouns, verbs, etc. can be stored in the dictionary, you can easily store them

  find to obtain the information sought.

  However, since variable forms of words exist for proper names, it is impossible to store them all

  in a dictionary. Therefore, these unregistered proper names

  very in the dictionary cannot be recognized as

proper names.

  In the case where a set of characters combined in a particular configuration is present, there is a high risk

  to produce an erroneous analysis if the processing

  tined to regular word units which is applied to sentences

  ordinary is then applied to a similar set of charac-

  tères, and this can lead to a meaningless translation.

  For example, in the case where there is a set of characters-

  tères en anglais "Sunday, 26 Jan., '80 (Sunday 26 January 1980) ', it is very difficult to translate them into Japanese, because you need them iL is very difficult to translate them into Japanese, because you need them

  express in the form "80-nen, 1-gatsu, 26-nichi, nichiyobi (80-

years, 1-month, 26-day, Sunday) ".

  In addition, for derived words, since no estimate

  tion was only done for the speech part and the nature

  particular semantics, it cannot, depending on the case, be obtained

no exact translation.

  The grammar used to analyze language, by context-independent grammar (cfg), has the disadvantage of providing many useless solutions which cannot ultimately be used even with the use of the method

  called from top to bottom and the method called from bottom to top.

  Many of these solutions are clearly recognized

  useless as mistakes when you actually read them. Any-

  times, since the transformation of the structure or the formation of the translation also takes place for these useless solutions, after which the validity of the result is determined during the respective processing steps, this results in a waste of the processing time. For example, the English word "let (leave, or particle of imperative)" has both directions of instruction and invitation, so that grammatical analysis must be carried out for the two respective possibilities, which reduce the

  yield. In addition, it is difficult to choose between them.

  Furthermore, in the English language, several words can be combined by means of hyphens, such as for example in "take-care-of-him attitude" for so freely form an adjective group. However, it is difficult to manipulate these words to

  using ordinary grammar analysis, as is usual

tuel.

  The situation is also identical for the

  While the form of the redundant query in the English language is extremely limited, it requires a

  extremely complex treatment in the usual analysis method.

  It is not easy to determine to which verb the interrogation

  redundant, that is to say "isn't it?" (isn't it?), relates.

  The invention was made in order to respond to the pro-

above exposed problems.

  The first object of the invention is to provide a language analyzer capable of determining the extent of the degree of coupling between two successive words and capable of determining

  whether or not they form a phrase on the basis of the result of the operation

ration.

  A second object of the invention is to provide an analysis

  language sister capable of carrying out a morphological analysis in an introduced sentence containing a set of composite characters, for example units of dimension, without

  having to store all of these character sets in a dictionary

tionnaire.

  A third object of the invention is to provide an analysis

  language sister capable of carrying out a moral analysis

  appropriate phology in relation to an expression containing

numerical values.

  A fourth object of the invention is to provide an analysis

  language sister capable of translating a proper name according to a

  meaning that is consistent with the context.

  A fifth object of the invention is to provide an analysis

  language sister capable of morphologically analyzing in a way

  correct strings for an expression containing more

successive proper names.

  A sixth object of the invention is to provide a language analyzer capable of processing an unregistered proper name and capable of performing an appropriate analysis for a proper name while taking into account the existing relationship

  with groups of words arranged before and after him.

  A seventh object of the invention is to provide an analysis

  language sister capable of correctly analyzing strings morphologically in the case of a set of characters expressing a specific meaning because they are combined according to

a special rule.

  An eighth object of the invention is to provide an analysis

  language sizer capable of giving an estimate of grammatical nature, meaning, etc. of a word recognized as being a derived word

  obtained according to a predetermined rule.

  A ninth object of the invention is to provide an analysis

  language sister capable of recognizing the structural particularity

  of a proposed sentence and perform the analysis

according to the particularity.

  The first object of the invention can be achieved using a language analyzer which includes a dictionary storing dictionary data containing the data associated with morphemes with regard to words, compound words and sentences, and

  a means of morphological analysis allowing to apply the analysis

  morphological lysis of a proposed sentence while referring to the dictionary, o the dictionary contains data relating to the degree of

  coupling indicating the degree of coupling between words that make up

  kill a compound word and a phrase, and

  the means of analysis refers to diction-

  for the respective words contained in the proposed phrase and, when several data from the dictionary have been found for a single word in combination with other words, it chooses the combination of words which has the highest degree of coupling

  while referring to the data relating to the degree of

  beach. The second object of the invention can be achieved using a language analyzer comprising

  an input means for introducing a set of characteristics

  ters of a predetermined language, a fundamental dictionary used for the operations of

  search relating to the set of characters entered and emma-

  wasting fundamental data, and a means of analysis for the set of introduced characters calling upon the search in the dictionary, o the means of analysis carries out the analysis of a set of characters introduced when a part of the 'together

  characters was found by search in the basic dictionary

  mental with regard to the set of characters introduced, by an analogous search in the fundamental dictionary for

  the other part of the character set.

  The third object of the invention can be achieved using a language analysis comprising: a dictionary storing dictionary data on all the search units in the dictionary, and an analysis means serving to divide a sentence introduced in dictionary reference units and to perform an analysis of elements related to the units

  dictionary reference while referring to diction-

nary, o

  the dictionary contains, under the data of the diction-

  nary, indications of distinction making it possible to indicate that the reference unit of the dictionary expressing numbers is intended to express numbers, and

  the means of analysis refers to diction-

  for the respective dictionary reference units contained in the entered sentence and, when the indications of distinction are contained in the extracted dictionary data, it combines the dictionary reference unit for which the indication of distinction was found with another

  dictionary reference unit with which another indication

  tion of distinction was found disposed in the vicinity of the university

  dictionary reference tee just mentioned, it calculates the numerical values signified by the two dictionary reference units so as to obtain a single numerical value and puts the two dictionary reference units in the form of a

single unit of analysis.

  The fourth object of the invention can be achieved using a language analyzer comprising:

2 604814

  an input means for introducing a set of characteristics

  mothers of a predetermined language,

  a dictionary used to find a set of characters

  tères introduced from the means of entry,

  a means of research used to consult the diction-

  in relation to the set of characters introduced, a means of producing nature information which produces information relating to nature for the set of characters not recorded in the dictionary and the set of characters whose information of nature is not recorded in the dictionary among the set of characters introduced, o the means producing information relating to nature is designed to produce several information of nature for

  the set of characters having no information of a nature.

  The fifth object of the invention can be achieved using a language analyzer comprising: a dictionary which stores dictionary data on all the reference units of the dictionary, and an analysis means serving to divide the sentence proposed in reference units of the dictionary and to perform a morphological analysis of input strings for the units of

  dictionary reference while referring to diction-

nary, o

  the dictionary contains, under the data of the diction-

  nary, information of distinction making it possible to specify an admissible position in a succession of proper names consisting of several successive proper names relative to the reference unit of the dictionary which signifies a proper name, and

  the means of analysis is designed to refer to diction-

  for the respective dictionary reference units

  contained in the proposed sentence and, when the information available

  extinction is contained in the dictionary data found, it

  combines the dictionary reference unit for which the information

  mation of distinction has been found with another reference unit of the dictionary which means another proper name and is located in the vicinity of the reference unit of the dictionary which has just been mentioned according to the position specified by the information of distinction to obtain a single unit of analysis. The sixth object of the invention can be achieved using a language analyzer comprising:

  an input means for introducing a set of characteristics

  mothers of a predetermined language,

  a dictionary used to find the set of characters

  tères introduced by the input means, and a means of analysis of the information relating to the nature, used to find in the dictionary what concerns the set of characters introduced, then to analyze the information of nature relative to the set of characters, o the means of analyzing information from

  nature is designed to analyze information

  defining the nature of a set of characters while taking into consideration the information relating to the nature of the set of characters which is arranged before and after the set of characters which has just been mentioned. The nature information analysis means is designed to analyze the nature information of multiple rows of letters

  by arranging the character sets collectively.

  The seventh object of the invention can be achieved using a language analyzer comprising: a dictionary which stores the dictionary data on all the reference units of the dictionary, and an analysis means serving to divide a sentence introduced in dictionary reference units and performing the morphological analysis while referring to the dictionary, o the analysis means is designed to distinguish a succession of dictionary reference units containing a

  element of specific meaning as an expressive assembly unit

  mant a specific meaning constituted under a certain

260 4 8 1 4

  rule and forming a succession of diction reference units-

  who have the elements of specific meaning in one

  single unit of grammatical analysis.

  The eighth object of the invention can be achieved using a language analyzer in which the grammatical nature, the semantic nature or the translated word are estimated for the words which are not recorded in a dictionary and for the words who

  are recognized as derivative words based on a particular

  rity proper to the morpheme, like an affix.

  The ninth object of the invention can be achieved using a language analyzer comprising:

  a first means of analysis used to morpho-

  logically analyze a proposed sentence of a predetermined language,

  a second means of analysis used for syntaxi-

  cally analyze the structure of a sentence of the language on the basis of the result of the morphological analysis by the first means of analysis, a dictionary of dictionary data for the language used with the first and second means of analysis and,

  control means for searching the diction-

  in order to bring the first and second means of analysis to perform the analysis, o

  the first means of analysis is designed to perform operations

  search rations in the dictionary, distinguish the particu-

  morphological differences of the proposed phase of the language in order to distinguish the arrangement in relation to the structure of the sentence, and estimate the nature to be the result of the analysis relating to the arrangement and the role of the arrangement with regard to the sentence structure that exists in the sentence, and

  the second means of analysis is designed for morpholo-

  gically analyze the structure of the surface layer of a sentence of the language by applying a grammatical rule based on the estimated nature and role, and to analyze a possible subsidiary relationship for the constituents contained

in the sentence.

  According to the language analyzer of the invention, an automatic translation can be obtained with high speed and

high quality.

  The following description, intended as an illustration of

  the invention aims to give a better understanding of its characteristics

  teristics and advantages; it is based on the appended drawings, among which: FIGS. 1 to 10 illustrate a first embodiment

  of the language analyzer according to the invention, applied to a device

  English to Japanese machine translation, and more particularly

  especially, Figure 1 is a functional block diagram illustrating

  an example of the detailed structure of a mor-

  phological; Figure 2 is a functional block diagram illustrating the entire structure; - Figure 3 is an explanatory view illustrating an example of the structure of a dictionary-type file attached using a flag with the highest preference; Figure 4 is a flowchart illustrating an example of a morphological analysis process;

  Figure 5 is a flowchart illustrating an example of the pro-

  entry stops found in the morpho- analysis process

  logic; FIG. 6 is an explanatory view illustrating an example of formatting of a set of characters introduced; FIG. 7 is an explanatory view illustrating an example of search in the dictionary; FIGS. 8A to 8D are flowcharts illustrating an example of the operation of eliminating the contradiction relating to the flag with the highest preference in the morphological analysis; Fig. 9 is an explanatory view illustrating an example of the contents of the dictionary information buffer found after it has been called the dictionary; FIG. 10 is an explanatory view illustrating an example of the

  contents of the dictionary information buffer found in the result-

  status of the operation to eliminate the contradiction for the

  skin with the greatest preference;

  Figures 11 to 16 illustrate the second embodiment

  tion of the invention, and, more particularly,

  Figure 11 is a block diagram of the embodiment

  tion; FIG. 12 is a diagram illustrating an example of data stored in a dictionary file; Figure 13 is a diagram illustrating an example of

  data stored in a dictionary file of basic units

  mental .; FIG. 14 is a diagram illustrating an example of data stored in a dictionary information storage table; Figure 15 is a flowchart illustrating the operation of the device; Figure 16 is a flowchart illustrating the recognition of units;

  Figures 17 to 29 illustrate the third embodiment

  tion of the language analyzer according to the invention, applied to an English-Japanese automatic translation device, and, more particularly, FIG. 17 is a functional block diagram

  illustrating an example of the detailed structure of the morpho- analysis

  logic; Figure 18 is a functional block diagram illustrating the structure as a whole; Figures 19A and 19B are flow charts illustrating an example of the morphological analysis; Fig. 20 is a flow chart illustrating an example of the collective arrangement of the currency symbol and the unit in morphological analysis; Figs. 21A and 21B are flow charts illustrating an example of the numbers linked by hyphens in the morphological analysis; Figures 22A and 22B are flowcharts illustrating a

  example of treatment of successive numbers in morpholo-

  gic; Figures 23A and 23B are flow charts illustrating a

  example of collective arrangement with the preceding numerical value

  dent in morphological analysis; FIG. 24 is an explanatory view illustrating an example of

  the structure of a dictionary file attached using a dra-

  digital skin; Fig. 25 is an explanatory view illustrating an example of a set of characters entered; FIGS. 26A to 26D are explanatory views illustrating

  the content of the diction information storage table-

  nary as it is extracted from the dictionary for the set of characters entered which is illustrated in FIG. 25 according to the processing steps; FIG. 27 is an explanatory view illustrating another example of the set of characters introduced; Figure 28 is an explanatory view illustrating the content

  a table of currency symbols, a table of notations

  tional, from a table of decimal points in the dictionary; Figures 29A to 29D are explanatory views illustrating

  an example of the diction information storage table-

  nary as it is extracted from the dictionary for the set of characters introduced represented in FIG. 27, according to the processing steps;

  Figures 30 to 36 illustrate the fourth embodiment

  tion according to the invention, and, more particularly,

  Figure 30 is a block diagram of the embodiment

  tion; FIG. 31 is a diagram illustrating an example of the data stored in a reference dictionary; Figure 32 is a flow diagram illustrating the operation of the entire device; FIG. 33 is a flowchart illustrating the processing of the proper name registered in a dictionary; FIG. 34 is a flow chart illustrating the processing of the proper name not registered in the dictionary; FIG. 35 is a flowchart illustrating the processing making it possible to produce information of a missing nature; FIG. 36 is a diagram illustrating an example in which the data stored in the dictionary information reserve table is modified after the processing of the entered sentence;

  FIGS. 37 to 46 illustrate the fifth embodiment of the language analyzer according to the invention, applied to a device

  English-Japanese machine translation, and, more particularly, Figure 37 is a functional block diagram illustrating

  an example of a detailed structure of the mor-

  phological; Figure 38 is a functional block diagram illustrating the structure as a whole; FIG. 39 is an explanatory view illustrating an embodiment for the structure of a dictionary file; Figure 40 is a flowchart illustrating an example of morphological analysis for a proper name; FIG. 41 is a flowchart illustrating an example of the collective arrangement for proper names recorded in a dictionary during the morphological analysis; FIGS. 42, 43 and 44 are flow diagrams illustrating an example of processing according to the position information of the proper name in the analysis of a proper name FIG. 45 is a flow diagram illustrating an example of the collective arrangement for proper names not recorded in a dictionary during morphological analysis; Figures 46A to 46F are explanatory views showing the

  content of the diction information storage table-

  naire, reported in the dictionary, for the example of a set of characters introduced, according to the processing steps; FIGS. 47 to 52 illustrate the sixth embodiment of the invention, and, more particularly,

  Figure 47 is a block diagram of the embodiment

  tion; Figure 48 is a diagram illustrating an example of data stored in a reference dictionary; Figure 49 is a flow diagram illustrating the operation of the entire device; Figure 50 is a flowchart illustrating the processing of proper names registered in the dictionary; Figure 51 is a flowchart illustrating the processing of proper names not registered in the dictionary; Fig. 52 is a diagram illustrating an example in which data stored in the dictionary information reserve table is changed after processing the entered sentence;

  Figures 53 to 57 illustrate the seventh embodiment

  tion of the language analyzer according to the invention, applied to an English-Japanese automatic translation device, and, more particularly, FIG. 53 is a functional block diagram illustrating an embodiment of the detailed structure of section d 'morphological analysis; Figure 54 is a functional block diagram illustrating the entire structure; Figures 55A and 55B are flow charts illustrating an example of the morphological analysis; FIG. 56 is an explanatory view illustrating an example of the

  content of the information table belonging to the production section

  morphemes processing information; Fig. 57 is an explanatory view illustrating an example of the contents of an adaptation table;

  Figures 58 to 63 illustrate the eighth embodiment

  tion according to the invention, and, more particularly, Figure 58 is a block diagram for explaining the entire structure; Figure 59 is a block diagram for explaining, an example of processing of derived words through a prefix; FIG. 60 is a block diagram for illustrating an example of processing of derived words by means of a suffix; Figure 61 is a block diagram showing complete details prepared by a synthesis of Figures 58, 59 and 60; Figure 62 is a block diagram showing other details for a complete unregistered word processing section of Figure 61; and Figure 63 is a block diagram for explaining an embodiment of an automatic translation device to which the invention is applied;

  Figures 64 to 90 illustrate the ninth embodiment

  tion of the language analyzer according to the invention, applied to a

  English-Japanese translation system, and, more particularly-

  Figure 64 is a functional block diagram illustrating the entire structure; FIG. 65 is a collective functional principle diagram showing the function of recognizing the structural arrangement of an English sentence introduced in one piece; FIG. 66 is a flowchart illustrating an example relating to the collective arrangement of a block in relation to the English sentence introduced; Figure 67 is a flow chart illustrating the details of a word output operation in the processing program; Fig. 68 is an explanatory view illustrating an example of Dictionary information for words or phrases of English stored in a dictionary of words; Figure 69 is an explanatory view illustrating an example

  table data for the head of block condition, the con-

  end of block edition and the condition of goal and role estimation, stored in an analysis rules file; FIG. 70 is an explanatory view illustrating an example of collective arrangement for the structure; FIG. 71 is an explanatory view illustrating an example of collective arrangement for the block; Fig. 72 is an explanatory view illustrating an example of information on English and information on words collectively arranged in a block;

  Figure 73 is a flowchart illustrating an example

  ple of an analysis operation illustrated in an analysis section; FIG. 74 is an explanatory view similar to that of FIG. 68, illustrating an example relating to the entry into a

  dictionary of expressions in the case where this embodiment pos-

* has an estimation function for identical cases; FIG. 75 is an explanatory view similar to that of FIG. 69, illustrating an example for the head and end condition of the block and a table of the block preparation information

  in the case where this embodiment has an estimation function

  identical cases; Figure 76 is a functional block diagram similar to that of Figure 64, illustrating the entire structure in a modification of this embodiment; FIG. 77 is a functional block diagram similar to that of FIG. 76 for the collection of the function of the analysis of the information on "let" in relation to the modified embodiment presented in FIG. 76;

2 6 0 4814

  Fig. 78 is an explanatory view illustrating an example of dictionary information which contains "Let" information for English words and phrases stored in the word dictionary of the modified embodiment; FIGS. 79 and 80 are explanatory views similar to that of FIG. 72, illustrating an example of the block information and the word information where a sentence in English containing the information on "Let" is collectively arranged in a block; FIG. 81 is a flowchart similar to that of FIG. 66 and of FIG. 73, illustrating an example of a program for the collective arrangement of information on "Let" with regard to the English sentence introduced; Figure 82 is a flowchart similar to that of

  figure 73, illustrating an example for the analysis processing con-

  keeping the information on "Let" executed in the analysis section for the modified embodiment; FIGS. 83A and 83B are flowcharts illustrating an example of the program for the analysis of the information on "Let" for the English sentence entered; Fig. 84 is a flowchart illustrating an example of the grammatical analysis program for hyphenated words in the case of the introduced English sentence; Figure 85 is an explanatory view similar to that of Figure 72, illustrating an example of the block information and word information which are collectively arranged based on the introduced English sentence containing the word with a dash union, in one block; Figure 86 is a functional block diagram similar to that of Figure 64, illustrating the entire structure of another modified embodiment; Figure 87 is a block diagram similar to that of Figure 65, where a redundant query analysis function for the English sentence entered is collectively arranged in the modified embodiment shown in Figure 86; Figures 88 and 89 are explanatory views similar to that of Figure 72, illustrating an example of block information and word information collectively arranged from the English sentence containing a redundant query, in one block; and Figures 90A and 90B are flow charts illustrating an example of a redundant query analysis program for

the English sentence introduced.

  The first embodiment according to the invention will be described. FIG. 2 shows the overall constitution of the first embodiment, in which a language analyzer according to the invention is applied to an English-Japanese automatic translation device. The invention can of course be applied effectively not only to a similar English-Japanese machine translation device which translates English into Japanese, but also to any of the language analyzers in

  which sentences of the input language are mainly ana-

  lysed when translating from one language to another. The embodiment shown in the figures has an input section 1010, through which an English text 1012 to be translated into Japanese is introduced. The input section 1010 includes, for example, a keyboard having character keys such as alphanumeric keys or function keys, an optical character reader (OCR) device for reading English text recorded on paper, and, or well, a file memory device for reading English text recorded on a memory medium, for example a magnetic disc. The English text introduced in the input section 1010 is read and is supplied to a pre-formatting section 1014, where a translation pre-processing is carried out. In this section, it is mainly carried out a sentence recognition processing for unknown words. This operation is part of

morphological analysis.

  Pre-formatted English data is transferred

  at the same time as the information obtained during pre-shaping, to a morphological analysis section 1016. Section 1016

  of morphological analysis divides the sentence by referring to a dictionary

  1018 word tutor, morphologically analyzes the chain of

  English phrase, performs different types of agen-

  cements, such as treatment of unknown words, proper names,

  expression of time, expression of numbers, etc. and made

  kills a process related to the entire sentence, such as

  redundant derogation and recognition of identical cases. The rule specific to morphological analysis is stored in

  a 1036 file of grammar analysis rules.

  English data obtained after morpholo-

  gic are transferred at the same time as the information relating to the dictionary obtained by morphological analysis to a section I of morphological analysis 1020. Section I of morphological analysis

  1020 is a functional section that performs morpholo-

  gic of the surface structure of an English sentence by applying a grammatical rule to English data and which puts

  all structural possibilities.

  After the analysis in section I of morphological analysis 1020, the English data is sent at the same time as the information of analysis to a section II of syntactic analysis 1022. In this section, a solution is chosen by application

  a description of the structure resulting from the result of the related analysis

  tive to the surface layer by section I analysis. A plausible analysis tree for the English sentence is thus prepared in order to form the structure thereof. These analysis rules are also stored in the file

1036 analysis rules.

  After the analysis, the English data is transferred, as data relating to the analysis tree, to a section

  1024 of structural transformation. Section 1024 of transform-

  structure mation prepares a corresponding sentence tree in Japanese from a structural tree, i.e. a structure

  intermediate of the English sentence and transforms it into a struc-

  underlying Japanese ture from which one can easily

translate into Japanese.

  Structural tree data showing structure

  underlying in Japanese which are thus subject to transforma-

  structure structure are sent to a translation production section 1026, where the translated sentence is formed. This is a working section for producing a Japanese sentence to

  from the tree structure of The structural tree in Japanese.

  The Japanese sentence data thus translated, that is,

  ie the data of the translated sentence is sent to a post-formatting section 1030. The post-formatting section 1030 modifies the translated sentence data by referring to the 1018 dictionary while using the information used in the translation process to arrive at a Japanese sentence.

  more natural. The Japanese sentence data is trans-

  linked to an output section 1032, then delivered, under the translated Japanese sentence 1034, by the output section 1032. The output section 1032 comprises for example a printer, a display and, or alternatively, a memory device to file, like a

magnetic disc.

  The progress of the successive translation operations is controlled by a control section 1038 which governs the operation of the device as a whole. The 1018 word dictionary stores dictionary data for words in the English and Japanese languages where not only the vocabulary, but also various information such as connection relationships, namely coexistence relationship, meanings, forms of the plural and singular, 'part of speech, etc., are established in this embodiment. In addition, the analysis rules file 1036 stores the rule data for morphological analysis and

syntactic analysis.

  The control section 1038 is connected to an operation display section 1040. The operation display section 1040 includes operation keys by which an operator gives various instructions to the device of the invention, such as keys translation instruction or cursor keys, a

  display or indicator that displays intro English text

  duit, the Japanese sentence which is the result of the translation,

  intermediate data such as diction information-

  various instructions for the operator, etc. The

  most of the operation display functions can be

26 0 4 8 1 4

  killed so as to be contained in a keyboard if it is arranged in the input section 1010 or in a display if it is

  supplied for output section 1032.

  We pass to FIG. 1, which shows the detailed structure of your section 1016 of morphological analysis. The morphological analysis section 1016 has an input device 1100, for example a keyboard for the input section 1010 and an input interface 1104 which interfaces with the file of the input document 1102. The interface input 1104 is arranged so as to include an input character set buffer receiving the data relating to the character set in English in the form of code data, for example ASCII, from the input device 1100 or from the input document file 1102, and it temporarily stores the data of the character set. The set of characters entered can be the one that

  is pre-shaped in pre-shaping section 1014.

  The morphological analysis section 1016, as shown in the figure, includes a processing section 1106, a dictionary transfer section 1108, a processing section for rules for eliminating contradictions, and a control section 1112. processing section 1106 is a functional analysis section for performing the morphological analysis and it includes an extracted dictionary information buffer, that is to say a dictionary information storage table 1120 (we can refer to figure 9). The morphological analysis is carried out on the basis of the following operations: the instruction is given to search the dictionary in an order which starts from the top of the set of input characters according to the set of characters of the search key, l dictionary information obtained from the 1108 dictionary search section

  relative is stored in the diction information buffer-

  nary extracted 1120, and the treatment of obtaining a degree

  preferably according to the flag of higher preference above

after described is executed.

  Dictionary search section 1108 is a working section used to retrieve information from the dictionary by searching the dictionary of words 1118 based on the set of search key characters that were the subject of an instruction from the processing section 1106 and to transfer it to the processing section

treatment 1106.

  The 1118 word dictionary stores grammatical information, such as parts of speech and La

  entry section for each of the words, as well as a prefecture flag

  highest frequency, as shown in Figure 3 for the example of the input information. The dictionary is considered to be a dictionary type file with a flag of

  highest preference. The "highest preference flag"

  vée "is a flag indicating the extent of coupling between words contained in a compound word or a phrase constituting the dictionary entry, and" 0 "indicates weak coupling or no coupling while" 1 "indicates coupling In this case, use as a phrase is considered valid for a compound word or

  a phrase considered to have strong coupling, otherwise the pos-

  sibility of use as separate words is also

considered in parallel.

  As illustrated by the example in Figure 3,

  each of the entries in the word dictionary 1118 is arranged res-

  pectively for the compound word, the locution and the distinct words constituting them, without making a difference between the distinct words

  and the word compound or phrase. In addition, each bending form

  It constitutes a separate entrance. If there are several inflectional forms, they are respectively recorded as different entries. The type of bending is displayed in the bending section. The situation is the same for parts of speech, where a recording is allowed for several parts of speech, and the information relating to parts of speech is

  contained for each of them. Like other information, the character-

  accounting or non-accounting of a noun, the fact that a verb is transitive or intransitive, or the fact that it is a word

translated, etc. are saved.

  For example, "get (and many others

  meaning) "is the infinitive form of a verb and the prefect flag

  The highest frequency is "0". The phrase "get up (get up, go up, etc.)" is an expression for an infinitive verb and its highest preference flag is "1". In addition, the prepositional group "up to" has the flag of highest preference "1", but a nominal group such as "white house" considered a composite name has the flag of more high preference "0", which shows, in the latter case, that the

  degree of coupling between words is low. In the figure, the sym-

  bole..U "indicates a vacancy of character.

  In this way, the dictionary information found in the dictionary search section 108 contains the flag of highest preference. In the case where, if "1" is set for the highest performance flag for sets of

  identical characters or overlapping sets of characters

  song, a similar contradiction must be eliminated. Section 1110 dealing with rules for eliminating contradictions

  eliminates the contradiction and processes it

  while referring to the rule of elimination of contradictions

  tions for the flag of highest preference which is stored

  in file 1036 of the analysis rules.

  The rule for eliminating contradictions is applied in the embodiment, for the way in which the preference is

  granted, in the order of (1) to (3).

  (1) The phrase or word whose part of the speech is a verb.

  (2) The compound word, the phrase or the word with many

constituent words.

  (3) The compound word, the phrase or the word located in the part

before the sentence.

  The usage used for the word thus selected, namely the analysis unit, is displayed as active information relating to the dictionary information buffer found 1120 by the processing section 1016. The active information shows that the unit of analysis is valid or effective if it is "1", while it shows that its possibility is not

used if it is "0".

  The control section 1112 is a functional section used to govern and control the operation and the processing

  in each of the functional sections of section 1016 of ana-

  morphological lysis. The section can be incorporated into the control section 1038 to control the

device as a whole.

  The result of the morphological analysis is transferred via an output interface 1114 to the analysis section I 1020. In the case where the result is not directly transferred to the analysis section I 1020, it is once stored in the analysis input file 1116 and

  in the information file of the analysis dictionary 1118.

  In this embodiment, while all the words, compound words or phrases starting from the cut-off part of the dictionary reference unit are taken during the morphological analysis, the information from the dictionary obtained for distinct words constituting the compound word or phrase for which it has been decided that it is a collective "unity" according to the flag of highest preference is rejected. That is, the extent of coupling between words in a sentence is determined while reference is made to the flag of highest preference

  for dictionary information obtained in morpho- analysis

  logic. Of the compound words or phrases, those that are determined to have strong coupling are considered to be used as phrases in the sentence and, if not, the possibility of their use as separate words

  is also considered in parallel.

  Such processing using the flag of highest preference is carried out using the sequence presented on the

  figure 4. Data relating to intro character sets

  duits are received from input section 1010 (1200),

  the input character set is cut by the reference unit

  dence dictionary dictionary search

  nary 1018 with the flag of highest preference (1201), it is searched in the dictionary 1018 according to this (1203), the search being carried out until the final position of the sentence indicated by the data of the set of characters introduced (1202), then the contradiction relating to the highest flag

  preference is eliminated (1204), and the result of the mor-

  is issued in section I of analysis 1020

(1205).

  In input processing 1200, the data is

  first read from or available from the 1102 input document file

  input sitive 1100 to be placed in the buffer of the input character set of the input interface 1104 (1210, see Figure 5). The data relating to the set of input characters are introduced for example in the form ASCII, and,

  when the EOF symbol (end of file) is read, the processing section

  tement 1106 writes the NULL code (empty) in the set's buffer

  input characters for the end position.

  Then, the processing section 1106 gives a new form to the set of input characters (1211). For example, if two or more characters belonging to the character corresponding to a space follow one another, they are modified and corrected into one

  vacancy character. The character corresponding to a space

  carries the vacancy character (represented by the symbol.U ..), the tabulation, the line break (represented by the symbol "" v), etc. In addition, these characters corresponding to a space appearing between the vertex and the first produced character which is other than the character corresponding to a space in the set buffer.

  input characters are eliminated.

  For example, the input character set "UUIUwillUUgetUuputoUgoUtoUaUwhiteUhouseU ...- (I will get up to go to a white house) is reset to the form shown in Figure 6, namely" IUwillUgetUup utou gouto U aUwhiteUhouse ... .. (NULL) ". The position of the symbol (NULL)

  indicates the final position in the buffer.

  Dictionary reference delimiters which are used for the 1201 clipping process of the reference unit

  of the dictionary are arranged at the position of the alphabetic character

  tick, numeric character, apostrophe, and characters such

  that the hyphen and the semicolon as well as for the apostro-

  phe following vacancy characters. The processing section

  1106 has an upper pointer for the reference of the

  dictionary, which is first placed at the head of the buffer.

  Then section 1108 search in the diction-

  naire consults the dictionary1118 file attached using the

  flag of highest preference using the character set

  tères which goes from the character indicated by the head pointer to the character preceding the following delimiter as a set of search key characters. Dictionary entry field and search key character set

  are subject to comparison and, if all are the same, the information

  Dictionary entry is accepted (1203). This agreement is considered good when all the character set of the entry of the diction-

  naire is in agreement with at least part of the set of characters starting from the upper end and if the part which comes just after this part is a reference delimiter of the

  dictionary, an apostrophe or period. For example, as shown

  felt in figure 7, if the head pointer indicates the character

  lead g "in the search key character set," get upon (get on) "," get (get) "and" get

  upon "from the dictionary entry agree with that.

  Dictionary information retrieved is stored

  in buffer 1120 of dictionary information found apparently

  related to processing section 1106. At the same time as what is read, the recognized initial position and the final position of

  the set of characters are stored. This clarifies the posi-

  tion of the characters in the input buffer in an order starting from the head. An accumulation region for active information is arranged in the dictionary information buffer 1120, which is information indicating whether the information of the

  dictionary found is effective or not for The rest of the process

  All of these are set to "1" at this stage.

  Then, the head pointer is updated on all the references in the dictionary and it is positioned on the

  character just after the delimiter that appears next to it

  Immediate swim of the present pointer when looking at the set of characters from left to right. The postponement to

  dictionary is therefore performed successfully. In the example above

  above mentioned, the character at the head of the dictionary reference unit is indicated by using "I" for "I" at the beginning, then "w" for "will", then finally "g" for "get". When the head pointer arrives at NULL, it is determined that it

  is in the final position (1202).

  Figure 9 shows an example of information from the dic-

  found in this way for the example of the whole

  of English characters introduced which has been described above.

  We will now give explanations concerning the treatment for eliminating contradictions 1204 carried out by the

  section 1110 of treatment of the rules for eliminating contradictions

  at the same time as referring to the file of election rules

  mination of contradictions for the file of greatest prefe-

  rence 1036, in connection with FIGS. 8A to 8D. The flowchart presented in FIGS. 8A and 8B illustrates the processing carried out in the case where the positions relative to the words where the flags of highest preference are positioned overlap with one another,

  while the flowchart of Figures 8C and 8D illustrate the process

  elimination applied to the units of analysis, i.e.

  say items with the flag of highest preference, that is

  to say a treatment allowing to transform the active information into "0". In these flowcharts, the reference "≤" represents the substitution, "'" represents the reference, and "Po-x" represents the content of x that the entry field of the pointer p respectively has. First, a group of words which each have a higher reference flag of "1" and whose positions in the sentence overlap each other are detected (steps 1220 to 1223). Then, the rule of elimination of the flag of highest preference is applied to each of the groups detected, and those which, among them, prove effective are selected (steps

1224 to 1235).

  For example, in the embodiment described above, the flag of highest preference "1" is set to "get up" for the start position "8" and the end position "13" and to "up to" for the initial position "12" and the end position "16" with regard to the group of characters "get up to", as shown in FIG. 9, and the positions of the characters

  overlap each other. Then, rule (1) as below

  above described is first applied to serve to determine whether or not it is a verb at the same time as reference is made to the speech part of the "psave" reserve pointer and to the part of speech from the pointer p (1224). In this example, since it corresponds to a verb, the combination "get up" is chosen. If the rule (1) is not satisfied, the rule (2) is applied (1228), and the length (lens) of the character set

  associated with the entry of the psave reserve pointer is com-

  trimmed with the length (len) of the set of characters associated with the entry of the pointer p. In addition, if rule (2) is not satisfied either, rule (3) is applied (1229), and the start position associated with the initial position of the psave reserve pointer is compared with the start of position associated with the initial pointer position p. Then, when any one of the rules (1) to (3) for eliminating contradictions has been satisfied, the application being done in this order, the active information in the section

  input input, i.e. not effective, is trans-

  formed as "0" (1232), while the other, i.e. the active information of the other entry item, i.e. of the actual entry, is left as it was, at "1" (1231). Such a

  application of the rule of elimination of contradictions is applied

  quée in succession while the pointer p advances by degrees (1234, 1235) to the final position for each of the successive entry headings, the active information being put in the form "1" in the only case of a effective entry section. The state corresponding to the example mentioned above is presented in FIG. 10. For example, for the entry item "up to", its

  active information is put in the form "0".

  Then, the elements being overlapped, even partial, for the position associated with the combination in which the active information and the flag of highest preference are both "1" are detected (1236 to 1241), and their active information is returned to "0" (1142, 1249). The application of such a rule for eliminating contradictions is executed in the order given for each of the entry headings while the pointer p advances by degrees (1243, 1248) to the final position and the active information of ineffective entry items is reduced to "0". Thus, the active information which relates to "get" and to "up" is brought back to "0", for example, for the entry item "get up" (FIG. 10). For the "white", "white house" and "house" entries, since all their flags of highest preference are "0" even if the positions are overlapping, their information

active are kept at "1".

  In this way, when the processing has been carried out just before the final position (NULL), the content of the input buffer of the input interface 1104 and of the dictionary information buffer found 1120 are delivered by the output interface 1114 to analysis section I 1116. The content of the dictionary information buffer found 1120 is only delivered for the input item for which "1" is indicated as active information. For example, the content of the

  input buffer can be written to the input file of the analysis

  lyse 1116, while the content of the information buffer

  dictionary entry found 1120 can be written in the

  analysis dictionary information file 1118.

  In this case, since the active information and the flag of higher preference are both also delivered, the structure of the analysis dictionary information file 1118 is identical to that of the dictionary information buffer found. It may also happen that active information and

  the flag of highest preference is not issued.

  We will now describe more specifically the second

embodiment.

  FIG. 11 illustrates the second embodiment of the Language analyzer according to the invention, applied to a device

  English-Japanese machine translation.

  This embodiment has an input section 2014, to which data is delivered from a device

  2010 entry file or a 2012 entry document file.

  sitif 2010 includes for example a keyboard having character keys such as for example elphanumeric keys or

  function keys, and an optical character reading device

  tères allowing to read a text in English recorded on a

  sheet of paper. The 2012 input document file is a dis-

  memory positive in which the English text is saved

  on a memory medium, for example a magnetic disc.

  The 2014 input section includes a 2014a input character set buffer and stores the input English phrase that is delivered by the 2010 input device or the

  2012 input document file to the character set buffer

  input ters 2014a. The 2014 section reads the input sentence emma-

  stored in the 2014 input character set buffer and the

  delivers to a 2016 processing section.

  The 2016 treatment section is a functional section.

  nelle performing the morphological analysis relative to the sentence provided which is delivered by the entrance section 2014

  by performing a search in a dictionary file, i.e.

  say by consulting a dictionary file or by trying to find what you are looking for. The 2016 treatment section

  includes a table 2016a for storing information in the dictionary

  shareholder and it stores the information obtained by the consulta-

  tion of a dictionary file 2022 or a dictionary file of fundamental units 2026 which will be described below to place it

  in the 2016a table for storing diction information-

  nary. The 2016 processing section consults a set of consultation key characters in a single unit in the case

  of consulting the dictionary from the set of characters

  tères constituting the entry sentence which was issued by the entry section 2014. The set of consultation key characters is consulted in an order starting from the first character of the set of characters constituting the sentence

  entry, based on a predetermined consultation rule.

  For example, the input sentence is split from the top of the sentence, in order, using delimiters, such as spaces, commas, etc., and the sets of divided characters are used as a set, respectively. of consultation key characters. In this case, the sets of characters expressing units such as m, km, m / s respectively consist of a set of consultation key characters. The processing section 2016 sends the consultation key character set, which is found from the character set constituting the input sentence, to the consultation section 2020

from the dictionary.

  The 2020 search section in the dictionary con-

  a 2022 dictionary file is consulted on the basis of the sets of consultation characters which are delivered by the processing section 2016. In the 2022 dictionary file, the entry items and grammatical information such as the speech parts are stored as indicated on Figure 12. If there is an entry, or entry item, in the dictionary file

  2022, the consultation section of the 2020 dictionary reads the information

  mation of the speech, etc. part of this entry and delivers it to the 2016 processing section. If there is no entry in the

  dictionary file 2022 following consultation of the dic- file

  2022, the dictionary consultation section 2020

  inform the 2016 processing section of the situation.

  The 2016 processing section stores the speech part information, etc., found by the consultation section of

  2020 dictionary in a 2016a information storage table

  Dictionary mation. If there is no entry for the set of lookup key characters in the dictionary file

  2022, the 2016 processing section delivers the entire character-

  touch keys to a recognition section

unit 2024.

  The unit recognition section 2024 consults a dictionary file of fundamental units 2026 on the basis of the set of consultation key characters issued by The processing section 2016. The entries of fundamental units are stored in the dictionary file of 2026 fundamental units, as shown in Figure 13 if there is no entry in the 2026 dictionary file of fundamental units, or units

* basic measurement, the set of consultation key characters

  tion is divided into several sets of characters, as described below, and the 2026 dictionary of base units is consulted several times. Then, if base unit entries are present more than once in the search

  performed in the 2026 base unit dictionary file, more

  Several unit information is obtained from the base unit entries. If there is no base unit entry in any of the different search results, iL is obtained information indicating that the object sought is not saved in

the dictionary.

  The 2024 unit recognition section provides the base unit entry, the compound unit information, and the information that the word is not recorded in the

  dictionary to the processing section 2016. The processing section

  2016 stores this information delivered by the 2024 unit recognition section in the information reserve table of the 2016a dictionary. Dictionary storage table 2016a stores for

  conservation the entry section relating to character sets

  lookup key terms and grammatical information, for example the speech part obtained by searching in the dictionary file 2022 or the base unit dictionary file 2026 relating to the set of lookup key characters. After the data has been stored in the information storage table of the dictionary 2016a, the processing section 2016 delivers this data at the same time as the sentence entered at the output interface 2018. The output interface 2018 delivers the input sentence and the data relating to the morphological analysis supplied by the processing section 2016 to an output device 2030, for example a printer or a display, or else to a memory file 2032,

for example a magnetic disc.

  According to another possibility, the input sentence and the data of the morphological analysis delivered by the processing section 2016 are supplied directly to an analysis means (not shown) for carrying out the analysis relating to the input sentence in the means of analysis and, moreover,

  prepare a translated sentence based on the analysis.

  The control section 2028 is used to control the function-

  each of the functional sections of the device and can

  advantageously be constituted by a microcomputer.

  We will now explain the operation of the device.

  tif in relation to the organization chart presented in figure 15.

  At the start, the English sentence provided as input is read in the 2010 input device or the 2012 input document file to be placed in the 2014 input section (2100). The input sentence read and placed in the 2014 input section is

  stored in the input character set buffer 2014a.

  The input sentence stored in the character set buffer

  entry points 2014a is then read and issued to the

2016 treatment.

  In the processing section 2016, when the input sentence is delivered, the dictionary search unit is subjected to a cutting operation (2102). Thus, the set of characters constituting the input sentence is divided, in the name of a

  predetermined rule, in sets of key characters

  sultation which form the search units in the dictionary file 2022 or the dictionary file of basic units 2026, these units being successively taken from the head of the set of characters. It is then decided whether or not the split set of consultation characters is present (2104) and, if it is present, the set of consultation key characters is sent to the

  Dictionary Lookup Section 2020.

  When the search key character set is sent to the dictionary search section 2020,

  the latter consults the 2022 dictionary file to search for

  expensive lookup key character set (2106). It is determined whether the lookup key character set

  is present or not in the entry section of the diction- file

  nary 2022, as shown in FIG. 12, and, if the entry is present, grammatical information, for example a portion of speech, stored in the dictionary file 2022 is read and this read data is sent to the processing section 2016 to be stored in the dictionary information storage table 2016a (2110). Then, the program returns to step 2102 and the dictionary consultation unit is

  again subject to a cut operation.

  If there is no entry in the dictionary file 2022, the dictionary consultation section 2020 returns the set of consultation key characters to the processing section 2016, and the processing section sends the set of consultation key characters in the unit recognition section

  2024, where units are recognized (2112).

  In the case where the set of consultation key characters sent to the consultation section of the 2020 dictionary consists of ordinary words, such as nouns and verbs, since there are entries for most of them in the dictionary file 2022, the grammatical information such as the speech part is read from the dictionary file 2022 and the corresponding data is sent to the processing section 2016 and recorded in the dictionary information storage table 2016a. As described above, the entry fields for common words such as nouns and verbs are formed, but no entry is formed for the set of characters expressing units in the 2022 dictionary file. Therefore, in the case where the set of characters of consultation key is a set of characters expressing a unit such as kg or m / s, since there is no entry in the dictionary file 2022, The program goes to

  step 2112 for the purpose of recognizing the units.

  The unit recognition operation carried out in step 20112 will now be explained in relation to the

figure 16.

  When the set of characters from the consultation key

  tion, for which no entry is present in the dictionary file 2022 when consulting the dictionary file 2022, is sent from the processing section 2016 to the unit recognition section 2024, the pointer P is positioned on the character lying at the head part of the set of

  characters of c Lookup in the recognition section

sance of units 2024 (2200).

  Then, the unit recognition section 2024 consults the base unit dictionary file 2026 for

  the set of characters starting from the character for which the point-

  P P was positioned (2201). In this search, it is decided whether the base unit for which the entry exists in the 2026 base unit dictionary file appears or not as a

  complete character set in character set par-

  both of the character for which the pointer P was positioned and whether

  starts or not starts with the character for which the pointer P was posi-

  tionné. So, it is tempted to find if a set of characters

  comprising one or more characters starting from the character-

  tère for which the pointer P was positioned conforms to any one of the fundamental units for which it exists

  an entry item in the dictionary of basic units

  2026. For example, in the case where the characters on which the pointer P is positioned are k, m, s, etc., input items exist in the base unit dictionary file 2026 for these particular characters. starting from the character on which the pointer P was positioned, as represented on La

figure 13.

  The unit recognition section 2024 determines whether or not the entry exists in the base unit dictionary file 2026 as a result of the search carried out in the base unit dictionary file 2026 (2204) and, if the entry exists, the pointer P advances the Length of the recognized base unit (2208). Consequently, in the case where the basic unit is k, m, s, etc., the pointer P advances by a single character, then positions itself on the next character of the set of characters of consultation key. The 2024 unit recognition section determines whether

  a set of characters starting from the character on which the point-

  P P is positioned or not yet present (2208). In the case where such a set of characters is still present, the program returns to step 2202, o the dictionary file of basic units

  2026 is again consulted in relation to the character set

  tères which starts from the character for which the pointer P is posi-

  tionné. Then, the section decides whether the entry item exists or not in the base unit (2204) as a result of the search carried out in the dictionary file of base units 2026 and, if the entry is present, the pointer P advance the length of the unit

recognized base.

  advance the length of the recognized base unit.

  In step 2208, if the set of characters starting from

  character for which the pointer P is positioned is no longer pre-

  sent, the search in the dictionary file of basic units

  2026 comes to an end, i.e. the recognition of unity

  was successfully completed.

  For example, in the case where the set of consultation key characters delivered to the unit recognition section 2024 is km / s, which represents a unit, the entry is not present in the dictionary file d 'base units 2026, since km / s is itself a complex unit. Thus, the pointer P is first placed on k (2200), and k is found in the dictionary of base units 2026 to confirm the presence of

the entrance (2202).

  Then, the pointer is positioned on m (2206), and m is found in the base unit dictionary file 2026

  (2202) to confirm the presence of the entry, in the same way.

  Since the 2024 unit recognition section considers the forward slash (/), period (.), Etc. as part of the unit, the pointer then positions itself on the s by jumping the "/" symbol from km / s (2206). So, s was found in the base unit dictionary file 2026 to confirm the presence of

  the entrance, in the same way, (2202). As a result, since the pre-

  sence is confirmed for each of the characters k, m and s during the search in the dictionary file of fundamental units 2026, it is decided that km / s is a set of characters expressing a unit. In this way, in the case where entry headings exist in the base unit dictionary file 2026 for all the characters constituting the set of characters of the consultation line, or else in the case of headings entry exists in the basic unit dictionary file 2026 for all characters except symbols such as slash and period, etc. which are considered part of the unit, it is determined that the character set of key

  consultation is a set of characters that express unity.

  When the 2024 unit recognition section has finished consulting the 2026 base unit dictionary file

  and succeeded in recognizing the compound unity, it delivers information

  tion unit thus obtained in the processing section 2016, to store it in the dictionary information storage table 2016a (2210). The recognition of the units is therefore over. In step 2204, if no entry is present in the base unit dictionary file 2026 as a result of the

2 6 0 4 8 1 4

  search in the 2026 base unit dictionary file for

  the set of characters starting from the character on which the point

  tor P is set, this means that the character set cannot be recognized as a basic unit or a composite unit. Thus, the unit recognition section 2024 sends information to indicate that the character set is a

  word not registered in the dictionary, that is to say an informa-

  tion indicating that the word does not express a unit, intended for the processing section 2016, this information being put in

  reserve in the diction information reserve table-

  naire 2016a of The treatment section 2016, which puts an end to

  the unit recognition operation.

  Referring now to FIG. 15. When the recognition of the units (2112) has ended, the program returns to step 2101 and a cut-out intended to isolate the reference unit from the dictionary is again carried out by the section of processing 2016. After cutting the reference unit from the dictionary, the processing section 2016 determines whether or not the isolated unit is still present (2104) and, if the isolated unit, i.e. a

  set of consultation key characters, is no longer pre-

  smells, it delivers the information stored in the dictionary information storage table 2016a to the

  output via the 2018 output interface (2114).

  Analysis of the input sentence was therefore carried out.

  As described above according to this embodiment, the English sentence supplied as input is divided into sets of key characters for consultation and is first searched for in an ordinary dictionary file 2022 and, if there is no corresponding entry item in the 2022 dictionary file, units are recognized. In unit recognition, the set of consultation key characters is divided and identified by the pointer P, and the dictionary file of the basic units 2026 is consulted for each of the divided character sets. Then, those which are recorded in the dictionary file of fundamental units 2026 or those which are composed of a succession of sets recorded in the dictionary file of fundamental units 2026 are determined.

  as sets of characters expressing units.

  Since it is possible to recognize

  sance of units even for a set of characters expressing a complex unit by combining basic units stored in the basic unit dictionary file 2026, an analysis corresponding to the multiform expression of the units can be performed. In addition, since it suffices that the base unit dictionary file 2026 stores basic units, for example k, m, s, etc., and that there is no need to store complex compound units from combinations of basic units, for example km, km / s, etc., it is possible to reduce the capacity of the

dictionary file.

26 0 4 8 1 4

  The third embodiment will now be explained. FIG. 18 illustrates the general structure of the third embodiment, in which the language analysis according to the invention is applied to an English-Japanese automatic translation device. This embodiment includes an input section 3010, through which an English text 3012 to be translated into Japanese is introduced. The input section 3010 can comprise, for example, a keyboard having character keys, for example alphanumeric keys, or function keys, an optical character reader (OCR) making it possible to read the English text recorded on a sheet. paper and / or a file memory device for reading the recorded English text

  on a memory medium, for example a magnetic disc.

  The English text introduced by the input section 3010 is read and supplied to a pre-formatting section 3014 where pre-processing is carried out for translation. In this case, the recognition of the sentence and the processing of unknown words are mainly carried out there. It works in such a way as to

  part of the morphological analysis.

  The English data from the pre-layout is transferred to a section 3016 of morphological analysis at the same time as the information obtained during the pre-layout.

  form. Section 3016 Morphological Analysis Morphological Analysis

  only the English sentence chain and divides it into effect

  killing a consultation of a dictionary of words 3018, performs various arrangements such as a processing of unknown words, proper names, expressions of time, numbers, etc., and performs a processing of the sentence as a whole, such recognition of the redundant interrogation and recognition of the apposition. The rules for miorphological analysis are

  stored in an analysis rules file 3036.

  English data obtained after morpholo-

  gics are transferred to an analysis section I 3020 at the same time as the dictionary information obtained following the morphological analysis. Analysis section 3020 is a functional section that analyzes the structure of the sentence-related surface layer by applying grammatical rules to English data in order to

  to discover all the possibilities concerning the structure.

  The English data which is submitted for analysis in analysis section I 3020 is transferred to analysis section II 3022 together with the information obtained by the analysis. In this section, a solution is

  chosen thanks to the application of a structural description repo-

  on the result of the analysis relating to the structure

  ture of the surface layer carried out by section I of syntactic analysis, thanks to which a plausible tree of analysis of the English sentence is prepared so as to form its structure. These analysis rules are also

  stored in analysis rules file 3036.

  English data submitted for analysis is trans-

  fées in a section 3024 of structural transformation at the same time as the data relating to the analysis tree. The structure transformation section 3024 prepares a structure tree for

  a corresponding Japanese phrase from the struc-

  ture, which constitutes an intermediate structure of the sentence

  of English, in order to transform the sentence from Japanese into a struc-

  underlying ture compared to the Japanese language, from which one can easily obtain a Japanese translation sentence. Structure tree data indicating structure

  underlying in Japanese which have thus been transformed are deli-

  vrees to a translation training section 3026, where the sentence

  translated is formed. This is a functional section allowing

  so much to form a sentence in Japanese from the structure of

  the sentence structure tree in Japanese.

  Japanese sentence data that is prepared as a translated sentence, i.e. data

2 6 0 4 8 1 4

  translated, are sent to a post-shaping section 3030.

  The post-shaping section 3030 modifies the work data.

  duction by searching in a dictionary 3018 on the basis of the information used in the translation processing to finally arrive at a more natural Japanese sentence. Data relating to the Japanese sentence is transferred to an output section 3032 and is delivered by the latter in the form of a sentence translated into Japanese 3034. The output section 3032 comprises for example a printer, a display and, or ,

  a file memory device, for example a magnetic disc

  tick. The sequence of the translation processing operations is controlled by a control section 3038, which

  governs the control of the device as a whole.

  In this embodiment, the word dictionary 3018 stores dictionary data, or lexicological data, for English and Japanese words, but it provides various information such as connection relationships, namely coexistence relationships, meanings, plural form or singular form, parts of speech, etc., in addition to the words of

  vocabulary. In addition, the analysis rule file 3036 stores

  sine rule data for morphological analysis and syntactic analysis. The control section 3038 is connected to an operation display section 3040. The operation display section 3040 includes, for example, operation keys, such as, for example, translation indication keys or

  cursor, by which the operator can give various instructions

  the device of the invention, a display or an indicator visually indicating the English text entered, the sentence of

  Japanese which is the result of translation, inter-

  mediators such as dictionary information, or informa-

  lexicological information, etc., as well as various indications of-

  to the operator. Many of these operations indicating functions can be so constituted that they are included

260 4814

  in a keyboard, if one is arranged for the input section 3010, or else appear on a display, if one is arranged

in outlet section 3032.

  Reference is made to FIG. 17. It shows an example of detailed structures relating to the processing of numbers in the

  section 3016 of morphological analysis. The morpholo-

  gique 3016 naturally includes other functional analysis sections, but the sections which relate more directly to the understanding of the invention will be illustrated below. Morphological analysis is performed by the fact that it is given instruction to search in the dictionary starting with the head of the set of input characters and

  by proceeding successively in relation to the character set

  search key, or consultation key, and by the fact that the processing of lexicological information obtained from

  the part of the dictionary consultation section 3104 is exe

  cut according to a digital flag which will be described below.

  Morphological analysis section 3016

  includes an input processing section 3100 for receiving

  view and process input character set data from the pre-processing section 3014. The input processing section 3100 has an input character set buffer which receives as input the data of the English character set in the form of a code data, for example the ASCII code, and it stores temporarily

  the character set data.

  The input character set data which is temporarily stored in the input processing section 3100 is sent to a unit cutting, or isolation section 3102 which cuts the character set data d entry into dictionary lookup units, such as

  words. The unit cutting section 3102 is a functional section.

  used to distinguish the reference units from diction-

  which constitute a set of key characters of

  consultation at the time of the subsequent consultation of the diction-

  nary 3018 belonging to section 3104 of consultation of dictionary

  tionnaire. The dictionary reference delimiter, which is used to cut out the dictionary reference unit, is arranged at the position of an English character, a numeric bone, an apostrophe, a character other than the hyphen and period, as well as the apostrophe following a vacancy character. This delimiter is stored in a delimiter table 3108 to which we refer when cutting the dictionary reference unit in the section

cutting units 3102.

  The dictionary of words 3018 contains, very particular-

  the information associated with the search for the discovery unit

  page. As shown in the example of the entry item information in Figure 24, storage takes place for the item

  input relating to each of the diction reference units-

  nary, namely words, grammatical information such as parts of speech, as well as indications of distinction showing that a word represents a number, that is to say a flag of number,

  and digital value information indicating the digital value

  risk relating to a word indicating a number.

  As shown in the figure, both the singular and plural forms are described together with each

  entries from the dictionary of words 3018, each of which

tituates a particular entry.

  The number flag indicates that a word means a number when "1" is set for this purpose. Like other information,

  we find for example the character of accounting or non

  accounting of a name, identification relating to the transitivity or non-transitivity of a verb, translated words, etc. If we refer for example to "thousand (thousand)", since

  is a name representing a number, the associated digital flag

  cie is "1" and the numeric value "1000". In addition, if we refer to "thread (filet)", since it is a name, but not a number name, that is to say a number, the digital flag

is saved as "0".

  The recognition of a number is carried out using the

  associated digital flag in case it is a word rec-

  stored in the 3018 dictionary, for example as "one (one)" or "thousand (miLLe)". Even unregistered words, such as a succession of numeric characters such as "123", two successive groups of numbers separated by a numeric dot (in English, but in French, it is a numeric comma)

  for example a small number such as "10.2 (10.2)", and a success

  sion of numeric characters separated by a comma (in English, but in French, a dot or a space is used) such as for example "1,000,000 (1,000,000)", are also recognized

  like numbers. In the present description, the expression

  "numeric character" usually includes not only Arabic numerals, but also numbers expressed in full

like "thirteen (thirteen)".

  As shown in Figure 28, the dictionary 3018

  includes a table of currency symbols 3018a, where are stored

  various currency symbols, a notation symbol table 3018b where notation symbols such as the comma ",", the period ".", the space "", and a decimal punctuation mark table 3018c are recorded, where the decimal point "." (respectively, in French, a decimal point), the decimal point "," (in French, we leave a space or use a period), etc. The tables relating to notation symbols or decimal punctuation are provided to the extent that ","

  as a notation symbol or else use "." as a deci-

  poorly in Japanese and English, while we use main-

  either a space or "." for symbolic notation and we use

  read "," as a decimal point in other European languages such as French or German, so that the use of

  symbols differ depending on the language to be treated.

  Dictionary lookup section 3104 is a functional section which cuts dictionary information by looking up the word dictionary 3018 based on the set of lookup characters introduced from the lookup section.

2 6 0 4814

  cutting of units 3102 and which transfers this to the sections of

processing 3110, 3112 and 3116.

  The arrangement of the succession of numeric characters is carried out using the following two treatments. First

  place, if a word is recognized as a number, as follows

  above indicated, then, when the reference unit of diction-

  The next year to which we refer is also recognized as a number, they are collectively arranged so that they can be synthesized into a single number. The operation is repeated as long as numbers follow one another. For example, "30 thousand (thousand)" is changed to "30,000" and "1.5 million (1.5 million)" is changed to "1,500,000". Then when

  digital expression always continues by means of the interpo-

  sition of "and (and)", if all the digits to the left of "and" which correspond to each of the digits of numerical values indicated by the pointer to the right of "and" are "zeros", in

  the purpose of making sense of the digital expression they are syn-

  thetized in a single number. For example, "one hundred and thirty (hundred and thirty)" is synthesized in "130", while "30 thousand and two hundred (30 thousand and two hundred)" is synthesized in

"30,200".

  After the recognition of such a number, a necessary local analysis is also carried out. During this operation, a series of analysis units activated by the morpheme activation information for each of the analysis units is collectively arranged in a single analysis unit on the basis of a rule of local analysis. For example, a currency symbol and a numeric value, namely "41,000", are collectively arranged in the form "1,000 yen", while a numerical value and a unit, namely

  "1.5 km", are collectively arranged in the form "1.5 kilo-

meter (1.5 kilometers) ".

  These arrangements are made in processing sections 3110 to 3122. Processing section 3110 is a functional section used to collectively arrange a number with a currency symbol or a unit. The processing section 3112 is a functional section performing the numerical form of the number. In addition, processing section 3114 is a functional section for processing numbers linked by a hyphen. In addition, processing section 3116 is a functional section for processing numeric characters.

succeeding each other.

  With regard to the number obtained after arrangement with a currency symbol or a unit, the arrangement of the currency symbol and the numerical value into a single name is performed in the processing section 3118. In addition, the arrangement of the unit and numeric value in a single name is done in processing section 3120. In addition, in the case of a number that has been subjected to numeric processing, a number linked by one or several hyphens and a continuous number, the arrangement processing with the previous numeric value is

  performed in processing section 3122. Lexical information

  logic relating to the introduced character set which has final-

  If it has been processed using this process, it is stored in the sorted lexicological information buffer, namely the bet table.

  in lexicological information reserve 3124.

  The results of the morphological analysis are

  transferred from the lexicolo- information storage table

  see 3124 in section I of analysis 3020.

  The digital flag processing is performed in the sequence shown in Figures 19A and 19B. Data relating to the input character set is received in the input processing section 3100, where input processing is performed (3200). Next, the unit splitting section 3102 splits the set of input characters into a dictionary reference unit, or lexicological units for the purpose of

  consultation of the dictionary 3018 (3201). Section 3104 of con-

  Dictionary Lookup performs a search operation in dictionary 3018 in relation to this data (3203) and, if there is a dictionary entry field (3204), it examines the digital flag (3205). If the digital flag is not positioned, since the word is not a number, the dictionary information, or Lexicological information, is stored in the lexicological information reserve table (3124). If "1" has been set in the digital flag, the number is scanned,

  that is to say put in digital form, in the processing section

  3112 (3206), and the collective arrangement 3207 with the preceding numerical value is carried out in the processing section 3122. When these operations have been carried out up to the final position of a sentence indicated by the data of the set of input characters (3202), the collective arrangement (3209) with the

  currency symbol or unit is made in the processing section

  3118, 3120, and the result of the morphological analysis is finally delivered in section I of syntactic analysis

(3210).

  Following the transfer to the dictionary, when there is no entry item, in step 3204, and, if the element is linked using a hyphen (3212), the processing reserved for numbers linked by hyphens (3213) is performed in processing section 3114. If the first character is not a character linked by a hyphen, but a currency symbol (3214), the currency symbol only is placed in reserve in lexicological information storage table 3124 (3216) and the currency symbol is rejected (3217) from the reference unit of the

  dictionary. If the first character is not a mon-

  shut up (3214), the processing for numeric characters is

  next 3215 is performed in processing section 3116.

  The operation continues until the final position

(3202).

  The collective arrangement 3209 with the monetary signal or the unit is carried out in the processing section 3110 according to the processing flow diagram presented in FIG. 20. First, during the initial processing 3220, the head pointer used to the treatment is positioned on the head of the tampon. If the item indicated by the pointer is not a numeric value 3221, the pointer advances by degrees (3226). In the case where the character is a numerical value, but where there is not previously a monetary signal, nor, Thereafter, a unit, the pointer is also advanced by degrees (3222, 3224). Processing continues o5 until the final position of the dictionary reference unit

(3227).

  If the-character is a numeric value (3222), the sym-

  monetary bole and numeric value are collectively arranged in one name (3223). For example, the currency symbol and the numeric character "41.00" is arranged in a single name. Additionally, if the preceding character is not a currency symbol and the following character is a unit, the numeric value and the unit are collectively arranged in a single name (3225). For example,

  the numeric character and the unit "1.5 km (1.5 km)" are collectively

  vely arranged in a single name. Processing continues until the final position of the dictionary consultation unit

(3227).

* The processing intended for the numbers linked by a hyphen is carried out in the processing section 3114 according to the flow chart presented in FIGS. 21A and 21B. First, the dictionary reference unit linked by a hyphen is stored in the buffer during initial processing 3230. In addition, the numeric value "O" is put in reserve and the hyphen contained in the reference unit of the initial dictionary

is changed to a space.

  Then, the dictionary consultation unit is

  cut out (3231) so that the diction-

  nary (3235). Following consultation of the dictionary, if no entry item is found, i.e. if the word is not registered in the dictionary (3236), the reference unit of the linked dictionary by a hyphen is, as a whole, put in reserve, as a word not registered in the dictionary, in the

  lexicological information storage table 3124 (3237).

  Following consultation of the dictionary, if an entry item (3236) is obtained, it is checked whether the associated digital flag is "1" or not. If the digital flag is not

  "1" means that the character is not a numeric character

  risk, and the dictionary reference unit linked by a hyphen is, as a whole, put in reserve, as a word not recorded in the dictionary, in the reserve table

  Information from dictionary 3124 (3237).

  When the value "1" has been set in the digital flag for the dictionary entry item, the processing section 3012 scans the number based on the data from the

  entry section (3239). Then, the numerical value thus numer-

  laugh is added to a numeric value stored at the time considered (3240), and the result of the addition is stored

  (3241). So, for example "two (two)" of "twenty-two (thirty-

  two) "is added to" 3020 "coming from" twenty (twenty) "just before to form" 3022 ". Processing continues until the position

  final of the dictionary consultation unit (3232).

  When it has advanced by degrees to the final position, the flowchart proceeds to processing 3233 in step 3232, and the numerical value set aside is transformed into a numerical value for the entire reference unit of the linked dictionary with a hyphen. Then the collective arrangement 3207 of the value

  numeric with the previous numeric value is performed.

  We will now explain the treatment 3215 of the character-

  next digital tere, which is executed in the processing section

  ment 3116, in relation to FIGS. 22A and 22B. In this organ-

  nigram, the symbol "=" means a substitution. First of all, an initialization 3250 is carried out, during which the numerical value placed in reserve, ie "val-save", is set to "O", the parameter "i" is set to "1", and the pointer p is positioned on the top of the character set of the dictionary reference unit. Then it is checked whether the character p indicated by the

  pointer p is a numeric character (3251), a notar character

  tion (3252) or a decimal point (which is, in French, a decimal point) (3253) and, if it does not match any of them, the entire character set is stored, as an unregistered word

  in the dictionary, in The information storage table

  lexicological section 3124 (3255). If it is a decimal point (which is a decimal point in French) (3253), the parameter "i" is multiplied by 10 (3254), and step 3258 is executed. In step 3258, the numerical value "num" (* p) of the character * p is added to the numerical value placed in reserve "val-save" in order to prepare a new numerical value put in reserve. The numerical value

  "num" (p) is a value by which the character (C p) is consi-

derived as a numeric value.

  In step 3251 or 3252, if the character is a numeric character or a notation character, step 3257 is executed. In step 3257, the numerical value set aside "val-save" is multiplied by 10, and the numerical value "num" (* p) of the character *

  p is added to the result to prepare a new numerical value

risk put in reserve.

  After these operations, the pointer advances by degrees (3259) and the process repeats to the final position of the dictionary reference unit (3260). If it is in the final position of the character set, the numerical value relative to the entire character set is placed as a reserved numeric value (3261), and a collective arrangement

  3207 with the previous numerical value is executed in the sec-

  processing operation 3122. By this operation, numeric characters

  successive risks such as "1,000.5" are analyzed as corresponding

  corresponding to the numerical value "1000.5 (1000.5)".

  Collective arrangement 3207 with the previous numerical value is carried out as follows in the section of

  processing 3122. First, the pointer to the diction-

  naire is positioned on a position preceding the dictionary reference unit (3270). If there is nothing at this position, it means that the first position of the reserve table indicates the numerical value, after which the numerical value

  of the dictionary reference unit that is being used

  dation is recorded in the diction reserve table-

  Mayor 3124 (3284). The recorded position is The position which follows the position currently indicated for the pointer P. If, at step 3271, a word is present at the previous position, if the entry item indicated by the pointer p is not not "and (and)" (3272), and if the pointer p does not indicate the numerical value (3273), the numerical value relative to the reference unit of the dictionary in use is recorded following the position in use which is indicated by the pointer p in the reserve table of the dictionary 3124 (3284). For example, in the example "to him two ...", "two (two)" is re-saved as the numeric value "2".

  In step 3273, if the pointer p indicates a numerical value

  rique, the numerical value p - v relative to the entry item indicated by the pointer p is multiplied by the numerical value "v-now" which is relative to the dictionary reference unit in use in order to form a new numerical value p - v relating to the entry item indicated by the pointer p (3274). In the case of "two thousand (two thousand)" for example, the operation "2 x 1,000 = 2,000" is executed in order to combine the expression "two thousand (two thousand)" into one. Then, the end position of the dictionary reference unit in use is set as the end position for the item

  input of the pointer p, that is to say the position "p-end" (3282).

  In step 3272, if the entry item indicated by the pointer p is "and (and)", the pointer p is transferred to the dictionary consultation unit located before (3275). If it is not at the final position (3276) and if it is a numeric value (3277), the numerical value "vnow" of the dictionary reference unit in use is produced and is rounded to the most significant figure, to be set to a value "vl". If the numeric value "v-now" of the dictionary reference unit in use is for example "8", "8.1 (8.1)", "98" or "11",

  the respective value of vl is "10", "10", "100" or else "100".

2 6 0 4 8 1 4

  Then, it is checked if the excess, or remainder, obtained by dividing by vl the numerical value "p-v" for the entry item indicated by the pointer p, what we will call mod (p-v,

  vl), is or is not "O". If it is not "0", the pointer p is incremented

  mented (3283), and the numerical value relating to the unit of reference

  The current dictionary reference is saved at the current position indicated by the pointer p in the dictionary reserve table 3124 (3284). In the case of "I and two" by

  example, "two" is re-stored as the numeric value

than "2".

  If the excess is "0" in step 3279, the numerical value "v-now" relative to the dictionary reference unit in use is added to the numerical value "p - v" relative to the entry item indicated by the pointer p to form a new numeric value "p - v" relative to the entry indicated by the pointer p (3280). In the case of "two thousand and two (two thousand and two)" for example, "two thousand (two thousand)" is already arranged in the form "2000" at this point. Then it is added

  with the "2" of "two" by the addition 3200 so as to arrange collectively

  the whole part in the form "2002". Then the informa-

  tion "and (and)" indicated by the pointer p + l is rejected from the

  information storage table 3124 (3281) and the organization

gram goes to step 3282.

  We will now present explanations in relation

  with an example. When the dictionary consultation is carried out

  killed for example for the set of input characters "to him two thousand and twenty-two ... (to him two thousand and twenty-two ...)", as shown in figure 25, the information of the dictionary entry item is entered in the dictionary information storage table 3124, as shown in FIG. 26A. For "him (him)" for example, the starting position is "4", the final position is "6", and the speech part is a pronoun. During digital processing, it is first decided with regard to "two (two)" that the digital flag is "1"

  (3205) and that the numerical value is therefore "2". Since the character-

  tère preceding "two (deux)" in this set of characters is not a numeric value, it is stored directly in the

table 3124 (3206, 3208, 3284).

  Then, the pointer is incremented in order to pass to S05 processing of "thousand (thousand)". The digital flag is "1" and the numeric value is "1000" (3205, 3206). In addition, since the numeric value of the previous character is "2" (3207, 3273), the 2x1000 multiplication is executed (3274), and its result is stored in table 3124 (see Figure 26B). Then, for

  the next word "and (and)", the dictionary information is tempo-

  usually stored without change in table 3124 (see

Figure 26C).

  The pointer advances again to process "twenty-two (twenty-two)". Since the words are linked by a hyphen and form a word which is not found in the entry fields of the dictionary (3212), the operation "20 + 2 = 22" is executed by the processing 3213 relating to numbers linked by hyphens (3237, 3239 to 3241). Since the previous word is "and (and)" (3272) and the numerical value before it is "2000" (3277), the numerical value "22" is rounded to the most significant digit so as to form " 100 "(3278) and the division operation 3273 is executed. Since the excess, or remainder, is "0", the addition 3280 between "2000" and "22" is performed. The information relating to "and (and)" is eliminated (3282) from the reserve table 3124, and the result of the addition "2022" is reserved as a numerical value in table 3124, whereby "two thousand and twenty-two (two thousand and twenty-two)" is recognized as "2022". Thus, the collective arrangement with the

  previous numeric value has been executed (3207).

  We will present another example. As shown in Figure 27, a grammatical analysis is performed for the input character set "you said $ 1,000.5 thousand was ... (you said 1000.5 thousand dollars were ...)". The word "p1,000.5"

  is not saved in the 3018 dictionary. The first character

  tère is a currency symbol "$" which can be recognized as a currency symbol in the dictionary entry fields. This is stored independently in reserve table 3124

(3214, 3216, Figure 29A).

  Then, "1,000.5" is transformed into a numerical value "1000.5" by Processing 3215 intended for the successive numeric characters. Since the preceding character is the symbol "t" and not a numerical value, the numerical value is recorded without

  modification (3280-3273, 3284, Figure 29B).

  The next word "thousand (thousand)" is a number and its numeric value is "1000". Since the preceding character is a numeric value (3272, 3273), the calculation "1000.5x1000 = 1000500"

(3274) is executed (FIG. 29C).

  After the dictionary search has been completed, the data in the reserve table

  Information files from dictionary 3174 are examined successively.

  Since the currency symbol "" is present just before the numeric value "1000500", both are collectively arranged and "$ 1000500" is produced to form an entry with a unique name (3209, 32213223, Figure 29D) (it looks like in French

1000 50 00).

  The fourth embodiment of the invention will be explained. FIG. 30 shows the fourth embodiment of the language analyzer according to the invention, applied to a device

  English-Japanese machine translation.

  This embodiment has an input processing section 4014, and the input processing section 4014 receives data from an input device 4012. The input device 4012 for example includes a keyboard having character keys, for example alphanumeric keys, and function keys, an optical character reading device making it possible to read English text recorded on a

  sheet of paper, and a reader, for example for magnetic disc

  tick. The input processing section 4014 has an input character set buffer 4014a and stores the input English phrase provided by the input device 4012 in the

  input character set buffer 4014a. The processing section

  input header 4014 reads an input phrase stored in the character set buffer 4014a and outputs it to a unit cutting section 4016. The unit cutting section 4016 is a functional section which cuts units lexicographic reference code for the input sentence delivered by the input processing section 4014. A delimiter storage table 4018 stores delimiters such as spaces and commas. The unit splitting section 4016 reads delimiters from the delimiter table '4018 and divides the input sentence sent by the input processing section 4014 into character sets, which are taken as units in relation to the search in a reference dictionary 4020, by separating the sentence at the level of the parts where the delimiters are present. Divided character sets are provided to a search section in

  dictionary, or dictionary lookup, 4022.

  The dictionary consultation section 4022 consults the reference dictionary 4020 in relation to the input sentence divided into lexicographic reference units provided by the unit splitting section 4016. The dictionary

  reference 4020 contains entry items, or entries, relating to

  from sets of English characters, the "speech parts" formed by them, information about their nature, etc., for example, as shown in Figure 31. The reference dictionary 4020 also stores, in addition proper names as shown in Figure 31, sets of characters for

  other parts of speech, for example verbs and adjec-

  tifs. Recording proper names as parts of speech

  aims at an application linked to the processing of registered proper names

  very described below, but does not try to express the usual grammatical value of proper names. In addition, information on the nature indicates what the proper name represents, which does not always reduce only to an entity, since, as indicated below, a proper name can represent several entities according to

the use made of it.

  The dictionary consultation section 4022 consults the reference dictionary 4020 relative to the set of characters divided into lexicographic reference units and, if the character set is a proper name, it delivers it to the section 4024 for processing names o a processing for proper names, which is described below, is carried out. If it is not a proper name, the assembly is supplied to a processing section 4036 and is put in reserve in a setting table.

  vocabulary information store 4036a of the processing section

4036.

  The proper names processing section 4024 comprises a section 4026 for processing the end of the preceding sentence, a section 4028 for processing the previous proper name, a section

  4030 proper name processing itself, a processing section

  4032 applying to the previous proper name and to the proper name itself, and a section 4034 producing information of

missing feature.

  Previous end-of-sentence processing section 4026

  decides whether the character set preceding the character set

  search terms provided by the lookup section of the 4022 dictionary may or may not be at the end of the sentence and, if all of

  preceding characters is found at the end of the sentence, it trans-

  forms the capital letter at the top of the set of characters to be treated in lowercase, delivers it to the section of

  4022 dictionary lookup and brings up the lookup section

  dictionary 4022 to consult the reference dictionary 4020 again. The character set not found even after the

  second search is deemed to be an unregistered proper name

  tré, which is sent to processing section 4036, then is stored in the information storage table of the

  dictionary 4036a. In addition, in the case where the character set

  tères precedent is not found at the end of the sentence, it is delivered to the processing section 4036 under the proper name for which there is no particularity information and is recorded in the reserve table d dictionary information 4036a,

as described below.

  The previous proper name processing section 4028 performs an analysis on the particularity information for the preceding character set provided by the preceding end of sentence processing section 4026 and delivers the result to the proper name processing section 4030. himself. The proper name processing section 4030 examines the particularity information of the proper name to be analyzed and, as described below, whether the particularity information of the proper name as well as the previous proper name is not recorded, it analyzes the proper noun and the previous proper noun in a collective way using the information recorded for the other words and puts it in reserve in the lexicographic information storage table 4036a of the section treatment

4036.

  The processing section 4032 of the previous proper name and

  of the proper name itself examines the common part of the information-

  particularity of the proper name and the previous proper name, analyzes these proper names from the common part, delivers the result to the processing section 4036, then

  stores in the lexico- information storage table

  graph 4036a of processing section 4036.

  Section 4034 for the production of particular information

  missing rity provides particularity information of a nom-

  pre which, after having been read in the lexicographic information reserve table 4036a, was sent via the unit cutting section 4016 to the consultation section of the dictionary 4022 and was found not to have the specific information when consulting the reference dictionary 4020 of the dictionary consultation section 4022. Since it is true that a proper name has different types of features depending on the situation in which

  uses, all information of a nature considered necessary

  are planned. For example, we provide "person, place,

  group, and others ". After having given the proper name of an information

  particularity, section 4034 for producing missing particularity information sends the data to the section

  4036 and stores them in the information storage table

  lexicographic section 4036a of processing section 4036.

  The processing section 4036, comprising the lexicographic information storage table 4036a, stores the data sent by the processing section 4032 of the proper name.

  previous and proper name itself, section 4034 of produc-

  tion of missing particularity information or the section of con-

  sultation 4022 so that they are placed in the lexicographic information storage table 4036a, after which it reads the stored data and delivers them to the analysis section 4038. The analysis section 4038 performs the analysis relating to the input sentence after it has been subjected to morphological analysis and read in the

  lexicographic information storage table 4036a.

  We will now explain how this arrangement works.

  refer to the flowchart in Figure 32.

  First, an English entry sentence from the

  input device 4032 is read and placed in the processing section

  entry 4014 (4100). The input sentence read and placed in the input processing section 4014 is then loaded into the input character set buffer 4014a. The input sentence loaded into the input character set buffer 4014a and read to be delivered to the unit splitting section

4016.

  When the input sentence has been entered, the unit splitting section 4016 reads the delimiters from the table.

  delimiters 4018 in order to proceed to the cutting of the reference units

  dictionary reference (4102). So the character sets

  constituting the input sentence entered are divided successively-

  ment, from the head of the sets of characters, into sets of consultation key characters, intended to constitute the units allowing the consultation of the reference dictionary 4020, by the fact that they are divided at the level of the parts o are found delimiters such as a space or a colon. This section determines whether the dictionary reference units obtained by the division, i.e. the sets of lookup key characters, have expired or not (4104) and, if there are still sets of characters. consultation key, it sends the character set of

  consultation in the consultation section of the 4022 dictionary.

  When a set of consultation key characters

  tion is sent to the consultation section of the dictionary 4022, the latter consults the reference dictionary 4020 in order to

  processing the lookup key character set (4106).

  The section determines whether or not the set of consultation key characters is present in the input items of the reference dictionary 4020, as shown in FIG. 31,

  (4108) and, if there is an entry section, it reads the information

  tion relating to the part of speech which is stored in the reference dictionary 4020 and decides whether the set of characters of consultation key is or not a proper name

(4110).

  If the consultation line character set does not

  is not a proper name, the consultation section of the diction-

  naire 4022 sends the data read from the reference dictionary

  rence 4020 to processing section 4036 and records them in the lexicographic information storage table 4036a

  (4112). When the data is stored in the storage table

  gination of lexicographic information 4036a, the input data indicating that the data have been stored in the unit cutting section 4016 and the data relating to the set of

  consultation key characters set aside just before-

  vant are provided by the processing section 4036. Then, the flowchart returns to step 4102 and the dictionary reference unit is again cut out in the cutting section

4016 units.

  In step 4110, if the set of consultation key characters is a proper name, the dictionary consultation section 4022 sends the data relating to the proper name read in the reference dictionary 4020 (hereinafter simply called the name proper) to the proper name processing section 4024, along with the data for the previous lookup key character set provided by the layout table

  vocabulary information store 4036a of the processing section

  4036 via the unit splitting section 4016 to the dictionary consultation section 4022, and the processing of the proper name registered in the dictionary is carried out in the section

  proper name processing 4024 (4124).

  We will now explain the processing of the proper name recorded in the dictionary in relation to the flowchart

shown in Figure 33.

2 6 0 4 8 1 4

  The data sent from the dictionary lookup section 4022 to the proper noun processing section 4024 is provided by means of the preceding end-of-sentence processing section 4026 to the previous proper noun processing section 4028. During processing of the proper name registered in The dictionary, the previous end-of-sentence processing section

4026 does not work.

  The previous proper name processing section 4028 determines whether or not the entry item preceding the proper name is a proper name not registered in the reference dictionary 4020, that is to say whether or not it is applied to it. the processing intended for proper names not registered in the dictionary,

  described below, (4100). If it is an unregistered proper name

  very, the processing section decides that all the part of the proper name and the previous unregistered proper name is like a proper name having the proper nature information (4202), sends the data to the processing section 4036 and the stores in the dictionary information storage table 4036a

(4218).

  If the preceding proper name processing section 4028 decides, at step 4200, that the entry item preceding the proper name is not a registered proper name, it determines whether the entry item preceding the proper name is or not a proper name registered in the reference dictionary 4020 (4204). If the entry item which just precedes the proper proper name is a registered proper name, it is determined whether the particularity information of the preceding proper name is or is not unknown, that is to say whether or not it is, not registered in the reference dictionary 4020

(4206).

  In the case where the specific information of the previous proper name is unknown, the flowchart goes to step 4202 and decides that the whole part of the proper name itself and of the name

  own just previous as being a proper name which has an inform-

  special feature (4202), and the processing section 4028 of the previous proper name sends the data to the processing section 4036. The data sent to the processing section 4036 is

  recorded in the Lexico- information storage table

graphic 4036a (4218).

  If the processing section of the previous proper name 4028 determines in step 4206 that the information of particularity of the previous proper name is not unknown, that is to say if it is recorded in the reference dictionary 4020, the data

  are sent from section 4028 of the previous proper name processing

  tooth in section 4030 proper name processing itself. The proper name processing section 4030 itself determines whether

  information of the proper name is unknown or not (4208).

  If the particularity information of the proper name is unknown, section 4030 of processing the proper name itself determines the entire part of the proper name itself and, of the proper name

  just preceding as being a proper name which has information

  specific feature of the proper name (4210), and sends the data to the processing section 4036. The data sent to the processing section 4036 is recorded in the reserve table

  for lexicographic information 4036a.

  If the proper name processing section 4030 in itself

  even determines that the particularity information of the proper name itself

  even is not unknown, i.e. if it is registered in the reference dictionary 4020, the name processing section

  own 4030 sends the data to the processing section

  ment 4032 of the previous proper name and of the proper name itself. The processing section 4032 of the previous proper name and of the proper name itself determines whether or not there is a common characteristic in the information of nature between the proper name itself and the just preceding proper name (4212) and , if there is a common particularity, it determines the entire part of the proper name itself and of the proper name just preceding as being a proper name which has the particularity information of the common part (4214) and sends the data to the processing section 4036. The data sent to the

2 604814

  processing section 4036 are recorded in the stake table

  in lexicographic information reserve 4036a (4218).

  If there is nothing in common between the information of particularity of the proper name itself and of the proper name which immediately precedes it, the processing section determines that the proper name is a proper name having specific information found in the dictionary 4020 which differs from the proper name immediately preceding it (4216) and sends the data to the processing section 4036. The data sent to the processing section 4036 are recorded in the reserve table

  of lexicographic information 4036a (4218).

  We return to FIG. 32. In step 4108, if there is no set of consultation key characters in the entry item of the reference dictionary 4020, it is determined whether the

  first character in the set of consultation key characters

  tation is or is not a capital letter (4116) and, if it is not

  not a capital letter, the consultation section of the diction-

  naire 4022 determines the character set of consultation key

  as an unregistered word, send it to processing section 4036 to save it to the reserve table

  for lexicographic information 4036a (4118).

  If the first character is a capital letter, the

  data relating to the set of consultation key characters

  tation are sent to the proper name processing section together with the data relating to the character set of

  previous consultation key from the consultation section

  dictionary 4022, the processing for the proper name not registered in the dictionary being carried out in this section

proper name processing (4120). We will now explain the treatment of unregistered proper names

  in the dictionary, in relation to FIG. 34. The data relating to the set of characters of

  consultation key are sent to processing section 4026

  preceding end of sentence together with the data

  relating to the set of pre-consultation key characters

  yield, and the preceding end-of-sentence processing section 4026 determines whether the end of the preceding entry item is or

  not likely to constitute the end of the sentence (4300). The deci-

  sion taken on the possibility that it constitutes or not the end of the sentence is taken according to the determination of the fact that the end

  of the previous entry item may or may not constitute

  substitute the end of the sentence, for example that the possible presence of a separate dot ".", etc.

  If the end of the previous entry item is suspected

  tible of constituting the end of the sentence, the data is sent from section 4026 of processing of the end of the preceding sentence to section 4028 of processing of the previous proper name, and the section 4028 of processing of the previous proper name determines the item

  previous entry as the end of the sentence (4302), trans-

  forms the first character in the lookup key character set in lowercase and sends it to the section of

consultation of the dictionary 4022.

  The consultation section of the dictionary 4022 consults the reference dictionary 4020 relative to the set of consultation key characters which has been transformed into a lowercase character (4304) and determines whether or not it is an item d entry from reference dictionary 4020 (4306). If it is an entry item, the consultation section of the dictionary 4022 sends the data found in the reference dictionary 4020 to the processing section 4036 and stores it in the information storage table. lexicographic

  4036a (4308). If it is not an entry item, the sec-

  consultation of the 4022 dictionary gives the first character

  tère of the set of consultation characters in the form of a capital letter and sends this one, under proper name

  not registered, to processing section 4036 for registration

  registration in the lexico- information storage table

graphic 4036a (4310).

26 0 4 8 1 4

  In step 4300, if the preceding sentence processing section 4026 determines that the end of the previous entry item is not likely to constitute the end of the sentence, the data is sent from section 4026 of previous end-of-sentence processing in Section 4028 of previous proper noun processing, and section 4028 of previous proper noun processing determines the previous entry item as not being the end of the sentence (4312). Data is sent from section 4028

  processing of the previous proper name in section 4030 of the treaty

  proper name itself, and proper name processing section 4030 determines the lookup key character set to be a proper name whose information

* particularity is unknown (4314).

  The proper name processing section 4030 itself returns the data to the previous proper name processing section 4028, and the previous proper name processing section 4028 performs the processing for the proper names registered in the dictionary (4316). The treatment of proper names registered in the dictionary is the same as that shown in the figure

33.

  We return again to Figure 32. If the reference unit

  rence of cut dictionary is at the end at the time of step 4104, the consultation section of the dictionary 4022 sends a signal indicating to the section 4034 of production of information of missing particularity that the reference unit is at the end, and the missing feature information section 4034 reads The information recorded in the layout table

  vocabulary information store 4036a of the processing section

  4036 and gives the proper name of the particularity information

missing (4122).

  We will now explain the operation which consists in endowing the proper name with the information of missing particularity,

using Figure 35.

  In section 4034 of production of the missing particle information, a pointer is first positioned on the top

2 6 0 4 8 1 4

  data in the Lexico- information storage table

  graphic 4036a (4400). Thus, the pointer is positioned on the entry heading of the sentence supplied divided into entry headings, which are respectively provided with information from the

  searches the 4020 reference dictionary.

  Then, it is determined whether or not the entry item indicated by the pointer is a proper name (4402) and, if it is a proper name, it is determined whether the particular information of the proper name is unknown or not (4404). if it is not a proper name, the flowchart goes to step 4408 and the pointer advances

in the next entry section.

  In step 4404, if the particularity information of the proper name is unknown, the missing particularity information is supplied (4406). Regarding the production of the missing particularity information, the latter is supplied to a proper name whose specificity information is unknown as shown in the lower part of figure 36. For example, the proper name " Johnson "whose particularity information is

  unknown has all kinds of specific information

  rity, that is to say "person, place, group, and others". By providing-

  sant all the types of information of particularity with proper name, one can leave the place allowing to analyze te propername according to any of the various particularities during the

further syntactic analysis.

  At step 4404, if the specific information of the proper name is not unknown, the flowchart goes to step 4408

  and the pointer advances to a next entry item.

  It is determined whether the entry item indicated by the pointer is at the end or not (4408) and, if it is not at the end, the flowchart returns to step 4402, and it is then determined whether or not the next entry item is a proper name. If the entry item indicated by the pointer is located

  at the end, the production of the missing information is finished.

  Once completed, the production of information from

  missing feature for the proper name, the data recorded

  entered in the lexicogra information storage table

  4036a are delivered by the processing section 4036 to the

  S05 syntactic analysis section (4124), whereby MorphoLo- analysis

  gic in this embodiment is completed.

  We will now give the explanations concerning the operation of the above-mentioned device in relation to

an example of a sentence provided.

  The explanation will be given in case the input phrase "In Tokyo Station Mr. Walter met Johnson (In Tokyo Station Mr. Walter met Johnson)" is provided. First, an input processing (4100) allowing to read the input sentence and

  placing it in the input processing section 4014 is done

  kill. Then, the dictionary reference unit is split (4102) and the input sentence is divided into words respectively by spaces. The reference dictionary 4020 is first consulted for "In (In)" (4106). Since there is no entry item for "In" in this precise form in the reference dictionary 4020, the step leads to the processing of proper names not recorded in the dictionary. However, since the previous part is at the top of the set of input characters, it is considered likely to constitute the end of the sentence, and is the subject of a search, in the form "in" after

  conversion from I to id in the 4020 reference dictionary.

  Since it is not a proper name (4110), the data sought

  in the reference dictionary 4020 are recorded in the lexicographic information storage table 4036

(4112).

  Then, the reference dictionary 4020 is consulted for "Tokyo" (4106). Since there is no entry heading in the 4020 reference dictionary for "Tokyo" (4108) and the first character is an uppercase letter (4116), a processing relating to proper names not registered in the dictionary East

  performed (4120). Then, the flowchart goes to figure 34.

  Since the previous part is "In (Dans)" and it is not a part likely to constitute the end of the sentence (4300), it is determined that "In" is not the end of the sentence (4312),

  then "Tokyo" is recognized as a proper name whose informa-

  special feature is unknown (4314), and processing for names

  own saved in Dictionary is performed (4316).

  The flowchart goes to figure 33. Since "In" which occupies the previous part is not an unregistered proper name

  (4200) and is not a registered proper name (4204), it is registered

  registered as a proper name possessing specific information by itself, i.e. as a proper name whose informationd

particularity is unknown (4216).

  We then return to FIG. 32. The search in the reference dictionary 4020 is then carried out for "Station

  (Station) "(4106). Since there is an entry section in the dictionary

  reference tenant 4020 for "Station" (4108) and since it is a proper name (4110), processing for proper names

  saved in The dictionary is performed (4114). The organization

  gram goes to figure 33. Since "Tokyo", from the previous part

  dente, is an unregistered proper name (4200), the whole "Tokyo Station (Tokyo Station)" is recorded as being a proper name which, globally, presents the specific information "place, group" which is that of "Station" (4202). Then, the reference dictionary 4020 is consulted, as shown in FIG. 32, for "Mr." (4106). Since "Mr." has an entry item in the reference dictionary 4020 and that it is a proper name (4110). The processing reserved for registered proper names of the dictionary is carried out (4114). Next, the flow chart proceeds to Figure 33. Since "Station" in the previous section is not an unregistered proper name (4200) but a registered proper name (4204), it has been determined that the specific information

  (place, group) is known (4206). Since "Mr." has information

  particularity "person" and is not unknown 44208), it is determined whether or not it is a common part for information of a nature between "Station" from the previous part and "Mr." (4212). Since "Station" means "place, group", while "Mr." means "person" and there is no common part between them, "Mr." is registered alone as a proper name having

  particularity information for a "person" (4216).

  So the flowchart comes back to figure 32, and the dic-

  reference shareholder 4220 is consulted for "Walter" (4106).

  Since there is an entry item for "Walter" in the reference dictionary 4020 (4108) and it is a proper name (4110), processing for proper names registered in the dictionary is performed ( 4114). The flowchart goes to Figure 33. Since "Mr." of the previous part is not an unregistered proper name (4200) and is a registered proper name (4204), and since the information of particularity, that of a person, is not unknown (4206) and that the specific information of "Walter", which presents itself as "person, place, group" is not

  no longer unknown (4208), it is determined that there is a com-

  mune for specific information (4112). Since the "person" part is present for specific information, "Mr.Walter" is registered collectively as a compound name which has

  the specific information of a "person" (4214).

  Then, "met" is looked up in the reference dictionary

  rence 4020 and, since there is an entry item (4108) and it is not a proper name (4110), the data found in the reference dictionary 4020 is recorded in the bet table in lexicographic information reserve 4036a

(4112).

  Further on, "Johnson" is searched for in the reference dictionary 4020 (4106). Since there is no entry rubric

  for "Johnson" (4108) and that the first character a letter uppercase-

  cule (4116), the proper names not registered in the dictionary are processed (4120). Next, the flow chart goes to Figure 34. Since "met (meet)" from the previous section is unlikely to constitute the end of the sentence (4300), it is determined that "met" is not the end of the sentence (4312), "Johnson" is considered as a proper name having information of unknown particularity (4314), and the processing of proper names registered in the dictionary is carried out (4316). Then the flowchart goes to Figure 33. Since "met" from The previous section is neither an unregistered proper name (4200), nor a registered proper name (4204), "Johnson" is recorded alone as a

  proper name whose particularity information is unknown.

  Once the previous treatments have been carried out, the information

  the missing feature is provided in the proper name of the

  as shown in figure 35.

  The pointer is positioned at "In (In)" at the upper dictionary reference unit (4400). Since it is not

  not a proper name (4402), the pointer advances (4408) and posi-

  operates on "Tokyo Station". Since "Tokyo Station" is a proper name (4402) and the information of particularity is not unknown, because the whole part "Tokyo Station" was recognized as being a place or a group in the previous treatment of proper names registered (4404), the

  pointer advances (4408) and is positioned on "Mr. WaLter".

  Since "Mr. Walter" is also a proper name (4402) and the information of particularity is not unknown, (4404), the pointer advances (4408). Since "met (meet)" is not a proper name (4402), the pointer advances (4408). Since "Johnson" is a proper name (4402) and the particularity information is unknown (4404), missing particularity information is provided (4406) and "Johnson" is provided with specificity information "person, location, group and others ", as shown on

Figure 36.

  As described above in this embodiment, an input English sentence is divided into sets of lookup key characters, which are then searched in the reference dictionary 4020 and, if there is one

  an entry item for proper name in the reference dictionary

  rence 4020, processing for registered proper names is carried out. In the processing of registered proper names, account is taken of the previous set of consultation key characters and, if the previous set of consultation key characters are a proper name, the specific information of the set of Previous lookup key characters and proper noun considered are examined. If there is no information from

  particularity, other particularity information is pro-

  picks, while, if any particularity information is present for the set and the name, the common part is considered as the particularity information for these names

  clean. Therefore, it is possible to endow appropriately

  nable a proper name not having information of particularity of an information of appropriate particularity, as well as to restrict in a suitable way the information of particularity provided to a information of particularity more appropriate. This allows more effective analysis during parsing

  later and leads to an appropriate translation.

  In addition, for the character set not registered in the reference dictionary 4020, if the first character is an uppercase letter and the preceding character set has been determined to be the end of the sentence, since the letter

  uppercase is transformed into a lowercase letter and that the dic-

  reference 4020 is consulted again, it is also possible to consult for the set of characters at the head of the sentence in the reference dictionary 4020. In addition, in the case where a set of characters begins by a capital letter appears in a part other than the head of the sentence, it is determined as being a proper name and the information of particularity of this proper name is provided by means of a proper name having information of particularity registered if it exists before and after it. Therefore, it is possible to analyze to some extent a

  proper name not registered in the reference dictionary 4020.

  In addition, since a proper name which does not have a

  particularity information is provided with all information

  useful special features and that the particu-

  unnecessary arity are removed when processing the word it is

26048 14

  possible to analyze a proper name whose particular information

  rity is unknown or an unregistered proper name.

  In addition, since several information of particularity is provided to a specific proper name and that information of

  appropriate features are selected based on the informa-

  particularity of the proper name before or after

  it is possible to choose particular information

  the appropriateity in the case of the analysis of a proper noun posses-

  dant several kinds of information of particularity depending on the relationship that exists with other words placed before and after it, which allows an effective analysis of the sentence provided.

  We will now explain the fifth mode of realization

sation of the invention.

  FIG. 38 illustrates the general structure of the fifth embodiment of the language analyzer according to the invention,

  applied to an angLais- machine translation device

  Japanese. This embodiment includes an input section

  5010, and an English text 5012 to translate into Japanese is intro-

  duit by this section. The input section 5010 may for example comprise a keyboard which has character keys, such as alphanumeric keys, or functional keys, an optical character reading device (OCR) making it possible to read the English text recorded on a sheet of paper, and, or, a file memory device for reading English text saved on a memory medium such as a

magnetic support.

  English text provided by entry section 5010

  is read in a pre-shaping section 5014 o a pre-

  processing is carried out for translation. In this case, recognition of sentences and processing of unknown words is

  mainly performed. This function is part of the analysis.

morphological lysis.

  The pre-formatted English data is transferred to a morphological analysis section 5016 at the same time as the information obtained during the pre-formatting. Section 5016

  of morphological analysis divides the data into sentences in search

  song in a 5018 word dictionary, morphologically analyzes the input string, performs the processing of unknown words, proper names, as well as various arrangements such as the expression of

  time and numbers, and also performs the related processing

  the entire sentence such as recognition of apposi-

  repetitive or redundant questions and interrogations (which one designates in English grammars under the name "tag questions". The rules of morphological analysis are stored

  in an analysis rules file 5036.

  The English data coming after the morpholo-

  gics are transferred to a syntax analysis section I 5020 at the same time as the dictionary information obtained during the morphological analysis. Parsing section 5020 is a functional section which analyzes the surface structure of the sentence by applying grammar rules to English data

  and discovering all the structural possibilities.

  The English data submitted for parsing in Section I parsing 5020 is sent, together

  time that the information analyzed, in a section II of syn-

  taxic 5022, where a solution is chosen from the results of the syntactic analysis carried out in relation to the structure of

  surface by section I syntactic analysis, by means of the appli-

  cation of a structural description. A plausible analysis tree

  of the English sentence is therefore prepared and its structure is formed.

  Parsing rules are also stored in the

analysis rule file 5036.

  After being subjected to parsing, the English data is transferred, in the form of analysis tree data, to a structure transformation section 5024. In the structure transformation section 5024, a structural tree of the corresponding Japanese sentence is prepared from the structural tree which is an intermediate structure of the English sentence and is converted into an underlying Japanese structure from from which the Japanese phrase can be

obtained with ease.

  The data relating to the structural tree indicating the underlying Japanese structure having thus been subjected to the structural transformation are sent to a translation training section 5026, where a translated sentence is produced. This is a functional section that produces a Japanese sentence from the tree structure of the Japanese structureL tree.

  The Japanese data obtained as a result of the training

  tion of the translation, i.e. the data associated with the

2 6 0 4 8 1 4

  translated sentence, are sent to a post-formatting section 5030. The post-formatting section 5030 modifies the translated data by searching in a dictionary 5018 using the information used in the translation processing in order to arrive at a more natural Japanese sentence. Data for the Japanese phrase is transferred to an output section 5032 and output, as a translated Japanese phrase 5034,

  via the 5032 outlet section. The 5032 outlet section can

  wear for example a printer, a display, and a device

  from memory to file, for example a magnetic disk.

  The progress of the series of translation operations is controlled by a control section 5038 which governs the

  control of the device as a whole.

  The word dictionary 5018 stores dictionary data, or lexicographic data, relative to English and Japanese words, and various information is established there, such as connection relationships, namely coexistence relationships, as well as meanings, singular and plural forms, parts of speech, etc., in addition to vocabulary. In addition, analysis rule file 5036 stores data

  rules for Morphological analysis and syntactic analysis.

  The control section 5038 is connected to a section

  5040 display of operations. The operation display section

  tions 5040 has operation keys by which an operation

  can provide various indications to the device, for example a translation indication key, a cursor key, etc., and a display or indicator which visually displays the English text supplied, the Japanese text which is the result translation, intermediate data such as lexicological information, various indications intended for the operator, etc. The device can be constructed so that most of the operation indication functions are incorporated into a keyboard if this is arranged on the side of the input section 5010, or else into a display if it is arranged from the side of the

outlet section 5032.

  Reference is then made to FIG. 37, which shows an example of detailed structures relating to the treatment of proper names in section 5016 of morphological analysis. The morphological analysis section 5016 is illustrated for the part which has a direct link with the understanding of the invention, although it naturally has other functional analysis sections. Morphological analysis is carried out by the fact that instructions are given to search the dictionary from the head of the set of input characters and in following succession

  the set of consultation key characters and that The

  is executed for the LexicoLogical information obtained on

  this base coming from the consultation section of the diction-

  naire 5104 according to the position information of the proper name,

as described below.

  The morphological analysis section 5016 has an input processing section 5100 used to receive the data relating to the set of input characters which are supplied by the pre-processing section 5014 and to perform the input processing . The input processing section 5100 is provided with an input character set buffer which receives the data from the English character set in the form of code data, for example the ASCII code. , and which temporarily accumulates

  character set data.

  The data relating to the set of input characters which are temporarily stored in the input processing section 5100 is sent to a unit splitting section 5012 which splits your data into reference units of the dictionary, for example words . The 5102 unit splitting section is a functional section which distinguishes between the dictionary reference units constituting the sets of consultation key characters when the dictionary 5018 is subsequently consulted in the consultation section of the 5104 dictionary. reference delimiters which are used in the treatment

  the dictionary reference units are

  placed in the position of English character, number, apos-

  trophe, characters other than the hyphen and the point, as well as the apostrophe following a vacancy character. They are stored in a delimiter table 5108 to which iL is referenced when cutting the reference units from the dictionary in the section for cutting units 5102.

  The reference dictionary 5018 stored in particular

  link the information intended for the search for the discovery unit

  page. For example, as shown in FIG. 38, for the example of the information associated with the entry heading, grammatical information, such as the speech part, is contained in each of the reference units of the dictionary, for example the entry section of a word. The speech part information contains, for the name, an indication specifying whether it is a proper name or a common name. For the proper name, a

  indication of distinction showing how to limit the posi-

  tion in the sentence, that is to say relative position information

  tive in the proper name, is stored. This will be described more particularly

  especially below. Other information is also recorded.

  such as the character of accounting or non-accounting

  accounting of a noun, the distinctive character of transitivity or intransitive of a verb, translation of the word, etc. There are four types of position information for the proper name, namely configurations "0" to "3" in this embodiment. Configuration "0" indicates a proper name which is not subject to any position restriction, for example "City

  (City or City) "or the name of a person" Walter ".

  tion "1" indicates that it is a proper name, for example "Mr.", located at the head of an isolated proper name, or else a succession of several proper names, that is to say a proper name which is arranged in a unique group of proper names. Configuration "2" indicates that it is a proper name, for example "Station (Station)" or "Bay (Bay)", located at the end of an isolated proper name, or a proper name which has been arranged in a single word forming a group of proper names, and which does not enter into the case of configuration "3" described below. The configuration "3" indicates that it is a proper name, such as "River (Rivière)" in "the Sumida River (La Rivière Sumida)", which is similar to the configuration "2", but is accompanied by the definite article "the (the, the or them)" at the head of a proper name arranged so as to form a unit

  based on a group of proper names.

  The consultation section of the dictionary 5104 is a functional section which takes lexicological information by consulting the word dictionary 5013 on the basis of the set of consultation key characters provided by the unit cutting section 5102 and transfers it to the setting table

  lexicological information reserve 5124, in the processing section

  position information 5110 and to the determination section

  tion of the end of the preceding sentence 5112.

  Processing using configurations "0" to "3" depending on the position information of the proper names sought

  in dictionary 5018 is done by the milking sections

  proper names ment 5114, 5116 and 5118. The proper names are processed in connection with configuration I in the proper names processing section 5114, in connection with configurations 2 and 3 in the proper names processing section 5116 and in connection with configuration 0 in the sections of

  processing of proper names 5118 and 5114.

  In this embodiment, the collective arrangement of proper names is carried out using, as a key, a word which is a part of the set of proper names grouped into a single proper name and which is subject to a position restriction when it is arranged as a proper name. Thus, even if several proper names are present continuously, we can arrange them by

  appropriately in relation to the context without doing the agen-

  erroneous cement which consists in always arranging them simply as a single group of proper names. For this purpose, processing is carried out by the proper names processing sections 5114, 5116

and 5118.

  To a certain extent proper names are registered

  very in the reference dictionary 5018. These proper names registered in the dictionary are subjected to an analysis in the position information processing section 5110 and in the

  proper names processing sections 5114, 5116 and 5118. CeLles-

  these constitute a section for processing registered proper names

  trés in Dictionary. Proper names not registered in the 5018 dictionary are analyzed in the determination section.

  the end of the previous sentence 5112 and in the milking section

  names proper to configuration "0" 5118. These consist of

  kill the processing section of unregistered proper names

in Dictionary.

  The processing applied to a proper name is carried out according to the two stages below presented. First, a proper name belonging to a set of input characters is recognized. In the case of a word recorded in the dictionary 5018, this is achieved by the fact that the proper name is indicated in its morpheme activation information. In addition, in the case of a word not registered in the 5018 dictionary, this is accomplished by the fact that the character located at the head is an English capital letter, for example "John" or "U.S.", etc.

  Then a group of proper names is submitted to an agent.

  collective cement which will transform the whole part considered into a single proper name. When the latter is recognized as a proper name using the dictionary information and if the new dictionary reference unit is also a proper name, the whole part considered is collectively synthesized

in one proper name.

  For example, "Mr. Weber" is analyzed as a single proper name. The result of this analysis leads to a form

  capable of being grouped into an idiomatic expression comprising

  both proper names during local analysis.

  Then a necessary local analysis is carried out.

  In this case, a succession of analysis units activated by the addition of morpheme activation information for each of the analysis units are collectively arranged in a single analysis unit thanks to the use of a Local analysis rule, and, for example, an arrangement is made. For example "Mr. Brown" is collectively arranged in "Brown shi". In addition, words constituting part of a geographical name are also arranged collectively. For example, "Lake Biwa (Lake Biwa)" is collectively arranged in "Biwaco". Likewise, words that

  form part of a group name are also collected

  tively arranged. For example, "Yale University (University of

  Yale) "is analyzed as" Yale Daigaku ".

  In the case of the proper name "Mr ...." or "Lake ...", just before the word, there is always an end having regard to the context. Therefore, in the English phrase "With Tom Mr. Brown went to ...", there is some form of ending between "Tom" and "Mr." having regard to the context. Therefore, if "Tom Mr. Brown" was collectively arranged under one proper name, an error would be produced in the subsequent analysis. For example, the proper name "University" is always followed by an end, coming immediately after. For example, in the English phrase "At Yale University Tom is ...", it is recognized that there is punctuation between "University" and

  "Tom". In this embodiment, the information relating to the posi-

  tion o of the respective proper names of a succession of proper names undergo a position restriction is stored in the dictionary 5018 as position information as described above, that is to say by the configurations " 0 "to

  "3". The collective arrangement which uses this position information

  tion is carried out in processing sections 5110, 5112, 5114, 5116 and 5118. The lexicological information relating to the set of input characters is, once these treatments are completed, stored in the lexicological information buffer consulted, that is to say the information storage table

lexicological 5124.

  The result of the morphological analysis is transferred from the lexicological information storage table 5124 to the

  section I parsing 5020.

  The processing using the proper noun position information is performed via the sequence shown in Figure 40. The initial processing is performed by receiving the data from the input character set in the input processing section 5100 (5200). Next, the unit splitting section 5102 splits the set of input characters into units of

  dictionary reference for diction search-

  Mayor 5018 (5201). Dictionary lookup section 5104 searches dictionary 5018 on the basis thus stated (5203) and, if there is a dictionary entry field

  (5204), she examines the part of speech that it contains (5205).

  If the speech part is not a proper name, the processing

  reserved to the proper name in this embodiment is not effective

  killed, but the lexicographic information is stored in the lexicographic information storage table 5124 (5206). If it is a proper name, proper names stored in dictionary 5207 are processed in section 5110 of position information processing and in sections 5114,

  5116 and 5118 for processing proper names. When these treat-

  have been made up to the final sentence position indicated by the data for the input character set (5202), the result of the morphological analysis is delivered to the

  syntax analysis section I of sentence 5020 (5210).

  As a result of consulting the dictionary, if there is no entry item at step 5204 and if the element begins with a capital letter (5212), it is recognized as being an unregistered proper name in the dictionary and the treatment of proper names not registered in the dictionary 5213 is

  performed in section 5112 determining the previous part

  dente and in section 5118 on proper names treatment. If the initial character is not an uppercase letter, since it is a name not registered in the dictionary 5018, it is placed in reserve, as an unregistered word, in the reserve table d lexicographic information 5124 (5214). This treatment

  is executed up to the final position (5202).

  The processing of proper names recorded in Dictionary 5207 is carried out in Processing sections 5110, 5114, 5116 and 5118 according to the flow chart presented in FIG. 41. First of all, reference is made to the information of

  position obtained which is contained in the lexicogra- information

  phique. Then, the processing of proper names 5221 for the confi-

  guration 0 is carried out if a configuration "0" is indicated, the processing of proper names 5222 is carried out for configuration 1 if it is indicated a configuration "1", and the processing of proper names 5223 is carried out for configurations 2 or 3 if it is

  indicated respectively a configuration "2" or "3".

  The processing of proper names 5221 for the configuration

  tion 0 is executed in processing section 5114. Processing

  ment is applied to a proper name which contains no restrictions

  position. First of all, if a part preceding the reference unit of the dictionary is an unregistered proper name (5230), the whole part is collectively arranged in a proper name in the proper noun position information, ie "pos ", is" 1 "and it is stored in the Lexicographic information storage table 5124 (5233). If the previous part is a proper noun whose position information, ie "pos", is equal to "1" (5231),

  The treatment is carried out in the same way.

* If the previous part is a proper name having position information "O" (5231), the entire part is collectively arranged in a proper name having position information "0" and is stored in the layout table. Dictionary information reserve 5124 (5235). In addition, if the party

  previous is not a proper noun having position information

  tion "0", the entire game is stored alone as a proper name with position information "0" in the betting table

  in lexicographic information reserve 5124 (5134).

  The processing of proper names 5222 relative to configuration 1 is carried out as described below. This

  processing is applied to a proper name such as for example "Mr."

  which is arranged in the head position of an individual proper name or else in

2 6 0 4 8 1 4

  head of a proper name arranged in a single group from a success-

  transfer of several proper names. First of all, if the part preceding the reference unit of the dictionary considered is an unregistered proper name (5240), the word is transformed into non

  registered (5241). If it is not an unregistered proper name

  very, it is stored alone, as a proper name which has position information "pos" worth "1" in the layout table

  lexicographic information pool 5124 (5242).

  We will now provide explanations for what

  concerns the treatment of proper names 5223 for configura-

  2 and 3, in relation to FIG. 44. The processing is applied for example to a proper name such as "Station (Gare)" or "River (Rivière)" which is located at the end of an isolated proper name or of a proper name arranged in a single group from a succession of several proper names. First of all, if the part preceding the reference unit of the dictionary considered is an unregistered proper name (5250), it is collectively arranged with the preceding word to form a proper name having, as proper name position information " pos "proper position-specific position information" pos-self ", and it is stored in the lexicographic information storage table 5124 (5225). In addition, the processing is carried out in the same way if

  the previous part is a proper name with position information

"1" (5251).

  If the previous part is not a proper name having position information "0" (5252), it is collectively arranged alone in a proper name having, as proper name position information, proper name position information "pos-self" which is specific to it, and it is stored in the lexicographic information storage table 5124

(5257).

  In step 5252, if the preceding part is a proper name having the position information "O", the proper position position information "pos-self" which is specific to it is checked (5253), and the processing 5255 is performed if it is the configuration

2 6 0 4 8 1 4

  "2". If the proper position-specific position information "pos-self" is configuration "3", it is also checked

  if the element before the diction reference unit-

  naire is or is not "the (le, la or les)". If it is not the defined article "the", processing 5255 is carried out. If it is "the", the group going from "the" to the particular element is collectively arranged in a proper name having the proper noun position information "3" and it is stored in the table of

  storage of lexicographic information 5124 (5256).

  For a word starting with a capital letter and

  recognized as an unregistered word for which no word

  input brick was only obtained in the reference dictionary

  5018 following consultation of the dictionary 5203, the organization

  gram passes, via steps 5204 and 5212, to processing 5213, and processing 5213 is executed in the preceding end of sentence determination section 5112. First of all, if the previous part

  the dictionary reference unit considered is not suscep-

  tible to make the end of the sentence, the processing of proper names 5221 relating to configuration 0, as described above, is

  performed in processing section 5118.

  The preceding part may be likely to constitute the end of the sentence in the following four cases. First, there is The case where a separate point "." is present. Then there is the case where the previous entry item has a period at the end and where the position information of the proper name is not "1". This case includes for example abbreviations such as "U.S.A.". In addition, there is the case of the sign "two points" noted ":", the case of the semicolon ";", that of the succession of a point and an apostrophe ". '" And that of the succession of a point and a quotation mark "." ". This last case is that corresponding to the head of the

  input character set buffer.

  If the previous part is one of the four cases above

  mentioned, this previous part, which is likely to constitute

  kill the end of the sentence, is recognized as effectively being the end of the sentence (5261) and the dictionary is consulted after transformation of the capital letter of this word into a lowercase (5262). As a result of the consultation, When a

  entry entry in The dictionary (5263), The word is saved

  very in Dictionary storage table 5124 (5264).

  If this is not the case, iL is recorded in the reserve table of the dictionary 5124 as an unregistered proper name, that is to say not registered, the character placed at the head remaining

  unchanged, i.e. remaining a capital letter (5265).

  We will now provide explanations with reference to an example. For example, when using the dictionary for a set of input characters such as "Along the Sumida River Paul and Mr. Gold Smith went ... (Along the Sumida River Paul and Mr. Gold Smith went ...) ", The dictionary input information is first written to the lexicographic information storage table 5124, as shown in Figure 46A. Regarding for example "the (la)", the initial position in the sentence is "7" and the final position is "9", and the speech part is called a definite article. No entry rubric can be obtained for the word "Along" for the head position of the input character set when consulting the dictionary 5203, and this word is determined as n ' being not registered. However, since the previous part can satisfy the conditions specific to the end of a sentence, because it is at the top, or at the top, of the input buffer (5260), the capital letter "A" at the top is transformed into a lowercase character and searching the 5262 dictionary leads to "along".

  Then, the pointer is incremented to pass to the process.

  "Sumida". In this embodiment, this word is not included in the dictionary 5018. Since the previous part

  is not likely to constitute the end of the sentence, the organizer

  gram goes to processing of proper names 5221 for configuration

  tion 0. As shown in Figure 46A, a proper name and "0" are found respectively for the speech part and for

  proper name position information.

  The following dictionary reference unit "River" is a proper name with name position information

  own "3". The previous part is a proper name having the informa-

  tion of position "0" and the antepenultimate part is "the (la)". In view of what has been stated above, "the Sumida River" is collectively arranged under a single proper name

  through steps 5250 to 5254 and is stored, as inform-

  proper noun position position "3", in the layout table

  lexicographic information store 5124 (figure 46B).

  Then the next dictionary reference unit

  "Paul" is a proper unregistered noun having position information

  proper noun "0", for which processing 5213 is executed.

  Although the previous name is a proper name, since the position information for this word is "3", it is not collectively

  arranged with it, but the lexicographic information is accumulated

  muLé unchanged in table 5124 (Figure 46C). A treaty-

  ordinary is applied to the conjunction "and (and)" which follows it.

  The next word "Mr." is a proper name having the informa-

  tion of position "1", which is accumulated without modification in the lexicographic information reserve table 5124 (FIG. 46D). Even if the previous part is a proper name "Paul", since a punctuation between words is present just before "Mr.", this word is put in reserve without modification in the bet table

  in lexicographic information reserve 5124.

  Then, the word "Gold" is a proper name not registered in the dictionary and is treated 5213. Since the word "Mr." which precedes it has position information "1", both are collectively arranged, and The entire part is transformed into a proper name having position information "1" (Figure 46E). Then the treatment is applied in the same way to the next word "Smith" (Figure 46F). The next word "went" is a verb form of the past and a common analysis

is done after that.

  As described above in this embodiment

  tion, The proper names are collectively arranged using the

  sation, as key, of a word which constitutes a part of a group of proper names collectively arranged in a single proper name and which

  is subject to a position restriction when they are collecti-

  vely arranged in a proper name. In this way, even if more

  if their proper names are present continuously, it is possible to carry out a collective arrangement suitably according to the context without producing the erroneous collective arrangement which would consist in simply grouping them into a set of proper names. In the above example, "the Sumida River" is analyzed as a group of names

  which is separated from the proper name "Paul". In addition, "Mr.

  Gold Smith "is also analyzed as a group of names

clean.

2 6 0 4814

  We will then explain the sixth embodiment

according to the invention.

  FIG. 47 illustrates the sixth embodiment of the language analyzer according to the invention, applied to a device

  English-Japanese machine translation.

  The present embodiment has an input processing section 6014 and data is supplied to the input processing section 6014 by an input device 6012. The input device 6012 for example comprises a keyboard comprising keys characters, for example alphanumeric keys, and function keys, an optical character reading device for reading English text recorded on a

  sheet of paper, and a reading device for a magnetic disc

  tick. The input processing section 6014 has an input character set buffer 6014a and stores the English input phrase supplied by the input device 6012 in the character set buffer 6014a. The input processing section 6014 reads the input sentence stored in the input character set buffer 6014a and outputs it to a unit splitting section 6016. The unit splitting section 6016 is a functional section which splits dictionary reference units in the input sentence sent by the input processing section 6014 by searching in a table of

  6018 delimiters. The 6018 delimiters table contains delimi-

  such as spacing, point, etc. The unit splitting section 6016 reads the delimiters in the delimiter table 6018 and divides the input sentence

  sent by processing section 6014 in character sets

  tères which constitute the units used for searching in a reference dictionary 6020 by dividing the sentence at the level of the parts where delimiters are present. Divided character sets are sent to a consultation section of

Dictionary 6022.

  The dictionary look-up section 6022 consults the reference dictionary 6020 to process the input sentence sent by the unit splitting section 6016, which has been divided into dictionary reference units. The reference dictionary 6020 contains, as shown for example in FIG. 48, input items for the sets of characters, parts of speech corresponding to these, specific information, and other information useful for interpreting the English phrase. The reference dictionary 6020 contains, in addition to the proper names indicated in the figure, sets of characters referring to other parts of speech, for example verbs, adjectives, etc. The proper names considered as parts of speech in this figure correspond to those to which the treatment of the inscribed proper names described below is applied, but do not refer to the proper names as usual grammatical entities. In addition, the information relating to the particularity indicates what the proper name

  concerned expresses and is not limited to a single category.

  Dictionary lookup section 6022 consults the reference dictionary 6020 for the character set divided into dictionary reference units and, if the character set is a proper name, it delivers it to a name processing section 6024 for the processing of proper names described below. In addition, if it is not a proper name, it is issued to a processing section 6036 and is

  stored in the lexico- information storage table

  graph 6036a of processing section 6036.

  The proper names processing section 6024 includes a preceding end of sentence processing section 6026, a previous proper names processing section 6028, and a section

  6030 proper name processing itself. The preceding end-of-sentence processing section 6026 determines whether a

  character set preceding the supplied character set searched for in the dictionary lookup section 6022 may or may not be at the end of the sentence and, if the preceding character set is at the end of the sentence, it sends to Dictionary consultation section 6022 after transforming the capital letter at the head of the set of characters to be processed into a lowercase letter and has the 6020 reference dictionary searched again by the

  Dictionary lookup section 6022. The character set

  tères not found during the second search is considered

  being a proper name not registered, is sent to the processing section

  6036 and is stored in the lexicographic information storage table 6036a. In addition, if the character set preceding the supplied character set is not at the end of the sentence, it is sent to the processing section 6036 as being a proper name for which the information relating to the particularity is not known and is entered in the lexicographic information storage table 6036a as described below. Section 6028 of proper name treatment previous analysis Information on the peculiarity of

  the previous set of characters from the milking section

  ment at the end of the previous sentence 6026 and delivers the result to section 6030 for processing the proper name itself. The proper name processing section 6030 itself verifies the information relating to the particularity of the proper name to be analyzed and,

  as described below, if the information relating to the particu -

  The entity is not entered for the proper name or else for the previous proper name, that is to say for one of the two, it analyzes together the proper name and the previous proper name using the information of peculiarity of the other and stores the

  result in the lexicogra- information storage table

  ph 6036a from processing section 6036.

  The processing section 6036, which has the lexicographic information storage table 6036a, stores therein the data sent by the previous proper name processing section 6028 or the dictionary consultation section 6022, then reads the data. thus stored and delivers them to a section 6038 of structural analysis, or syntactic analysis. The syntax analysis section 6038 performs the structural analysis of the input sentence, read in the Lexicographic information storage table 6036a and submitted to the analysis.

morphological.

  We will now explain the operation of the

  refer to the flowchart illustrated in Figure 49.

  First, an English input sentence from the input device 6012 is read in The input processing section 6014 (6100). The input sentence read in The processing section

  input 6014 is stored in the character set buffer

  entry points 6014a. The input phrase stored in The input character set buffer 6014a is read and placed in the

  unit cutting section 6016.

  When the input sentence has been entered, the unit splitting section 6018 Reads the delimiters in the table

  6018 delimiters in order to split the reference units

  dictionary reference (6102). Thus, the sets of characters constituting the input sentence provided are successively divided, starting from the top of the sentence, that is to say from its head, into sets of consultation key characters which constitute the

  units used when consulting the reference dictionary

  rence 6020, thanks to a division operated at the level of positions or delimiters, for example spacings and signs "colon", are present. It is determined whether or not the dictionary reference units thus divided, that is, the lookup key character sets, have expired or not (6104) and, if there is still a set of lookup key characters, lookup key character set

  is sent to the consultation section of dictionary 6022.

  When the lookup key character set is sent to the lookup section of dictionary 6022, the

  consultation section of the dictionary 6022 consult the diction-

  nary of reference 6020 relative to the set of characters of consultation key (6106). It is determined whether or not the set of consultation key characters is present in the

  input items from the reference dictionary 6020, as repre-

  felt in FIG. 48, (6108) and, if there is an entry item, the information relating to the part of speech which is stored in the reference dictionary 6020 is read and it

  can be determined whether the character set of consultation key

  tion is or is not a proper name (6110).

  In the case where the set of consul- tation key characters

  tation is not a proper name, the consultation section of the dictionary 6022 delivers the data read in the reference dictionary 6020 to the processing section 6036 and stores them in the lexicographic information storage table 6036a (6112) . When the data is stored in the dictionary information storage table 6036a, an input signal indicating that the input data is stored and the data of the consultation key character set stored just before is sent by the processing section 6036 to the unit cutting section 6016. Then, the flowchart returns to step 6012 and the cutting of the dictionary reference units is carried out again in the cutting section

6016 units.

  In step 6110, if the key character character set of

  consultation is a proper name, the consultation section of the dictionary

  tionnaire 6022 sends the proper name read in the reference dictionary 6020 (which will be referred to hereinafter simply as the proper name), together with the data from the previous set of consultation key characters provided by the table

  for securing information from the 6036a dictionary of the sec-

  treatment 6036 in the consultation section of the diction-

  naire 6022 via the unit cutting section

  6016, in the proper names processing section 6024, where the

  the proper name entered in the dictionary is carried out

(6114).

  Next, we will explain the treatment applied to names

  own registered in the dictionary by referring to the organization

gram of Figure 50.

  The data sent by the dictionary consultation section 6022 to the proper names processing section 6024 is delivered, via the end of sentence processing section 6026

  above, in section 6028 of processing of the proper proper name

  tooth. For the treatment of the proper name registered in the diction-

  nary, section 6026 of the preceding end-of-sentence processing does not

plays no role.

  In section 6028 of the proper proper name treatment

  tooth, it is determined whether or not the set of consultation key characters which precedes the proper name is a proper name not included in the reference dictionary 6020, that is to say whether or not it is a name own subject to name processing

  own not listed in the dictionary (6200), as described above

  after. If it is a proper name not registered, the whole part consisting of the proper name and the proper name not registered

  previous is determined to be a proper noun having the in-

  peculiarity formation of a proper name (6202), and the data is sent to the processing section 6036 and stored in the lexicographic information storage table 6036a

(6214).

  In section 6028 of the proper proper name treatment

  tooth, if the set of lookup key characters preceding the proper name is determined to be an unregistered proper name, at step 6200 it is determined whether the set of lookup key characters before the proper name is or not a proper name registered in the reference dictionary 6020 (6204). If the lookup key character set preceding the proper name is a registered proper name, it is determined whether the information

  particularity of the previous proper name is unknown or not, that is

  ie whether or not it is not registered in the reference dictionary

6020 (6206).

  If the particular information of the previous proper name is unknown, the flowchart proceeds to Step 6202, o The whole part consisting of the proper name and the previous proper name is determined to be a unique proper name having the information of peculiarity of the proper name (6202), and the previous proper name processing section 6028 sends the data to the processing section 6036. The data sent to the processing section 6036 is

  stored in the lexico- information storage table

chart 6036a (6214).

  In section 6028 of the proper proper name treatment

  tooth, if the particularity information of this name is deter-

  mined as not not unknown, that is, if it is determined to be registered in the reference dictionary 6020, the data is sent by the processing section 6028 of the unique proper name preceding to the processing section of the proper name 6030. In the proper name processing section 6030, it is determined whether the particularity information of this name is not unknown or not (6208). If the proper name specificity information is unknown, the proper name processing section 6030 determines that the entire part consisting of the proper name

  and of the previous proper name is assigned specific information

  the name of the previous proper name (6210), and sends the data to the processing section 6036. The data delivered to the processing section 6036 are recorded in the reserve table

  lexicographical information 6036a (6214).

  In section 6030 of the proper name processing section, if it is determined that the particularity information of this name is not unknown, that is, if it is entered

  in reference dictionary 6020, section 6030 of processing

  proper name determines that the proper name is a proper name having specific information found in the reference dictionary 6020, independently of the previous proper name (6212), and it delivers the data to the processing section 6036. The data sent to the processing section 6036 are recorded in the lexicographic information storage table

6036a (6214).

  We return to FIG. 49. If there is no set of consultation key characters in the entry items of the reference dictionary 6020 at step 6108, it is determined whether

  the first character in the character set of the consultation key

  tation is or is not a capital letter (6116) and, if it is not

  not a capital letter, the consultation section of the diction-

  nary 6022 determines that the character set of consultation key

  tation is an unregistered word, it sends it to the processing section

  6036 and stores it in the reserve table

  of lexicographic information 6036a (6118).

  If the first character is a capital letter, the data relating to the set of consultation key characters is sent, together with the data relating to the previous set of consultation key characters, from the consultation section of dictionary 6022 in the proper name processing section 6024, o the treatment reserved for the unregistered proper name

  in the dictionary is performed (6120).

  We will now explain the treatment applied to the proper name not registered in the dictionary, in relation to the

figure 51.

  The data relating to the set of consultation key characters is sent, together with the data relating to the previous entry item recorded in the lexicographic information reserve table, in section 6026 for processing previous end of sentence o it is determined whether the end of the previous entry item saved

  in the information storage table is or is not suscep-

  tible to be the end of the sentence (6300). This decision as to whether it is the end of the sentence is obtained by determining that the end of the previous entry rubric

  recorded in the lexico- information storage table

  graphic may or may not constitute the end of the sentence, for example by the existence of a separate dot ".", etc. If the end of the previous entry item recorded in the lexicographic information storage table is likely to constitute the end of the sentence, the data are

  sent by Section 6026 of end-of-sentence processing

  to the previous proper name processing section 6028, and The processing section 6028 determines that The entry topic

  stored in the information storage table

  lexicographic tion constitutes the end of the sentence (6302) and it sends the data to the consultation section of dictionary 6022 after changing the capital letter at the head of the consultation key character set to a lowercase letter. The consultation section of the 6022 dictionary consults the reference dictionary 6020 for the set of consultation key characters transformed by the change from capital letters to lowercase (6304), and it determines whether there is an entry item in the reference dictionary 6020 (6306). If there is an entry item, the consultation section of the dictionary 6022 sends the data found in the reference dictionary 6020 to the processing section 6036 and stores it in the lexicographic information storage table 6036a

  (6308). If there is no entry section, the consultation section

  tation from the dictionary 6022 transforms the first character of the set of consultation key characters into an uppercase letter, delivers it as a proper name not entered in the processing section 6036 and stores it in the reserve table

  lexicographical information 6036a (6310).

  In step 6300, if the preceding sentence processing section 6026 determines that the end of the entry item

  stored in the information storage table

  lexicographic tion is not likely to constitute the end of

  the sentence, the data is sent from processing section 6026

  preceding end-of-sentence ment in section 6028 of previous proper noun processing, and section 6028 of previous proper noun processing determines that the previous entry item

  recorded in the lexico- information storage table

  graphic is not the end of the sentence (6312). Data is sent from the previous proper name processing section 6028 to

  Section 6030 of the proper name processing, and the processing section

  6030 determines that the key character set of

  consultation is a proper name for which the information of parti-

circularity is unknown (6314).

  The proper name processing section 6030 returns the data to the previous proper name processing section 6028 and the processing reserved for the proper proper name registered in the dictionary is carried out in the previous proper name processing section 6028 (6316). The treatment of the proper name entered in the dictionary is identical to what was presented on the

figure 50.

  Returning to FIG. 49, if the cutting of the dictionary reference units operated in step 6104 is finished, the data recorded in the lexicographic information reserve table 6036a are sent by the processing section 6036 to the parsing section 6038 (6122), what

  the morphological analysis according to the invention is completed.

  We will now explain the operation of the device of this embodiment, as described above, in

  relation to an example input sentence.

  The explanations will be given in relation to FIG. 52 in the case, taken by way of example, of the sentence

  "In Tokyo Station Mr. Walter ... (In Tokyo Station Mr.

  Walter ...) ". First, the entry processing 6100 aimed at

  read the input sentence and place it in the milking section

  Entry 6014 is performed. Then the dictionary splitting units are split 6102, so as to divide the input sentence into respective words by means of spaces. First of all, the reference dictionary 6020 is consulted for "In (In)" (6106). There is no entry heading for "In" in the reference dictionary 6020. When the organization chart passes to the processing of the proper name not registered in the dictionary, since

2 6 0 4 8 14

  The previous part is recognized as being The head of the sentence (the top of the file), "In" is transformed into "in '. Since" in "has an entry item in the reference dictionary 6020 and it doesn is not a proper name (6110), The data found in the reference dictionary 6020 are recorded in the lexicographic information storage table 6036a

(6112).

  Then, the reference dictionary 6020 is consulted for "Tokyo" (6106). Since there is no entry field for "Tokyo" in the reference dictionary 6020 (6108) and since the first character is a capital letter (6116), the processing intended for the proper name not registered in the dictionary is performs (6120). Then the flowchart goes to figure 51. Since the

  previous part is "In" and it is not likely to cons-

  titrate the end of the sentence (6300), "In" is recognized as not being the end of the sentence (6312), "Tokyo" is recognized as being a proper name whose particularity information is not known (6314), and the processing of the proper name entered in the dictionary

is performed (6316).

  Next, the flowchart goes to Figure 50. Since the previous word "In" is neither a registered proper name, nor an unregistered proper name (6204), "Tokyo" is recorded only as a proper name with a specific information of its own,

  that is to say a proper name for which the particular information

tee is unknown (6216).

  Then, the flowchart returns to FIG. 49, and the reference dictionary 6200 is consulted for "Station (Station)" (6106). Since there is an entry item for "Station" in the reference dictionary 6020 (6108) and this relates to a proper name (6110), the processing for the proper name entered in the dictionary is carried out (6114) . The flowchart goes to Figure 50. Since the previous word "Tokyo" is an unregistered proper name (6200), the entire part "Tokyo Station (Tokyo Station)" is recorded as a proper name that has

  the specific information "place" belonging to "Station" (6202).

  Then, as shown in Figure 49, The 6200 reference dictionary is consulted for "Mr." (6016). Since there is an entry section for "Mr." in the reference dictionary 6020 and that it is a proper name (6110), The treatment intended

  in the proper name entered in the dictionary is made (6114).

  Next, the flowchart passes to FIG. 50. The preceding word "Station" is not an unregistered proper name (6200), but a registered proper name (6204), and the specific information "place" does not is not unknown (6206). Since "Mr." has specific information "person" which is not unknown (6208), "Mr." East

  listed alone as a proper name having particular information

"person" rity (6212).

  Then we come back to figure 49, and the

  reference dictionary 6020 is consulted for "Walter" (6016).

  Since "Walter" has an entry item in the reference dictionary 6020 (6108) and it is a proper name (6110), the treatment reserved for proper names registered in the dictionary is

  performed (6114). Next, the flowchart goes to Figure 50.

  Since the previous word "Mr." is not an unregistered proper name (6200), but is a registered proper name (6204) having specificity information "person", which is not unknown, (6202), while specificity information from "Walter" is unknown (6208), "Mr. Walter" is collectively arranged and recorded

  as a proper name having the specific information "per-

rings "(6210).

* As described above, in this embodiment, the English input sentence is divided into sets of lookup key characters for which, first, the reference dictionary 6020 is looked up . If there is an entry item as a proper name in the reference dictionary 6020, the processing of the registered proper name is carried out. The processing of the registered proper name is carried out at the same time as the entry section is taken into consideration

  stored in the information storage table

  lexicographic tion. If the previous entry rubric

26 0 4 8 1 4

  registered in the lexicogra- information storage table

  phique is a proper name, it is a verification of the information

  particularity of the previous item recorded in the lexicographic information storage table and of the proper name being processed. If one of the two does not have specific information, the information of the nature of the other word is taken, while, if both have specific information, they are recognized as names own own separately specific information that their

are respectively clean.

  It is therefore possible to provide information

  of proper particularity a proper name with no information

  particularity as well as limiting the particularity information produced in a more suitable manner. This allows more efficient parsing during subsequent parsing

and an adequate translation. -

  In addition, for the set of characters not entered in the reference dictionary 6020, if the first character is a capital letter and the preceding set of characters is

  determined to be the end of the sentence, the dictionary of reference

  rence 6020 is consulted again after the capital letter has been converted to lowercase and, therefore, it is possible to consult the reference dictionary 6020 also for the set of characters at the head of the sentence. In addition, if a set of characters beginning with an uppercase letter appears at a position other than the head of the sentence, it is determined to be a proper name, and its proper name specificity information is provided by a proper name with information of particularity entered which is present before or after it. Consequently, it is possible to analyze to a certain extent a proper name not registered in the dictionary of

reference 6020.

26048 1 4

  We will now explain the seventh embodiment

tion of the invention.

  FIG. 54 illustrates the general structure of the seventh embodiment of the language analyzer according to the invention,

  applied to an English-Japanese machine translation device

  born. This embodiment has an input section 7010, through which an English text 7012 to be translated into Japanese is provided. The input section 7010 can for example comprise a keyboard having numeric keys, for example alphanumeric keys, and function keys, an optical character reading device (OCR) which reads the English text recorded on a sheet paper and / or a file memory device for reading the recorded English text

  on a memory medium, for example a magnetic disc.

  English text provided by entrance section 7010

  is read in a pre-shaping section 7014, where a pre-process

  Translation is done. In this case, sentence recognition and the processing of unknown words are mainly

  performed. Its function is to execute part of the analysis.

morphological lysis.

  The pre-formatted English data is transferred to a section 7016 of morphological analysis at the same time as the information obtained during the pre-formatting. The morphological analysis section 7016 morphologically analyzes the input string of the English sentence at the same time as it divides it by consulting a dictionary of words 7018, and performs various types

  arrangements such as the processing of unknown words, the expression

  proper names, time, numbers, etc., and performs the processing of the whole sentence, for example for redundant query, of the type "isn't it? (isn't it?) "

  and position recognition. Morpholo-

  are contained in a 7036 analysis rules file.

  English data subject to morphological analysis is transferred to a syntax analysis section I 7020 at the same

2 6 0 4 8 14

  time that the dictionary information obtained by moral analysis

  phological. Section I of syntax analysis 7020 is a functional section which analyzes the structure of the cb-surface of the sentence while applying grammar rules to the data.

  English and updates all structural possibilities.

  The English data submitted for analysis in section I of syntax analysis 7020 are sent, together with the analysis information, to a section II of syntax analysis 7022, where a solution is chosen on the basis of the analysis relating to the surface layer structure made in section I

  syntactic analysis by applying a structured description

  connects. Then, a plausible tree of analysis of the English sentence is prepared, and the structure is achieved. Parsing rules are also stored in the rules file

7036.

  After being submitted for analysis, the English data is transferred in the form of analysis tree data to a structure transformation section 7024. Section 7024 of structural transformation prepares a corresponding Japanese structure tree from the structure tree, which is an intermediate English structure, and transforms it into an underlying Japanese structure, from which a sentence

  from Japanese can be easily translated.

  Data from the structural tree showing the underlying structure of Japanese thus transformed is sent to a translation production section 7026, o a translated sentence

  is produced. This is a functional section used to pro-

  draw a Japanese sentence from the Japanese structure tree. Japanese data formed as a translated sentence, that is, data from the translated sentence, is sent to a post-shaping section 7030. The post-shaping section 7030 modifies the data from the sentence translated using the 7018 dictionary and also using the information used in the translation processing to finally obtain a more natural Japanese sentence. The data of the Japanese sentence is transferred to an output section 7032, then is delivered by the latter in the form of a translated Japanese sentence 7034. The output section 7032 can comprise for example a printer, a display, and, or a file memory device, for example a magnetic disk. The flow of the translation processing series is controlled by a control section 7038 which governs the

  control of the device as a whole.

  In this embodiment, the lexicographic data relating to Japanese English words are stored in the word dictionary 7018. These are the established vocabulary, as well as connection relationships, namely coexistence relationships and various information such as meanings, forms of singular and plural, "parts of speech", etc. In addition, the rules data for morphological analysis and syntax analysis are stored in the rules file.

7036.

  The control section 7038 is connected to the section

  7040 Operation Display Section. 7040 Operation Display Section

  tions includes operation keys by which an operator can give various instructions to the device, for example a translation instruction key and a cursor key, and a display or indicator which visually displays the text of the

  input English sentence, the resulting Japanese sentence

  translation status, intermediate data such as lexicographic information and various instructions for the operator. It can be arranged that most of the operation display functions are incorporated in a keyboard, which will then be placed in the input section 7010, or else in a

  display, which will then be placed in the output section 7032.

  Reference is made to FIG. 53, which shows examples of detailed structures relating to section 7016 of morphological analysis in the case of number processing. The parts which are directly related to the understanding of the invention are presented, even if the morphological analysis section 7016 has

  of course other functional analysis sections.

  Section 7016 of the morphological analysis has an input processing section 7100 which receives and processes the

  input character set data provided by sec-

  pre-shaping tion 7014. The input processing section

  7100 has a received input character set buffer

  the data relating to the English character set in the form of coded data, for example the ASCII code, and temporarily stores the data relating to the character set.

  The data for the tempo input character set

  often stored in the input processing section 7100 are sent to a unit splitting section 7102 which splits the data into dictionary reference units, for example by

  words. Section 7102 for unit cutting is a functional section.

  which distinguishes a reference unit from the dictionary

  constituting the set of characters during the subsequent consultation

  dictionary 7018 in dictionary lookup section 7106. Dictionary reference delimiters that

  are used in the process of cutting the reference units

  dictionary rence are placed at the position of a character

  English, numeric, single quote, character

  other than the hyphen and the dot, as well as the appos-

  trophe following a vacancy character. They are housed

  sines in a table of delimiters 7104 and it is called upon when cutting the dictionary reference units in the

  unit cutting section 7102.

  The 7018 dictionary contains specific information

  liers for the search for cut units. In addition, the diction-

  naire 7018 contains morpheme processing information, such as the name of the month, the name of the day of the week, the cardinal number representing only a numeric number, the

  ordinal number, the unit used to express The gram or a large

  analogous deur, Le temps, "the (definite article)", "of (de)", the vir-

  guLe "," Le point ".", etc. The consultation section of the dictionary 7106 is a functional section which consults the dictionary 7018 to extract lexicographic information relating to the set of characters provided by the unit splitting section 7102 and transfers these to a production section-7108 of information

morpheme treatment.

  Morpheme Processing Information Production Section 7108 Contains Morpheme Processing Information

  (see Figure 56) indicating that a succession of characters pre-

  feeling a certain morphological aspect has meaning

  time, such as time, year, month, etc., and information

  additional specific tion is provided for the character set

  which is recognized as containing a cardinal number or a

  meaning of time, in the diction consultation section-

  naire 7016. For example, information such as:

  "figure figure" means "year".

  The set of characters to which an information has been added

  mation in section 7108 for processing information production

  In addition, the morpheme is subject to the necessary local analysis.

  In this case, a group of units made up of reference units

  dictionary reference, for example a word, activated by information

  tion of morpheme activation is collectively arranged in a single unit by the use of local analysis rules. For example "month name", "numerical expression" are collectively arranged in "month name + numerical expression", that is to say that "Oct." and "18" are grouped in "Oct. 18 (in French, the practice requires saying 18 Oct.)". At the same time, there is also a collective arrangement such as, for example, "November the 2nd" (November 2nd, that is to say November 2) for "month name + the (the) - +

  digital expression "," 22 March (22 March) "for" digital expression

  rique + nom de mois "," the 23rd May "(23rd of May, ie 23

26048 14

  May) for "the (le) + numeric expression + month name", "the 11th of June (le 11ème de juin, i.e. 11 June)" for "the

  (le) + numeric expression + of (de) + nom de mois "," '86, Jan.

  27, Mon. (86, Jan. 27, Mon, that is Monday January 27, 1986) "for" year +, + month and days +, + day of the week "," Sunday, 26, Jan., 1986 (Sunday January 26, 1986) "for" day of the week +, + month and day +, + year "," 11:30 am (11:30 am) "for" number: number + am (that is ie before noon) (or pm (ie after noon)) "or" name of month + year "," name of month + of (of) + year ", etc. Processing relating to local analysis is carried out in a section 7110 for setting the initial value, a section 7112 for finding a match, a section 7114 for dividing up units, a section 7118 for producing morpheme processing information, search sections 7116 and 7120 and processing sections 7122 and 7124, as well as a matching table 7128 containing a morpheme processing indication table which is a distinction / reference table making it possible to distinguish the fact that '' a succession of units comprising of s time factors and numbers, as shown in Figure 57, is a unit of time factors that obeys a certain rule. Section 7110 for setting the initial value fixes the unit value of a counter n which counts the number of dictionary reference units with matching when searching for a succession of dictionary reference units forming a group units as described above in the section

  search engine 7112.

  Section 7112 concordance search con-

  see the matching table 7128 for each of the units

  dictionary reference in order to match

  dance. The unit splitting section 7114 distinguishes the dictionary reference units, which will be assumed to be "p", supplemented by the dictionary search carried out in the dictionary consultation section 7106, from the reference reference units. dictionary constituting the sets of characters after the dictionary reference units have been completed using the dictionary consultation by the

counter n.

  The consultation section 7116 is a functional section.

  with the same function as the consultation section of the dictionary 7106, that is to say that it consults the dictionary

  7018 to extract dictionary information, or informa-

  lexicographical terms, relating to the set of characters which is distinguished in the unit cutting section 7114 and the

  transfers to section 7118 for processing information production

  morpheme. The morpheme processing information producing section 7118 has the same function as the morpheme processing information producing section 7108, where additional specific information is provided for each of those recognized as ordinal numbers or of

  time factors in consultation section 7116.

  The consultation section 7120 and the treatment sections

  7122 and 7124 produce a collective arrangement for a succession of dictionary reference units up to "p + n" obtained from the search matching section 7112 via the processing carried out in section 7118 of production of morpheme processing information, in order to form a single reference unit of the dictionary. Then the result is

  stored in the lexico- information storage table

  graphic 7126, which is a buffer for storing information

  lexicographic tion finally obtained with the help of the consultation

from the dictionary.

  The result of the morphological analysis is transferred from the lexicographic information storage table 7126 to the

  section I parsing 7020.

  We will then explain the collective arrangement obtained through the morpheme processing information according to the invention, referring to the flowcharts presented on the

Figures 55A and 55B.

26 0 4 8 1 4

  Suppose for example that the set of characters

  Next is provided to the input processing section 7100 (7300).

Input character set:

  ".. 26 Jan., '80 he .. (.. 26 January 1980 it ..)".

  The unit splitting section 7102 splits the set of input characters into dictionary reference units for the purpose of consulting the dictionary 7018 (7302). In the input character set, "26" is considered to be one unit when cutting dictionary reference units. It is determined whether or not the division of the reference units of the dictionary is completed for the set of input characters and, if it is finished, the operation ends (7304), while, if it does not is not

  finished, the flowchart goes to the next operation 7306.

  The dictionary 7018 is consulted for "26", which represents

  feels, in the set of input characters, the unit cut out for the transfer to the dictionary, so that it is taken in the

  dictionary the information indicating that "26" is "digit" (7306).

  Next, morpheme processing information indicating that

  "figure" has a particular morphological nature, that is to say

  say that it is a succession of digits to be treated as a combined cardinal number (7308). It is determined whether or not the group associated with the dictionary information was provided with morpheme processing information during step 7308 (7310). Then, if this is the case, the flowchart goes to step 7314 for the further application of processing based on the local analysis rule, while a group which has not been endowed of this information is recorded in the lexicographic information storage table 7126 (7312), after which the flowchart returns to step 7302. Therefore, "26" with the morpheme processing information advances to step 7314.

  The processing of step 7314 is carried out according to the organ-

  flowchart shown in Figure 55B.

  First, an initial value "0" is positioned in a counter n which counts the number of matches

  dictionary reference units when the reference units

  References of the dictionary are sought in Section 7112 of search for concordance (7410). In addition, since the dictionary reference unit completed by consulting the

  dictionary performed in the consultation section of the diction-

  naire 7016 is positioned as being at "p", the matching table 7128 is consulted for the (p + n) th with (n = 0) dictionary reference unit, ie "26" , by the matching search section 7112 (7412). Since "26" was provided with morpheme processing information indicating that it

  is a cardinal number in step 7308 and since these agencies-

  ment "each having a cardinal number" at their head part are present in the second and after the second in the section on the arrangement in the mapping table 7128 (see Figure 57), the reference unit of dictionary "26" is equal to the information in the mapping table 7128 and is therefore in agreement with this information. In this case, the matching is done for Ms-Me (see Figure 57) at the same time as the second matching arrangement such as "Ms" is fixed, together with the last combination data item having the " cardinal number "at its head part such as" Me "in the table of

reconciliation 7128.

  On the basis of the results of the matching carried out in the matching table 7128 for the (p + n) th (with n = 0) reference unit of the dictionary, an attempt is made to decide on the matching state (7414) and, if it is determined that there is a match, the flow chart proceeds to step 7416, while, if it is determined that there is no match,

  the flow chart proceeds to step 7424.

  If it is determined that there is a match, "1" is posi-

  tioned in the counter n so that the dictionary cutting unit is cut out at the (p + l) th position (n = 1) in the set of input characters. The cutting is carried out in the same way as in step 7302. The dictionary 7018 is consulted for "jan.," In the set of input characters cut out.

2 6 0 4814

  as a dictionary reference unit in the immediate vicinity of "26" by the processing in order to produce the morpheme processing information (7420, 7422). These treatments are performed the same

  so as in step 7306 and step 7308.

  By repeating the procedures between step 7412 and step 7422 above, the flowchart loops to "Jan. 26, he (Jan. 26, 1980)." However, since "it" does not match in the matching with the matching table 7128 in step 7412, the flowchart advances via step 7414 to step 7424. This means that, while the data is, until "Jan 26, '80", matched with "cardinal number, month, year" in the match table 7128, the match

  concordance fails for "26, Jan, '80 he".

  In addition, if the sentence ends at the character set

  input tères, for example "26 Jan., '80", that is to say if there

  not a break in succession for the diction reference unit-

  next year, the flowchart advances via step 7418 to

step 7424.

  In the case where it is determined that there is no matching at step 7414, it is determined if the content of the counter n is not greater than 1 or is greater than 1 (7424), and, if it is not greater than 1, it is registered as the sole unit of

  dictionary reference in the information storage table

  lexicolographic mation 7126 (7434).

  If it is not less than 1, the matching is performed for p + n (n = 3), that is to say that "he (il)" from "Jan 26, he" is " EOS ", which indicates the end of the arrangement (7426, 7428). If there is no matching, the flowchart advances to step 7434. If there is matching, "Jan 26, '80" which

  constitutes the position p- (p + n-1) for the reference unit of the dic-

  , is collectively arranged according to the result of the arrangement corresponding to the Ms arrangement in the mapping table 7128, and the result is recorded in the table

  7126 for storing lexicographic information.

  Then, it is accepted that the reference units of the dictionary have reached their end, that is to say at the (p + n-1) th unit, and a repositioning of (p + n-1) is carried out ) to "p"

(7432).

  The eighth embodiment of the invention will now

nant be explained.

  FIG. 63 illustrates the general structure of the eighth embodiment of the language analyzer according to the invention,

  applied to an English machine translation device-

  Japanese. In FIG. 63, an input section 8001, an English text 8002, a pre-formatting section 8003, a morphological analysis section 8004, a syntactic analysis section 8005, a section are represented. Sentence syntax analysis II 8006, an operation display section 8007, a word dictionary 8008, an analysis rule file 8009, a

  8010 control section, a structural transformation section

  ture 8011, a translated sentence production section 8012, a post-shaping section 8013, an output section 8014, and

8015 Japanese text.

  The present translation device comprises, as shown in the figure, an input section 8001 by

  which the English text 8002 to translate into Japanese is intro-

  duit. The input section 8001 may for example include a keyboard having character keys, such as alphanumeric keys, and function keys, an optical character reading device (OCR) for reading the text.

  English, or alternatively a file memory device allows

  both read the English text saved on a medium.

  memory, for example a magnetic disc.

  English text provided by entrance section 8001

  is read in the pre-shaping section 8003, where a pre-

  translation processing is performed. In this case, the recognition

  cance of the sentence and the treatment of unknown words are main

  palement carried out. The role of this section is to

part of the morphological analysis.

  The pre-formatted English data is sent to the morphological analysis section 8004 along with the information obtained during the pre-formatting. Section 8004 of morphological analysis divides the data by referring to the dictionary of words 8008, morphologically analyzes the English string, performs

  various types of arrangements such as word processing incon-

  nudes, the expression of proper names, time and numbers, and performs a processing relating to the entire sentence, for example for the redundant query (which is known as

  the designation "tag question") and recognition of the affixing.

  Morphological analysis rules are contained in the file

of analysis rules 8009.

  After the English data has been subjected to morphological analysis, it is transferred to Section I of syntactic analysis 8005 together with the dictionary information, or lexicographic information, obtained by morphological analysis. Section I of parsing 8005 is a functional section which analyzes, at the syntax level, the structure of the surface layer of sentences by applying a grammatical rule to English data and updates all the

structural possibilities.

  After being subjected to analysis in section I parsing 8005, the data is sent, together with the analysis information, to section II parsing 8806, where a solution is chosen from the results of the syntactic analysis carried out in relation to the surface layer by section I of syntactic analysis, by means of the application of the

  structural description. Thus, an analysis tree is prepared

  plausible for the description in English so as to establish the

  structure. These analysis rules are also contained in the

analysis rules file 8009.

  The English data submitted to the syntax analysis is transferred, as analysis tree data, to the structure transformation section 8011. The transformation section of

  structure 8011 prepares a corresponding Japanese structure tree

  from the structure tree, which is an inter-

  medium of the English sentence, in order to transform it into an underlying Japanese structure from which the sentence

  from Japanese can be easily translated.

  The data relating to the structure tree indicating the underlying structure of Japanese thus transformed is sent to the translation production section 8012, where the translated sentence is produced. This is a functional section used to generate a Japanese sentence from the structure

  of the Japanese sentence structure tree.

  The Japanese sentence data thus translated, that is, the data relating to the translated sentence, is sent to the post-shaping section 8013. The post-shaping section 8013 modifies the data translated by consulting the dictionary 8008 while using the information used in the translation processing in order to finally arrive at a more natural Japanese sentence. The data relating to the translated Japanese sentence is transferred to the output section 8014 and delivered, as a translated Japanese sentence 8015, by the output section 8014. The output section 8014 comprises for example a printer, a display, and, or, a memory device to

  file, for example a magnetic disc.

  The flow of the translation processing series is controlled by the control section 8010, which governs the control of the device as a whole. The dictionary of words 8008 contains, in the case of the embodiment shown, lexicographical data relating to the words of English and Japanese, in which are established, in addition to the vocabulary,

  various information such as connection relationships,

  ie coexistence relationships, meanings, singular and plural forms, "parts of speech", etc. In addition, the analysis rules file 8009 contains data from

  rules for morphological analysis and syntactic analysis.

  The control section 8010 is connected to the section

  8007 Operation Display Section 8007 Operation Display Section

  tion comprises operation keys by which an operator can supply various instructions to the present device, for example a translation instruction key, a cursor key, etc., a display or indicator which visually displays the text of the sentence d input English, Japanese sentence obtained as a result of translation, intermediate data

  such as dictionary information and various instruc-

  for the operator. It can also be set up so that most of these operation display functions are contained in the keyboard if this is placed in the input section 8001, or else in the display, this is arranged. of

  side of outlet section 8014.

  The embodiment of the invention relates to the machine translation device described above, and it is designed so that, if a derived word is contained in the text

  of English 8002, a grammar rule, a peculiarity

  tick, the equivalent (in Japanese), etc., are estimated according to the conditions specific to this derived word, which is recognized as such through an affix, or an analogous element, all this allowing d '' increase the reliability of the analysis result obtained or the translation result obtained. An affix dictionary is used to provide estimates of dictionary information relating to unknown words during morphological analysis. As for treatment methods, three types of treatment must be used, namely treatment relating to

  prefix, suffix processing, and esti-

  suffixed. However, the data include sensitive-

  two types, namely estimation data using the

  prefix and estimation data using the suffix.

  We will first provide explanations for the estimation data by prefix and the estimation data by suffix,

cited above.

  A. Prefix estimation data.

  If the prefix part of a word not entered in the dictionary agrees with what is described below and the remaining part is present in the dictionary, the word is treated according to the dictionary data concerning its root. In 1 20

  Dictionary data, it is possible to add a particu-

  the internal aspect of the row concerning the internal particularity

  initial and add a Japanese suffix to the word for translation

  initial tion. For the input word "electrochemical (electro-

  chemical) "for example, the entry heading of the layout table

  dictionary information reserve constituted by "electro-

  chemical "can be formed, according to the dictionary data, by the entry item of the prefix type" electro (electro) "and by the entry item of the dictionary" chemical (chemical) ", as

*indicated below.

  Input item of the Input item of the prefix type dictionary electro chemical (electro) (chemical) Input word Input item of the reserve table

diction information-

  naire electrochemical electrochemical (electrochemical) (electrochemical) In this way, the entry section of the dictionary information storage table includes all the

  information from the dictionary to which the roots of the words correspond

lay.

  B. Estimated data by suffix.

  If the suffix part of a word not registered in the dictionary is in agreement with what is described below and the remaining part is present in the dictionary, this word is registered by production of a new datum of the dictionary in

  function of the information described in the suffix dictionary.

  In this case, the first equivalent (of Japanese) of the data from the 1 21 dictionary corresponding to the root part of the word is pr4s and used as an equivalent (in Japanese) in the new data.

from the dictionary.

  For example, for the input word "control (control-

  their) ", an entry item from the reserve table

  dictionary information consisting of "controler

  their) "is entered on the basis of the entry section of the suffix type" (e) r (ie in French -eur) "and the section

  input from the dictionary "control".

  Entry item of the Entry item of the suffix type dictionary (verb) - (e) r control ((verb) -eur) (to control) verb name Input word Entry item of the reserve table

diction information-

  naire Controler Controler (controller) (controller) name We will then give explanations on the main lines of the processing of a derived word and the processing of a word

unknown.

  (1) For a word not included in the dictionary, if a prefix or a suffix are respectively present at the head or at the end of the word and the remaining part of the word is registered in the dictionary, the information of "speech part" of English, the internal peculiarity and the equivalence of Japanese are synthesized on the basis of

  Dictionary information and affix information.

  (2) The prefixes and suffixes are respectively arranged in a

  film system and can be produced independently of a pro-

gram.

  (3) It is planned to try The prefix first, then, by

  if it fails, then try the suffix. If both are contained

  naked in the word, the test is not performed.

  (4) In the event that the test estimate of a word has not succeeded, a processing of the terminal part is carried out, as if it

it was an unknown word.

  We will now provide more specific explanations.

  Lières on the eighth embodiment of the invention while

referring to the drawings.

  Figure 58 is a block diagram for explaining an embodiment according to the invention. In this figure, an input processing section 8020, a unit cutting section 8021, a delimiter table 8022, a dictionary consultation section 8026, a derivative word processing section 8023, a dictionary are shown. reference 8024

  and a dictionary information storage table 8025.

  First, an English sentence is read and placed in The sec-

  input processing from an input device consisting of a file of the input document, a keyboard, or an optical character reading device, etc. Then the unit

  dictionary reference is cut out in the discovery section

  dictionary reference units page at the same time that it is

  makes use of the delimiter table and, if it is not successful -

  The dictionary is consulted using the reference dictionary. Then, if there is an entry item that can function as the result of the consultation, the result of the consultation is recorded in the information storage table of the dictionary, while the processing of a derived word is performed if there is no section

entry.

26 0 4 8 1 4

  Figure 59 is a block diagram for explaining an embodiment of the processing of a derived word using a prefix. In this figure, a cross-section 8030 is represented between the part placed at the head and the dictionary of prefixes, a dictionary of prefixes 8031, a dictionary consultation section 8032 used for consulting the dictionary outside the case where it this is the

  prefix part, an 8033 cross-matching section applies

  as for the speech part contained in the prefix dictionary and the speech part of the entry section, an information preparation section of the dictionary 8034

  based on prefix estimation, and a processing section

  derived word ment 8035 based on the use of the prefix.

  Figure 60 is a block diagram for illustrating an embodiment of the processing of a derived word using the suffix. This figure shows a section 8040 between the end part and the dictionary of

  suffixes, a dictionary of suffixes 8041, a processing section

  8042 for completely unregistered (completely unknown) words, a dictionary consultation section 8043 used to consult the dictionary except in the case where it is the suffix part, a section for matching 8044 between the

  speech part of the root and speech part of the ru-

  entry brick for the suffix dictionary, a processing section

  8045 dictionary information preparation ment using

  suffix estimation, a diction consultation section-

  naire 8046 who consults the dictionary by adding the change root cd in the dictionary of suffixes to the part other than the suffix part, and an unregistered word processing section 8047 by estimation by suffix. We first match the terminal part and the dictionary of suffixes. If a match is not obtained, the word is treated as a completely unregistered word (entirely unknown), while, if there is a match, the dictionary is consulted for the

  part other than the suffix part. As a result of the consultation

  tion of the dictionary, if there is no entry item, the search in the dictionary is carried out at the same time as the change of root in the dictionary of suffixes is added to a part other than the suffix part. As a result of the case where there is no entry item, processing for a word not entered is carried out using the estimate by suffix. Whereas, conversely, if there is an entry item, there is a matching between the speech part of the entry item and the speech part of The root in the dictionary of suffixes in the same way only in the case where the heading

  input is present as a result of research in diction-

  nary for the part other than the suffix part. If matching is found, dictionary information preparation processing is performed using suffix estimation, while, if there is no matching, processing for

  word not entered is done using suffix estimation.

  FIG. 61 is a block diagram showing the detail of the general structure obtained by synthesizing the parts of FIGS. 58, 59 and 60, and FIG. 62 illustrates the detail of section 8042 for processing word completely unregistered (quite unknown) which is shown in Fig. 61. The 8042 completely unregistered word processing section which is shown

  in Figure 61, includes a pre-processing section 8050

  estimated information fencing for a name and a section 8051 of

  information preparation processing estimated for a verb.

  However, since each of the parts of FIGS. 61 and 62 has already

  been described in particular, we will omit to provide explanations here

  cations. The ninth embodiment according to the invention will

now be described.

  FIG. 64 illustrates the general structure of the ninth embodiment according to the invention, applied to a device for

  English-Japanese automatic translation.

  This embodiment has an input section 9010, through which an English text 9012 to be translated into Japanese is introduced. The input section 9010 includes a keyboard having character keys, for example alphanumeric keys, and function keys, an optical character reading device (OCR) for reading English text recorded on a sheet of paper, and / or a file memory device for reading English text saved on a

  storage medium, for example a magnetic disc.

  The English text introduced by the input section 9010 is read in a pre-formatting section 9014, where a

  translation pre-processing is carried out. In this case, the recognition

  birth of the sentence and the processing of unknown words are mainly carried out there. The purpose of this section is to

  part of the morphological analysis.

  The pre-formatted English data is transferred to a morphological analysis section 9016 at the same time as the information obtained during the pre-formatting. The morphological analysis section 9016 divides the sentence by consulting a dictionary of words 9018, morphologically analyzes the chain of

  the English sentence, makes various arrangements, such as the

  ment of unknown words, the treatment of proper names, the expression

  time, numbers, etc., and also processes

  ment on the entire sentence, for example for the redundant query (known under the original grammatical name "tag

  question ") and recognition of apposition.

  morphological lysis are contained in a rules file

9036.

  The English data submitted to the morphological analysis are transferred to a syntax analysis section I 9020 at the same time as the dictionary information obtained during the morphological analysis. Section I of parsing 9020 is, in this embodiment, a functional section which analyzes the structure of the surface layer by proceeding from bottom to top and from right to left for the sentence by applying the cfg rule to the data. of English, and it updates all the structural possibilities. After analysis in Section I of parsing 9020, the English data is sent to Section II of parsing 9022 together with the information analyzed. The section selects a solution from the result of the analysis

  relating to the surface layer structure carried out in the sec-

  tion I of syntactic analysis by the application of a description

  structural. In this way, he is prepared for the sentence

  of English a plausible tree of analysis in order to establish its structure

  ture. These parsing rules are also stored

  in the analysis rules file 9036.

  After being subjected to the analysis, the data

  of English are transferred to the processing section of struc-

  ture 9024 at the same time as the data in the analysis tree. In the structure transformation section 9025, the analysis tree relating to the corresponding Japanese sentence is prepared from a structural tree which is an intermediate structure of an English sentence and is transformed into a structure under - adjacent Japanese, from which we can translate a sentence from

Japanese with ease.

  The data from the structural tree indicating the underlying structure of Japanese thus transformed is sent to a translation production section 9026 where the translated sentence is produced. This section is a sentence production function

  of Japanese from a tree structure of the struc-

  Japanese ture. First, the production of a sentence structure is carried out, in which the tree structure is modified by changing the order in order to be in agreement with that of Japanese, then a production of morphemes East

26 0 4 8 1 4

  performed in order to produce a translated sentence following a top-down and left-to-right manner in the tree of

sentence structure.

  The data relating to the Japanese sentence thus produced, that is to say the translation data, are sent to a post-shaping section 9030. The post-shaping section 9030 modifies the translation data by consulting dictionary 9018 based on the information used in the translation processing in order to finally arrive at a more natural Japanese sentence. The data of the Japanese sentence is transferred to an output section 9032, then delivered, as a translated Japanese sentence 9034, by the output section 9032. The output section 9032 comprises for example a printer, a display and, or , a file memory device such as a

magnetic disc.

  The progress of the translation processing series is controlled by a control section 9038 which governs the control of the entire device. The dictionary file 9018 stores lexicographic data for English and Japanese words, and the analysis rules file 9036 stores

  rule data for morphological analysis and syn-

  taxic, in this embodiment.

  The control section 9038 is connected to an operation display section 9040. The operation display section 9040 includes operation keys by which an operator can provide various instructions to the present device, for example a key translation instruction or a cursor key, and a display or indicator that visually displays the English text provided, the Japanese sentence which is the result

  translation, intermediate data such as information

  Dictionary information and various operator instructions. It can thus be constituted that most of the operations display functions are contained in a keyboard, if it is placed on the side of the input section 9010, or

26 0 4 8 1 4

  1 28 well in a display if it is placed on the side of the section

9032 output.

  In this way, in section I parsing 9020, the sfg rule is applied to the English sentence in the mode bottom up, right to left for the English data after

  morphological analysis, so that all solutions are deduced

  options for the sentence structure. The solutions are

  generally understood as a tree structure.

  This shows the relationship for the words or groups contained in each of the sentences which are linked together according to a

  subsidiary relation or of coexistence, like a relation of modi-

  a case relationship, for example a subsi-

  reciprocal diary as for parents, children and grandchildren, etc. Each of the words or groups is located at the node of

tree structure.

  In this embodiment, before parsing,

  distinctions are drawn up for the particularities related

  both in the form and in the vocabulary of a sentence so that a collective arrangement can be decided in relation to the structure of the sentence. It is treated here of the arrangement relating to the structure

  of the sentence in terms of "unit" and "block".

  The "unit" is a word or group of words constituting the minimum unit in the translation process, which is treated exactly as if it were a single word in the analysis, and the dictionary information for each of the constituent elements

  contained in it is not used.

  The "block" is a structural arrangement, for which the analysis carried out is that concerning the interior rather than that concerning the exterior, which is treated in the same way as a unit relating to the exterior of it -this. For example, it could be a phrase, a phrase,

  etc., as well as elements corresponding to inter-

  mediaries used in the grammar cfg. In addition, there can be a reception structure, that is to say that a block can contain another block. In addition, the concept of the block can also include a sentence, a paragraph, whole sentences, which can each be considered as a block. Treatment giving preference to

  Partial analysis is referred to here as "partial analysis".

  This aims to reduce the number of unnecessary structural solutions described above in order to improve the efficiency of

  analysis and obtain a more plausible analysis result.

  Two specific features of the block are defined in one embodiment. One of them is a symbol in the cfg rule, which

  is referred to as a "goal" in the description,

  which must be arranged based on the result of the analysis

  when this has been done for each of the constituent elements

  tifs inside the block, that is to say a symbol describing the structure or attribute of the block. The other is a symbol of the cfg rule, designated here as a "role", which symbol is carried by the block when the analysis concerning the outside of the block relating to the sentence, the phrase is carried out. or the phrase containing the block, i.e. a symbol

  describing the relationship or role of the block vis-à-vis others.

  For example, in the case of the English sentence: I said, "White House isn't white (jedisais," La Maison Blanche n'est pas blanche ")

  the goal is a sentence and the role is a name (phrase).

  While the purpose and role are generally the same in most cases, they can sometimes be different from one of

the other, as in this example.

  For the embodiment shown in Figure 64 ,.

  when the structural arrangement of the input English sentence is recognized as a block, the functional sections responsible for estimating its purpose and role have been summarized in the structure presented in Figure 65. As can be seen in this figure,

  the structural arrangement of the data of the English sentence pre-

  shaped in the pre-shaping section 9014 is distinguished in the morphological analysis section 9016 using the

  word dictionary 9018 and analysis rule file 9036.

  9018 word dictionary stores information

  lexicographic relating to English words and phrases.

  For example, as shown in Figure 68, input headings are formed for all variants of each word in this embodiment and all of the information is created. For

  information relating to the "part of speech" for example, more

  Several speech part information can be provided, as shown in this figure. It will be clear that the manner of constituting the 9018 dictionary is not limited to this

only example.

  The analysis rule file 9036 stores the data relating to the top situation, or head, indicating the top, or the head, of the block, to the end situation indicating the end, as well as the preparation information of block used to endow the block with purpose and role in the form of a table. An example is shown in Figure 69. For example, a block begins with ", conjunction" and ends at the end of the sentence. Therefore, a first block is formed from the head and goes to "," just preceding the conjunction, and the goal is a phrase, while the role is a phrase. In addition, another block is formed from the conjunction and goes to the end of the sentence, where the goal and the role are both a

phrase.

  In addition, a block begins with ", relative pronoun" and ends with "," or the end of the sentence. As in this case, the possibility that there are several end conditions is accepted for a head condition. In the case where the block ends with ",", a cluster starting from "," just before the relative pronoun and going until the next appearance of "," forms a block in which the

  but is a phrase and the role is an adverb or an adjec-

  tif. Thus, this means that the cluster functions as an adverbial cluster or an adjectival cluster. In the case of a termination at the end of the sentence, a cluster starting from "," just before the relative pronoun to go to the end of the sentence constitutes a block, in which the goal is a phrase and the role is an adverb or a

2 6 0 4 8 1 4

  adjective. All this is presented according to the conditions of

  formation of the phrase, the phrase or the phrase appa-

  falling into ordinary modern English phrases. On the

  figure, the symbol, U "represents a spacing.

  In this way, in the morphological analysis section 9016, the English text introduced by the pre-shaping section 9014 is first divided into sentences constituting the translation units. In this case, Incorrect spellings and unregistered words are detected. The 9018 dictionary is consulted for

  all Phrase units and related dictionary information

  tive to each of the constituents is extracted. Various agency modes-

  are based on this information from the diction-

  nary. Figure 66 shows a flowchart for the layout

  collective of a block performed in the morpholo- analysis section

  gic 9016. First of all, a position pointer indicating the position read for an English phrase is positioned on the top, or at the head, (9100). The head position designates not the word se

  finding at the head, but the (imaginary) head of the sentence imma-

  diatly previous. Word picking processing 9001 is performed at the position. As shown in FIG. 67, in the word picking process 9101, a word is picked by advancing the position by one step (9111) unless it is at the

  end of the sentence (9110), and consult the dictionary 9018 related

  word (9112) to write the word information (9113).

  In this way, When the word information is pre-

  thrown in the word picking processing 9101, the table 9036 relating to the head and end conditions of the block is requested, so that it is determined whether or not one exists which satisfies the head condition (9102). In this way, steps 9101 and 9102 are repeated until the word which satisfies the condition of

head has been detected.

  When the head condition is satisfied, the word imm-

  next diat and the succeeding words taken in number

  wanted are examined in order to verify the agreement with the condi-

  block head (9104). In this case, the dictionary is consulted for each of the words if necessary. The pointer

position does not advance.

  If the match takes place with the head condition of the block in step 9104, then a word which satisfies the end condition of the block, having regard to the head condition, is searched (9105). The steps 9014-9106 are repeated cyclically until the word which satisfies the end condition has been discovered. When a word satisfies the end condition (9106), a cluster containing the word is recognized as a block, and the block is written (9107). More specifically, a block is prepared based on the fact that it is judged that the condition for preparing the block is satisfied at the position where the end condition has first been satisfied. Then, with reference to the block preparation information table 9036, the position relative to the word indicated by the pointer at the position where its advance was stopped in processing 9103 is defined as being the head position of the block, and the position satisfying the end condition appearing first after that is defined as the end position of the block. Simultaneously,

  the purpose and role of the block are also written.

  As a result of this recognition of the block, if there is "..., conjunction ..." in the English sentence for example, as shown in FIG. 70, the cluster starting from the head of the sentence and going to the preceding part "," is recognized as being a first block, while the cluster which starts from ", conjunction" which

  goes to the end of the sentence is recognized as another.

  block. In the figure, what is inside the signs E] indicates a block. In this block, the goal and the role are both a

  phrasing. In addition, a cluster going from the word that comes after the conjuction-

  tion until the end of the sentence forms another block, in which

  The purpose and role are also unephrase. According to another possibility

  bility, the cluster going from the conjunction to the end of the sentence can be defined as a block. In this case, the goal is a member

  and the role is an adverb.

  The block can also be defined as starting from the position not containing ",". In addition, it is possible to exclude from the scope of the Analysis punctuation, or elements

  analogues, under the information possessed by the block.

  In the same way, if there is "... relative pronon ...", ", relative pronoun ..." can be recognized as a block. In this block, The goal is a phrase or phrase and the role is a

adverb or adjective.

  The block can naturally be a reception structure.

  For example, if the English sentence has a structure such as "(head of the sentence) ..., conjunction, ..., relative pronoun ... (end of the sentence)" as shown for example on Figure 71, the cluster starting from ", conjunction" and going until the end of

  the sentence constitutes a block BL1-BL1, in which ", pronoun rela-

  tif ..., "is another BL2-BL2 block contained in the first block.

  In this way, the morphological analysis section 9016 distinguishes The particularities of The sentence with regard to the

  form and vocabulary so as to be able to specialize vis-à-vis

  screw the rest of the structural arrangement as a block. Furthermore, in

  In addition to this block recognition, the morpho- analysis section

  logic 9016 also performs various processes such as the expression of a proper name, a derived word, an unknown word, an abbreviation, a numeric expression, a temporal expression, a word with hyphen, an apostrophe "'", as well as the estimation of an apposition and the processing of a redundant query, in order to prepare the morphological analysis data. The English sentence which has thus been subjected to a morphological analysis is transferred to the syntax analysis section I 9020 together with the analysis information. Figure 72 shows an example of the output data. This figure shows the result obtained when the English sentence: I said, "White House is'nt white" I said, "The White House is not white")

2 6 0 4 8 1 4

  was provided by The input section 9010 and analyzed in The morphological analysis section 9016. Block I starts at the word position n 4 and ends at the position no 10, in which case the goal and the role are all two, in this case, of a nature to choose. Similarly, block 2 starts at position n 5 and ends at position n 6, for which the goal is a nominal group, while the role is a proper name. Thus, the block "White House is'nt white (The White House is not white)" contains, as a reception structure, another block "White House (White House)". In a block, that is to say the smallest "White House" block, each of the constituents placed inside acts as a proper name, while it has a nominal phrase position with respect to the exterior, namely "is'nt white. (is not white.)". "White House" can be treated as

a unit.

  Along with a similar block of information, the word information found in the word dictionary 9018 is, after being added to the block information, sent by the morphological analysis section 9016 to section I analysis

syntax 9020.

  Section I parsing 9020 analyzes the structure

  surface ture of the English sentence by applying a cfg rule stored in the analysis rule file 9036 in order to update all possible structural trees. In this case, if a block is contained, the partial analysis described above is

  carried out while preference is given to local analysis.

  This can improve the performance and accuracy of the analysis.

  More particularly, the inclusion relationship of the block is prepared from the position information relating to the block. Then, the innermost block is analyzed. The block for which the analysis has been completed is considered as a unit and no more processing is carried out inside

  this one. In this way, the scope of the analysis gradually extends-

  towards the neighboring neighboring blocks. Finally, the whole sentence is analyzed. Analysis is performed on the basis of

26 0 4 8 1 4

  cfg rule following a mode down at the top, right to left in the English sentence. The analysis is carried out in such a way that it is possible to maintain all the possibilities offered by the

grammar rules.

  Figure 73 illustrates an example of such an analysis processing flowchart. First, on the basis of the English data provided in Section I of parsing 9020, all the structural arrangements relating to a sentence are recognized as being blocks, and there is an estimate of the purpose and the role 9120. The arrangement mode is as shown in the

  figure 70. Then, if there is no block present in such an arrangement

  cement (9121), the sentence is analyzed (9125), and a single sentence collectively arranged as a symbol is selected, and the analysis of the sentence is completed (9126). Since processes 9125 and 9126 are included in processes 9121-9124 if the whole sentence processing system in one block is used,

so they are not useless.

  If a block is present, the innermost block is analyzed first (9122). In the example presented in FIG. 71, the interior of the block BL2BL2 is analyzed. While various solutions are generally obtained by analysis, a solution in which the block is collectively arranged in a single symbol cfg and which is in agreement with the purpose of the block is chosen from these solutions (9123). In this case, for those who have a "choice" goal relative to the block, all those who are arranged in a

  single symbol are selected. Then those who have been selected

  treated in this way are treated as a single arrangement having the role relating to the block (9124). In the block that has

  a role called "at choice" a role of the symbol which is collected collectively

  in treatment 9123 is defined as the role.

  Treatments 9121-9124 are repeated successively.

  In this way, in the example of figure 71, the inte-

  the block BL2-BL2 is first analyzed, then the interior of BL1-

  BL1 is analyzed. In this case, the BL2-BL2 block is treated the same

  so that a single word, and none of the constituents that compose it

sent is not analyzed.

  In this way, when the data defining the structural arrangement and the subsidiary relationship is obtained, it is sent to the syntax analysis section II 9022. The data can be easily recognized in the form of a structure tree, as shown below. - described above. Data is brought back to

  the Japanese sentence structure in the transform section

  structure 9024 and in the translation production section

  tion 9026, a translated sentence is produced for each of the nodes it contains. Treatment of structural tree nodes

  is performed in the top to bottom mode, left to right.

  The translated sentence thus produced is subject to a post-

  formatted in the post-formatting section 9030, is visually displayed in the operation display section 9040 and, for example, is printed as a sentence of

* Japanese 9034 in the output section 9032.

  In this way, according to the present embodiment, the peculiarities of the English sentence relating to form and vocabulary are distinguished so as to be able to detach the structural arrangement in the form of a block. For the block, a goal which can be the result of the analysis and the structural role with

  which block works vis-à-vis the outside are estimated.

  Next, the surface layer structure of the English sentence is analyzed by applying a cfg rule so that all possible structureIs trees are updated. This allows

  reduce the number of unnecessary solutions and improve the return

  analysis as well as providing an analysis result

More reliable.

  In this way, since there are various configurations

  tions for the expression of apposition, it is difficult to recognize them in the analysis, in particular in the analysis of the cfg type. In these conditions, since it is generally difficult to recognize the apposition after the analysis, a

  ambiguous translation is inevitable. In addition, if a rule suscep-

  tibLe to recognize them was prepared, there would be a risk that no expression of apposition would be recognized as an identical case or that the number of possible combinations would become excessive. Thus, an unnecessary local analysis is carried out between the parts contained in the expression in apposition and the other parts. In relation to the present embodiment, the load imposed on the processing in the analysis step can be softened by recognizing the expression of the apposition thanks to the peculiarities of the sentence relating to the form or

  to the semantic peculiarities of words. An estimate of the appo-

  sition is performed by recognizing the configuration

  immediately following as constituting a block.

  For the English sentence structure "-, relative pronoun L._", the relative pronoun is recognized due to the existence of the code

  of "speech part" of the word, for example a specific code "R".

  In this case, the inner part framed by "," is considered as a block provided that there is no intersection with a block or a unit indicated in the pre-layout and that they do not contain "and (and)" or "or (or)" in the part coming after the second ",". For the English sentence structure "_, relative pronoun-.", The inner part surrounded by "," and "," is considered as a block. The point could be another

  symbol used to indicate the end of the sentence.

  To allow a similar estimate to be made

  apposition; dictionary 9018 is designed to store.

  siner a word meaning information. The significance information represents a means of discerning the article, the place, the person, etc., as shown in FIG. 74. In addition, also for the conditions of preparation of the block, the table 9036 is constituted in the manner shown in Figure 75 so that the head of the block is recognized by "proper name (person), name (person)" as the head condition, and the head of the block is U recognized by "proper name (person), item name (nobody)". It is therefore possible to produce an estimate of the expression of the apposition in relation to the morphological and

  without having to perform the analysis and perform the analysis

  lysis in consideration of the apposition estimate in

  the example presented in figure 64 for other treatments.

  In addition, in the English sentence, there are groups which carry extremely specific information and are used only in a restrictive manner. If they are analyzed in the same way as ordinary groups, they are analyzed in The sentence in an entirely different way and it is difficult to obtain

  your initial particular nature of the sentence using analysis.

  In addition, this results in significant losses.

  For example, "let's" or "let us" coming just after punctuation is analyzed as a sentence in the imperative using the

  causative verb "let" and it should be analyzed as a possessed group

  dant the particularity of the invention.The word "Let" also has various uses as a transitive verb and as a noun and its use is not limited to invitations as is the case for auxiliary verbs. Therefore, the analysis should be conducted on the basis of the respective respective possibilities, which reduces the efficiency. In addition, it is difficult to give preference to use as an invitation from the result of the analysis, since there is no difference between use in the causative sense and in the sense of the invitation based solely on the structure of the sentence and therefore it is difficult to

  distinguish them on the sole basis of the structure of the sentence.

  So, we can reduce the time lost in performing the analysis by excluding from the subject of the analysis "let's" or "let us" just after the punctuation. Moreover, by separating them from the fundamental use of the word, namely from the causative use, one can easily

perform semantic analysis.

  When "please", "let's" or "let us" appear at the head of the block, a flag is positioned, then

  block information and everything related to related information

  tive to the unit in each case are not issued. For example, the English phrase "let's go to school."

2 6048 14

  school.) "is treated as <go to school>

<attached to "let's">.

  To perform a similar analysis of "Let", iL is provided a section 9200 of information processing relating to "let" between the morphological analysis section 9016 and section I

  parsing 9020 according to a variant of this embodiment

  tion. Figure 77 collectively illustrates the sections

  derées. In these figures, the elements identical to those presented

  in Figure 64 are designated by the same reference numbers.

  In addition, the 9018 dictionary is constructed to store the "let" information for the word. As shown in Figure 78, the information on "let" provides "0" for ordinary words, "1" for "let's" and "let us", and "2" for "please (if

Please)".

  The function of the information processing section 9200 on "let" is to receive the result of the morphological analysis, at the same time as the input English sentence, from the morphological analysis section 9016 and to add the information on "let" as additional information to the information of a word during the analysis, as shown in FIG. 79. In this case, a block is presented for the sentence. In the example shown in this figure, block 0 is (start: 1, end: 10, goal: sentence, role: sentence). Thus, the block of this example includes for example a sentence in addition to a phrase, a group, etc. In this case, the concept of the block also includes a paragraph and the entire sentence, each of which can be considered as a single block. In addition, "transitive verb root (attached to 's)"

  is described as the speech part of "let's" for the informa-

  tion of the word, and the information on "let" is "1".

  As shown in FIG. 81, a processing for preparing a block of the sentence 9300 is carried out before the collective arrangement of the block for the sentence has been started.

  Entrance English. The following treatments can be

  ticks to those in the flowchart in Figure 66. For example, in the English sentence,

26 0 4 8 1 4

  I said, "Let's go to school. (I said, let's go to School.)", Block 0 (start: head of the sentence, end: end of the sentence,

  role: sentence, goal: sentence) is formed.

  As shown in FIG. 81, in the syntax analysis section I 9020, each of the structural arrangements is recognized as being a block on the basis of the English data which are delivered to it and the purpose and the role of this block are estimated (9120). If the block is not present in the layout (9121), the analysis ends. If blocks are present in the sentence

  first, the innermost block is first analyzed (9122).

  While various solutions are generally obtained by the ana-

  lysis, only the solution collectively arranged as a cfg symbol is retained among them (9123). The following treatments are

  identical to those presented in figure 73.

  Information processing on "Let" is carried out in section 9200 of information processing on "let" according to the flowcharts presented by way of example in FIGS. 83A and 83B. First, a pointer is positioned on the head block (9330), so that the word placed at the head of the block (9331) is checked. If the information on "let" is "0", the pointer advances by one

  notch (9339) to be transferred to the next word.

  If the information on "let" is not "0", the reference unit of the previous dictionary is checked (9322). If it is not punctuation or if the pointer does not indicate the head, the pointer advances by degrees (9339) to transfer to the next word. When the previous dictionary reference unit

  turns out to be punctuation, or if the point-

  tor indicates the head, the innermost layer block

  containing the word is assigned a mark (9333).

  Then, if the information on "let" is "1" (9334), since it is the expression "let's" or "let us" coming just after the punctuation, the role of the block identified is recognized as being an "invitation phrase" (9336). If the information is "2", since this corresponds to "please (please)", the role of

  block identified is recognized as a "request phrase" (9335).

  Then, the goal of the identified block is recognized as being a sentence in the imperative (9337) and the information relating to the word indicated by the pointer is canceled (9338). Then, the pointer advances by one notch (9339), in order to transfer to the next word. The treatment is

  executed for the word in the final position (9340).

  Figure 80 shows examples of

  result of the analysis for which the information processing on "let" was executed in application to the example of the above mentioned input sentence I said "let's go to school. (I said, let's go to the 'school.) "Section 9200 for processing information on" let ", when adding information on" Let "to the word information, eliminates the word information about the" let "in the table, and the block information is described as being an" imperative sentence "regarding the goal and as an" invitation sentence "regarding the role, as shown in the

figure 80.

  In addition, in the case where the dictionary 9018 is consulted for the set formed by several words coupled by hyphens, when one is in the case where words linked by hyphens are manipulated in the sentence of English, if their

  entry headings exist in dictionary 9018, the

  ment succeeds. For a hyphenated word not entered in the dictionary 9018, if the whole part is treated as being an unknown word, for example an adjective, since the dictionary information for each of the words coupled by hyphens belonging to the group of words linked by hyphens cannot be used, it cannot be translated. In addition, if a

  entry section for information relating to each of the

  kills from the group of words linked by hyphens is present in the dictionary 9018, it is not possible to do otherwise than to neglect them. In addition, if the analysis has been carried out by decomposition into each of the constituent elements, the coupling of the words linked by hyphens is then likely to take

extremely diverse forms.

  In view of what has been stated above, to eliminate the problem, the set formed by words linked by hyphens is analyzed as constituting an adjective in the sentence,

  at the same time as the analysis is carried out for the only integrated part

  of words linked by hyphens, the elements

  kills of the set formed of words linked by hyphens and

  the assembly thus obtained then being combined. This allows for

  kill the analysis of words linked by hyphens while using

  information relating to each of the constituent words. Thus, for words linked by hyphens which are not listed in the dictionary 9018, the set formed by all these words is treated in the same way as an adjective. Names linked by hyphens are searched for in the dictionary and the analysis is carried out in a closed form for the sole internal content of the

  group of words linked by hyphens.

  Thus, if the hyphenated word is a word not entered in the dictionary 9018, block information is delivered which relates to the entire group of words considered as a single block and the dictionary is consulted for each of the constituents of the interior of the block in order to obtain the information relating to the respective units, without taking account of the hyphens. For the word not entered in the dictionary, a final estimation processing is carried out among the processing of unknown words in the manner of sorting the processing of unknown words. This treatment of the hyphens can be carried out according to the example of the structure presented in FIG. 64. In this case, the position of the word in the sentence is not expressed by the number applied to the word, but by the number of characters counted from the head of the sentence, i.e. the number of characters. Figure 84 shows an example of related word processing

  by hyphens performed in the morpho- analysis section

  Logic 9016. For an input English sentence such as, for example

  example, "The anti-war attitude is her open-door policy.

  anti-war study (that is to say pacifist) is its policy of open door (that is to say of free access or free exchange).) ", The position pointer advances by degrees to bring out a word (9135) and searches the dictionary (9353). In this case, the hyphen is not used as a

delimiter for the word.

  If the entry item exists (9353), the information relating to the word is then written (9359). This is repeated until the

end of sentence.

  If the entry is not found when consulting the dictionary (9352), if it is not a word containing a hyphen (9354), the information relating to the word is then written (9359), while, if it is a word containing a hyphen, the block with the hyphen is written (9355). In the block with the hyphen, the starting position is the initial position of the hyphenated word and the

  end position is the final position of the hyphenated word.

  The goal is of the "choice" type and the role is "adjective / noun".

  Next, the hyphen is detached to extract each of the constituent words (9356), and the respective constituent words are searched for in the dictionary (9357). The information relating to the word obtained as a result of consulting the dictionary (9358) is then written. For the case of writing the word information in steps 9359 and 9358, it is written in the form "part of speech = word not entered in the dictionary" in the case of a

  word not listed in the dictionary.

  FIG. 85 shows examples relating to the block information and to the word information of the block of English collectively arranged in a block following the processing of the example of the

  input sentence. In this example, the hyphenated word "anti

  war (anti-guerre) "is inscribed in the dictionary 9018 and the word

  "open-door" is not listed in the dictionary.

  Therefore, the entry box for the hyphenated word "anti-war" is then written as word information. However, for hyphenated words "open-door ', they are broken down into" open (open) "and" door (porte) "and written as information for the words, and block 1 (start: 30 , end: 38, goal: optional, role: adjective / noun) is written as information

for the block.

  Furthermore, while the form of the question

  Dante of English is extremely limited, its treatment is extremely complex by the ordinary method of analysis. Of

  moreover, it is not easy to determine the verb to which the question

Additive gation relates.

  Having regard to what has just been stated, after It has been recognized that this was a redundant questioning on the basis of the particular aspect of the sentence considered from the point of view of its form, that -this is treated as information relating to the structural arrangement to which it belongs, thanks to which the verb targeted by the redundant query can be specified, that is to say that the part of the redundant query of the sentence of English is discovered as being a structural configuration and the analysis is carried out by considering the part of the redundant query as being simple information which

  has a certain type of structureL arrangement belonging to it.

  In the present embodiment, a unit or a block is described in the form of a symbol (from the point indication that it is an end point of a unit or block). The text of the input sentence is conformed during the morphological analysis during which the block recognition is also carried out. In the present embodiment, the C sign "quotation mark" is designated by "Q" and the sign of the brace is designated by "P". For example, these signs are specified in the following manner: .... 'by \ (Q' .... \) ', "...." by \ (Q ".... \) ", (....) by (\ (P .... \)),

<....> by <\ (P ....

by L \ (P .... \)} and

  E ....] by E \ (P .... \)] respectively.

  The block is therefore recognized in this way.

  The start symbol and the end symbol of the block are only applied in the context where the block opens or closes using these symbols. The part immediately preceding the start signal and the part immediately following the end signal

  must be symbols other than alphanumeric symbols.

  The above symbols which do not correspond to the above case are treated as simple symbols. The blocks can sometimes be in host relationships insofar as they do not present

no intersection between them.

  During the processing of the redundant query, if the group of words coming next follows a situation for which the pointer indicates ",", the cluster which leaves. after "," and who goes

  until "?" is rejected as a unit and a flag is posi-

  tioned as for a block. So the form of the phrase

  redundant gation includes: ", (auxiliary verb) + (personal pronoun)?" ", (auxiliary verb) n't + (personal pronoun)?" ", (auxiliary verb) + (personal pronoun) + not (i.e.

  negation particle used for conjugation)? ".

  In addition, auxiliary verbs include types such as: "am, is, are (forms of conjugation of the verb to be in the present)", "was, were (forms of conjugation of the verb to be in the past)", "do,

  does, did (verbal particles of negation and questioning

  tion) "," have, has, had (forms of the conjugation of the verb avqir in the present and the past) "," will, shall, would, should (verbal particles of the conjugation of the future and the conditional) "," can , cannot, could (forms of the verb power in the present (affirmative and negative) and the past) "," may, might (forms of another verb power in the present and the past) "," must, ought (various forms of the verb expressing obligation) "," won't, shan't (future and conditional particles in negative form) "," need (verb to need) "," dare (verb to dare) "," used ( verb form used to express habit) ". Personal pronouns include" I (I) "," you (you / you) "," he, she, it (he or she respectively for Man, Woman and animal or object) "," we (we) "," they (they

or them) ". All this is used as information for the block of the layer located

  the most inside to which these expressions belong. For example, in the English sentence: you said so, didn't you? (you said that, right?) this whole part is recognized as being a block in relation to the structural arrangement EYou said so (you said that)] <with redundant interrogation>. Similarly, in the English sentence: I said, "You said so, didn't you?" (I said, "You said that, didn't you?" the quotation mark "You said so, didn't you?" is

  recognized as being a block 1 on the basis of the structural arrangement

  turel and, moreover, the whole of the part is recognized as being a block 2 on the basis of the structural arrangement, that is to say [I

  said, [you said so,] <attached to a redundant query>.

  The abbreviated word "didn't" is processed after being developed

  loped to its full spelling form using a predetermined table. In the case of words having several forms

  developed, they are all supplied.

  To perform such interrogation processing

  redundant, a redundant query processing section 9210

  dante is arranged between the morphological analysis section 9016 and the syntax analysis section I 9020 according to another variant

  of this embodiment. Figure 87 shows collecti-

  the sections considered. In this figure, the elements identical to those presented in Figure 64 bear the same

reference numbers.

  The redundant interrogation processing section 9210 receives the result of the morphological analysis at the same time as the

  entry English sentence from the Moral Analysis Section

  phological 9016 and, as shown in figure 88, a block is made for the sentence. In the examples presented in this figure, block 0 is (start: 1, end: 12, goal: sentence, role: sentence). In this case, the word is represented by the word number for this modified embodiment). In this modified embodiment, the block includes for example a sentence in addition to the phrase and the group. In this case, the concept of the block also includes the paragraph and the entire sentence, which can

  each be considered a single block.

  The collective arrangement of the block of the input English sentence with the redundant query can be

  the same as that presented in the flowchart of figure 81 -ci-

  above described. Thus, the processing 9300 used to prepare the block of the sentence is carried out before the start of the processing. For example, in the English sentence:

  I said, "It is good, isn't it?" (I said, "It's good, isn't it-

  not?) the O block (start: head of the sentence, end: end of the sentence,

  role: sentence, goal: sentence) is formed.

  In section I of parsing 9020, the analysis is executed according to the same flowchart as that presented on the

figure 82.

  We will now explain the processing carried out in the processing section of the redundant query 9210 using FIGS. 90A and 90B. First, a pointer is positioned

  on the word at the head of the word information (9370).

  If it is not a comma, the pointer moves up a notch

  (9384) and it repeats until the end of the sentence (9371).

  Then, it is checked whether the word coming after the decimal point is a word belonging to group c or the word belonging to group e, while the pointer remains in the position it occupied (9373, 9379). In this case, it is defined that the words containing an auxiliary verb or the verb "to be" in the "speech part", when it is not a question of the negative form, are words belonging to the

26O4814

  group a, while those which contain the negative form of the auxiliary verb or the verb "to be" in the speech part are words belonging to group e. If the word does not belong to any

  of the two groups, the pointer advances by one notch (9484) and the operations

  tions repeat until the end of the sentence (9371).

  In the case where the word belongs to group a, step 9384 at which the pointer has advanced is executed if the word following the words of group a is not a pronoun. If it is a pronoun, it is checked whether or not the next word is "not (negation particle used in conjugation)" (9375) and, if it is not "not" , it is determined whether or not the word following the pronoun is a question mark (9377). If it is not the interrogation anoint, operation 9384 to which the pointer has been advanced is executed. If it is the question mark, the goal is rewritten in the form "negative sentence" and the role in the form "redundant question sentence" for the block of the layer most inside ( 9378), while ", ...? Is rejected from the word information table (9383). The block in the innermost layer indicates blocks satisfying The condition: start position 4 ( position of ","), as well as the condition: final position> (position of "?") for the position

  of the block, with a minimum interval for (final position -

initial position).

  If the word following the pronoun is "not" in step 9375, it is then checked whether the next word "not" is or not a

  question mark (9376). If it is not an information point

  terrogation, operation 9384 to which the pointer advanced is executed. If it is a question mark, the goal relating to the block of the innermost layer is rewritten in the form "affirmative sentence", while the role is rewritten in the form "sentence redundant query "(9382), and", ..? " East

  rejected from the word information table (9383).

  In step 9379, if the word following the decimal point is a word belonging to group a, operation 9384 until which the pointer has advanced is executed if the word following the group

26048 14

  is not a pronoun. If it is a pronoun, it is determined whether

  the word immediately following it or not is a point of inter-

  rogation (9383) and we go to step 9384 with which the pointer has advanced if it is not a question mark. If it is a question mark, the purpose of the innermost layer block is rewritten as "affirmative phrase",

  while the role is rewritten in the form "sentence of interroga-

  redundant tion "(9382), and", ...? "is rejected from the information table

  mation on the word (9383). Then the pointer moves up a notch

  (9384), and the operations are repeated until the end of the sentence.

  As an example, Figure 88 shows the block and word information provided by the morphology analysis section 9016 to the redundant query processing section 9210 for the English sentence:

  I said, "It is good, isn't it?" (I said, "It's good, isn't it-

  not? ") described above. The block information for block 1 is (start: 4, end: 12, goal: optional, role: optional). If it is subject to redundant query processing in the redundant interrogation processing section 9210, the block information relating to block 1 is rewritten in the form (start: 4, end:

  12, goal: affirmative sentence, role: redundant question sentence

  dante) and, at the same time, the word information relating to the redundant query going from number 8 to number 11 is

abandoned.

  Of course, those skilled in the art will be able to imagine,

  from the device whose description has just been given to

  title merely illustrative and not limiting, various

  variants and modifications not departing from the scope of the invention.

  We will have understood that the term "attitude", sometimes used,

  must be understood in the same sense as the term "goal".

260 4814

Claims (30)

  1. Language analyzer, characterized in that it comprises a dictionary (1018) in which are stored dictionary data containing morpheme data for words, compound words and phrases, and an analysis means (1016 ) making it possible to carry out a morphological analysis of an input sentence (1012) by referring to the dictionary, o: the dictionary contains data relating to the degree of coupling indicating the degree of coupling between each of the words constituting the compound words and the phrases, and   said means of analysis refers to the related dictionary   the respective words contained in the input sentence and, when several data from the dictionary are consulted for a   single word in combination with other words, it chooses the combination   naison of words having the highest degree of coupling in se   referring to the data relating to said degree of coupling.   2. Analyzer according to claim 1, characterized in that, when the data relating to the degree of coupling indicate a high degree of coupling for several combinations between a word found in the dictionary and other words, the analysis means prepares preference data in order to give preference to   one of said combinations based on a certain rule.   3. Language analyzer, characterized in that it comprises: an input means (2014) used to introduce a set of characters of a predetermined language, a fundamental dictionary (2024, 2026) used for searching for said set of characters entered and in which fundamental data are stored, and an analysis means (2016) making it possible to analyze said set of characters introduced by consulting the fundamental dictionary, where:   said means of anaLyse consults the basic dictionary   tal relative to said introduced character set and, if a part of said character set is found by him, he consults the fundamental dictionary in the same way for other parts of said character set in order to analyze said set of characters.   4. Analyzer according to claim 3, characterized in that   the fundamental dictionary is a dictionary of basic units   tales having data which allow to express dimension units stored in the dictionary, and in that the means   of analysis consult the dictionary of relative fundamental units   to said set of input characters so as to thus analyze whether or not said set of characters expresses a unit of dimension.   5. Analyzer according to claim 4, characterized in that the analysis means consults the dictionary of fundamental units   relative to the character set and, if the character set   tères consists only of the combination of sets of characters stored in the dictionary of fundamental units   which expresses a unity as a result of the consultation, it deter-   mine as a set of characters used to express unity.   6. Analyzer according to any one of claims 3 to   , characterized in that the analysis means has a pointer, said pointer is positioned on a character placed at the head of the set of characters entered and, if part of said set of characters is found by consulting the fundamental dictionary relatively to the set of characters from the character on which the pointer has been positioned, said pointer is positioned on a set of characters following the part of said set of characters found, then the fundamental dictionary is again consulted in relation to the together   next characters, on which said pointer was positioned.   7. Analyzer according to any one of claims 3 to   6, characterized in that the set of characters entered is found using a standard dictionary and is not stored in said dictionary.   8. Language analyzer, characterized in that iL comprises:   a dictionary (3018, 3104) in which are stored   dictionary data relating to all the reference units of the dictionary, and an analysis means (3016) making it possible to divide an introduced sentence into reference units of the dictionary and to carry out a morphological analysis of said reference units of the dictionary while with reference to said dictionary, o: the dictionary contains, under said dictionary data, an indication of distinction indicating that said dictionary reference units represent numbers for dictionary reference units representative of numbers, and said means of analysis refers to the dictionary for the respective dictionary reference units contained in the input sentence and, if said distinction indication is contained in said dictionary data found, it combines the dictionary reference unit for which said indication   of distinction was found with a reference unit of the dic-   present in the vicinity of said dictionary unit and for which another said indication of distinction has been found, then it calculates the numerical values which are signified by the two said dictionary reference units in order to form a single numerical value and forms the reference units   of the dictionary in a single unit of analysis.   9. Analyzer according to claim 8, characterized in that the analysis means arranges said analysis unit, if it is accompanied by a reference unit of the dictionary expressing a   currency symbol or unit of dimension, with the numerical value   risk to form a single unit of analysis. 26048 14   10. Language analyzer, characterized in that it comprises: an input means (4014) used to introduce a set of characters of a predetermined language, a dictionary (4020) used to search for said set of characters introduced by said entry way,   a means of consultation (4022) which consults the diction-   relatively to said set of characters introduced, and a means of producing particularity information, which produces specificity information of a   set of characters not recorded in the dictionary   tionnaire and a set of characters for which information   feature of particularity has not been recorded in the dictionary, among said sets of characters introduced as a result of the search carried out by said means of consultation, o: said means of production of particularity provides several information of particularity to said set of   characters without specific information.   11. Analyzer according to claim 10, characterized in that it further comprises: means for cutting units used for dividing   the set of input characters in units for consultation   tion of the dictionary, and an analysis means used to analyze all of   characters divided by the means of cutting units after consultation   tation of the dictionary by said means of consultation, at the same   time than the previous set of characters.   12. Analyzer according to claim 11, characterized in that, when the particularity information is present for only one of the sets formed by said set of characters and the set of characters preceding it, and that no particularity information is present. is present in the other set of characters, the analysis means provides the specific information of said first set of characters said set of characters   not having specific information.   13. Analyzer according to claim 11, characterized in that, if specific information is present in each of the sets of characters, namely said set of characters and the preceding set of characters, the analysis means provides information of common particularity as particularity information of said set of characters and   from the previous character set.   14. Analyzer according to any one of claims 11 to   13, characterized in that the set of characters and the preceding set of characters which are analyzed by said analysis means are proper names.   15. Analyzer according to claim 11, characterized in that, if the set of characters begins with a character in capital letters and that the preceding set of characters is not at the end of the sentence, the means of analysis transforms the capital letter at the head of said set of characters into a lowercase letter, then consults the dictionary using said means of consultation, and, if it does not appear as entered in the dictionary following said consultation , the analysis means analyzes said set of characters as being a name own not registered.   16. Language analyzer, characterized in that it comprises: a dictionary (5018) in which dictionary data relating to all the reference units of the dictionary are stored, and an analysis means (5016) making it possible to divide into dictionary reference units an introduced sentence and perform   morphological analysis by referring to the relative dictionary-   ment to said dictionary reference units, o: -   the dictionary contains, under the said dictionary data, information of distinction making it possible to specify the position at which several proper names are authorized to be found in a group of successive proper names for a unit for consulting the dictionary which indicates a proper name, and   Said means of analysis refers to the related dictionary   to the respective dictionary reference units which are contained in the entered sentence and, if Said distinction information is contained in the dictionary data found, it combines the dictionary reference unit for which said distinction information was found with the dictionary reference unit which is in the vicinity of said dictionary reference unit and signifies another proper name so as to form a single unit of analysis according to the   position specified by said distinction information.   17. Language analyzer, characterized in that it comprises: an input means (6014) used to introduce a set of characters of a predetermined language, a dictionary (6020) used to search for said set of characters introduced by the input means, and a feature information analysis means (6024)   who consults the dictionary with respect to said character set   ters introduced, then analyzes the particularity information of said set of characters, o: said means of analysis of particularity information analyzes the particularity information of said set of characters while taking into consideration the particularity information of the sets of characters before and after said set of characters. -   18. Analyzer according to claim 17, characterized in that the feature information analysis means analyzes the feature information of several sets of characters   by arranging collectively said sets of characters.   19. Analyzer according to claim 17, characterized in that, if the particularity information is present for only one of the sets formed of said set of characters and of a set of characters preceding it, and that the particularity information is   not present for the other, the means of parsing information   particularity provides the particularity information of said first set of characters, The set of characters not having specific information.   20. AnaLyser according to any one of claims 17 to   19, characterized in that the set of characters analyzed by said particularity information analysis means is a proper name. 21. Language analyzer, characterized in that it comprises: a dictionary (7018) in which are stored   dictionary data on all dic- reference units   tionnaire, and an analysis means (7016) making it possible to divide into dictionary reference units an introduced sentence and to carry out the morphological analysis relating to said dictionary reference units by consulting the dictionary, o: said analysis means distinguishes that an estate   dictionary reference units with a semantic element   particular tick is an assembly unit expressing a signi   specific specification consisting of a certain rule and forms a succession of reference units of the dictionary having said   specific semantic element in a single unit of analysis.   22. Analyzer according to claim 21, characterized in that the dictionary contains data making it possible to distinguish the reference units from the dictionary having a particular semantic element,   the analysis means includes a matching table   weighting of distinction making it possible to distinguish that a succession of reference units of the dictionary having said specific semantic element is a unit of assembly expressing a specific meaning composed according to a certain rule, and the means of analysis consults the dictionary relatively to the respective dictionary reference units contained in the sentence entered and forms the succession of dictionary reference units having said specific semantic element into a single analysis unit by performing a mapping using the matching table in correspondence of distinction   when it has been distinguished from the diction reference unit -   nary having said specific semantic element.   23. Language AnaLyser, characterized in that a particular   Grammaticality, a semantic particularity or a translated word is estimated to relate to a word not entered in a dictionary, Which is recognized as a derived word with regard to a particularity morphological such as an affix.   24. Language analyzer, characterized in that it comprises: a first analysis means (9020) making it possible to carry out the morphological analysis for a sentence introduced from a predetermined language, a second analysis means (9022) analyzing the sentence of said language on the basis of the results of the morphological analysis carried out by said first analysis means, a dictionary (9018) in which dictionary data for said language used for analysis are stored by said first and second analysis means, and   a control means (9038) used to consult the dictionary   and to have said first and second means of analysis perform the analysis, o: the first means of analysis consults the dictionary, distinguishes the structural arrangement by distinguishing the particularity of the sentence introduced from said language having regard to its shape and   produces the estimate of a certain attitude as being the result   state of the analysis and of the structural role of said arrangement which is exerted in said sentence for said arrangement, and said second means of analysis analyzes the structure of the surface layer of the sentence of said language by applying a grammatical rule based on said estimated attitude and on the role, and analyzes the possible subsidiary relationship of the constituents contained in said sentence.   25. Analyzer according to claim 24, characterized in that the second analysis means performs the analysis of the arrangement preferentially to others if said arrangement is contained in the sentence of said language.   26. Analyzer according to claim 24, characterized in that the first means of analysis distinguishes the peculiarity of the sentence of said language having regard to its form and its meaning and produces an estimate of an expression of the apposition on the   basis of said distinguished features.   27. An analyzer according to claim 24, characterized in that the predetermined language is English, and the dictionary contains, as dictionary data, information of distinction which distinguishes predetermined words containing "let's" and "let us" (these two expressions indicating that the following verb is used in the first person plural of the imperative, and having other meanings), and   when the first means of analysis consults the diction-   nary to obtain said information of distinction, it excludes "let's" as well as "let us" from the object of the analysis by the second means of analysis if a punctuation is present at the previous part.   28. Analyzer according to claim 27, characterized in that   that the first means of analysis estimates attitude as an instruc-   tion and role as an invitation to the arrangement containing the excluded part.   29. Analyzer according to claim 24, characterized in that the first analysis means, if no dictionary data is obtained during consultation of the dictionary relatively   to several words linked by hyphens, consult the diction-   nary for each of the different words, estimates the whole of the part containing the different words as being an arrangement and estimates the attitude of said arrangement as being an adjectival group. 30. Analyzer according to claim 24, characterized in that the predetermined language is English, and the first means of analysis distinguishes the peculiarity of   the sentence of English having regard to its form, distinguishes the group cons-