JP2010272079A - Machine translation device, method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform translation, when a specific term in original text data is witten down with the abbreviation of the term, by selecting a translation candidate wherein the abbreviation generated from a plurality of translation candidates of the term, and the abbreviation written in a text agree. <P>SOLUTION: A term/abbreviation extraction part 26 refers to an analysis result of the original text data accepted by an input part 22 and analyzed by an original text analysis/translation part 25 by referring to a dictionary part 31, and extracts an abbreviation and a term of an expanded expression corresponding to the abbreviation from the original text, and a generated abbreviation selection part 28 compares generated abbreviations generated by an abbreviation generation part 27 with the extracted abbreviation extracted from original text data by the term/abbreviation extraction part 26 to select a generated abbreviation completely coinciding the extracted abbreviation, and the original text analysis/translation part 25 uses a translation candidate in the second language, from which the selected abbreviation has been generated by the abbreviation generation part 27, as a translation result. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a machine translation apparatus, method, and program for translating an original sentence in a first language to be translated into a translated sentence in a second language.

  With the progress of internationalization, the need for quick information exchange using foreign languages is increasing, and machine translation is highly expected as a tool for such information exchange. The result of translation by the current machine translation technology is not at a level that requires manual correction, and it takes a lot of time and effort to correct the translation result.

  In particular, if a specific term in the original text is accompanied by a supplementary explanation of the word, the translation result may be unnatural or the original text may be interpreted incorrectly. For example, if the original text “This is an optical character reader” is translated in the conventional way, the term “This is an optical character reader (optical character reader).” It will be repeated.

  On the other hand, when two types of translations of an abbreviation and an expanded translation of a term are registered in the dictionary, the user can select which translation to use, or two types of translations depending on conditions There is one that uses differently (see, for example, Patent Document 1). Also, when a word such as “Security Protocol Council (SPC) Information” is translated by the conventional method, “protocol” is interpreted as a verb, and the result of translation is “Safety is planning Council (SPC) information”. There was a case that became. On the other hand, when the character string in parentheses is the first character of some words existing before the parentheses, the word part is treated as a noun phrase and the syntax of the original text is analyzed. There are some (see, for example, Patent Document 2). According to this proposal, the correct translation result of “Safety Council Council (SPC) Information” is obtained for the original “Security Protocol Council (SPC) Information”.

Japanese Patent Laid-Open No. 06-309356 JP 11-66068 A

  However, if an abbreviation (alphabet) of the word is written for a specific term in the original text, the word is not registered in the dictionary, and there are multiple ways to translate it, the English translation result of the term May not match the abbreviations that are written together. For example, consider a case where the following original text is translated.

  Original 1: This document describes issues with Solaris installation update (ITU) or driver update (DU).

Previous translation result 1: This document has described the problem on updating (ITU) at the time of installation or the renewal of a driver (DU) of Solaris.
If ITU and DU are not registered in the dictionary, “Updating on Installation (ITU)” becomes “update (ITU) at the time of installation” and “Driver Update (DU)” becomes “the renewal of a” driver (DU) ”, and the translation of each word does not match the corresponding abbreviation.

  Although the above example sentences are sentences in the computer field, new terms frequently occur in fields such as the computer field where technological innovation is thriving. Therefore, there are many cases where translated words of new words are not registered in the dictionary. In such a case, it is not possible to use the technique of Patent Document 1 on the assumption that the translated word is already registered in the dictionary. In addition, by applying the above-mentioned Patent Document 2, if the part of speech of the word before the parenthesis is limited to nouns and the syntax of the original sentence is narrowed down, knowledge on the translation and the translation method employed in creating the translation Since the balance with the words in the parentheses is not considered in selecting the information, the final translation is determined independently of the information on the abbreviations in the parentheses.

  An object of the present invention is to provide a translation candidate in which an abbreviation generated from a plurality of translation candidates of a term and an abbreviation described in the sentence match when the abbreviation of the word is written together with a specific term in the original text Machine translation apparatus, method, and program for performing translation by selecting an item.

  The machine translation device of the present invention includes a storage device that stores a machine translation program and a dictionary unit, an input device that inputs original text data of a first language to be translated, and an output that outputs translated text data of a second language after translation In a machine translation device comprising an apparatus and an arithmetic control device for computing and executing the machine translation program, a dictionary unit storing vocabulary / rules for translating and analyzing the first language into the second language; An input unit that accepts original text data in the first language input from the input device and stores it in the storage device; and an original text that performs language analysis and translation of the original text data received by the input unit with reference to the dictionary unit An abbreviation and a corresponding expanded notation term are extracted from the original text data with reference to the analysis / translation section and the language analysis result of the original text data by the original text analysis / translation section, and the extracted abbreviation A term / abbreviation extraction unit that translates the expanded notation term corresponding to the original sentence analysis / translation part, and the expansion notation term extracted by the term / abbreviation extraction part and translated by the original sentence analysis / translation part The abbreviation generation unit that generates abbreviations from bilingual translation candidates, the abbreviation generated by the abbreviation generation unit, and the extracted abbreviations extracted from the original text data by the term / abbreviation extraction unit are compared, and the degree of coincidence is 100% When a generation abbreviation selection unit that selects a generation abbreviation and a generation abbreviation with a degree of coincidence of 100% is selected by the generation abbreviation selection unit, the source text analysis / translation unit The second language translation candidate is adopted as the translation result.

  According to the present invention, when an abbreviation of a term is written together with a specific term in the original text data, a generated abbreviation generated from a plurality of translation candidates of the term and an extracted abbreviation extracted from the original text data Is selected, a translation result that matches an abbreviation written together for a specific term can be obtained.

  In addition, if the generated abbreviations generated from a plurality of translation candidates for terms and the extracted abbreviations extracted from the original text data do not match, the word of the term that is the source of the generated abbreviation with a matching degree of less than 100% is brute force To generate abbreviations extracted from the abbreviations, so even if abbreviations in a sentence are difficult to match with abbreviation candidates generated using ordinary translation knowledge, use a different knowledge from ordinary translation knowledge. Can create translation candidates.

  In addition, since the term extracted from the original text data and the translation candidate that is the origin of the generated abbreviation are registered in the dictionary part, if the same term appears in subsequent documents, the translation can be easily unified. .

The functional block diagram of the machine translation apparatus 11 concerning embodiment of this invention. The block block diagram which shows the hardware constitutions of the machine translation apparatus concerning embodiment of this invention. The flowchart which shows the processing content of the machine translation apparatus concerning embodiment of this invention. The flowchart which shows the content of the process for abbreviations of step S308 of FIG. FIG. 4 is an explanatory diagram of a syntax tree structure obtained by analyzing the original sentence 1 in the original sentence analyzing process in step S305 of FIG. 3.

  Embodiments of the present invention will be described below. FIG. 1 is a functional block diagram of a machine translation apparatus 11 according to the embodiment of the present invention, and FIG. 2 is a block configuration diagram showing a hardware configuration of the machine translation apparatus according to the embodiment of the present invention.

  First, the hardware configuration of the machine translation apparatus according to the embodiment of the present invention will be described. In FIG. 2, the machine translation apparatus 11 according to the present invention is realized by installing the machine translation program according to the present invention on, for example, a general computer and executing the program in the processor 18 of the arithmetic and control unit 12.

  A display device 13, a mouse 14, a keyboard 15, a disk drive 16 and an auxiliary storage device 17 are connected to the arithmetic control device 12. The display device 13 displays calculation results of the calculation control device 12 and the like, and input devices such as a mouse 14 and a keyboard 15 are used for inputting various commands to the calculation control device 12 via the display device 13. It is also possible to input an original document of the first language to be translated from the keyboard 15. The disk drive 16 inputs / outputs a file such as an original document of a first language to be translated to a storage medium, and is also used as an input device for inputting the original document of the first language to be translated.

  The auxiliary storage device 17 stores calculation results of the calculation control device 12, a translation dictionary, and the like, and includes, for example, a plurality of hard disk drives (HDDs). The arithmetic and control unit 12 includes a processor 18 and a memory 19. The memory 19 stores a program 20 related to data retrieval and a work area 21 used when arithmetic processing is executed by the processor 18.

  Next, a functional configuration of the machine translation apparatus 11 according to the embodiment of the present invention will be described. Each functional block in the arithmetic control device 12 shown in FIG. 1 functions as each functional block when the processor 18 shown in FIG. 2 executes the program 20.

  In FIG. 1, an input unit 22 receives input of original text data in a first language to be translated. For example, when original text data is directly input by the user from the keyboard 15 or the like, or input from the disk drive 16 via a storage medium, the original text data is input to the memory 19 or the storage device 17 in the arithmetic control device 12. To remember.

  The sentence division unit 23 divides the original sentence of the original sentence data input from the input unit 22 into sentence units, and stores it in the work area 21 of the memory 19. The original text divided in units of one sentence is delivered to the control unit 24. The control unit 24 controls the source text analysis / translation unit 25, the term / abbreviation extraction unit 26, the abbreviation generation unit 27, the generation abbreviation selection unit 28, the translation preliminary candidate generation unit 29, the dictionary registration unit 30, and the output unit 32. is there.

  The original sentence analysis / translation unit 25 analyzes a sentence with reference to the dictionary unit 31 for the original sentence input from the input unit 22 and divided into one sentence unit by the sentence division unit 23. The abbreviation generation unit 27 also generates a translation candidate for the second language, which is a source for generating the abbreviation, and also generates a final translation in the second language.

  The term / abbreviation extracting unit 26 applies a specific term in the original sentence to the original sentence divided by the sentence dividing unit 23 on the basis of the analysis result obtained by the original sentence analyzing / translating unit 25. The abbreviation corresponding to it is extracted.

  The abbreviation generation unit 27 generates abbreviations from the second language translation candidates obtained by the source text analysis / translation unit 25 for the first language terms extracted by the term / abbreviation extraction unit 26. .

  The generation abbreviation selection unit 28 compares the abbreviation generated by the abbreviation generation unit 27 with the abbreviation extracted from the original text by the term / abbreviation extraction unit 26 and selects a generation abbreviation having a high degree of coincidence. It is. Although details will be described later, a translation candidate that is a source of the generated abbreviation selected by the generated abbreviation selection unit 28 is selected as a final translation result by the original text analysis / translation unit 25.

  The preliminary translation candidate generation unit 29 extracts the abbreviation generated by the abbreviation generation unit 27 and the abbreviation extracted from the original text by the term / abbreviation extraction unit 26 from the original text when the degree of coincidence is less than 100%. Preliminary translation candidates that match 100% with the abbreviations generated are generated. The degree of coincidence refers to the rate at which character strings constituting abbreviations coincide. For example, if the number of abbreviations is 3 and all 3 characters match, the matching degree is 100%, and if 2 characters match, it is 66%.

  The dictionary registration unit 30 is a dictionary of translation candidates in the second language, which are the basis of the generated abbreviation selected by the generated abbreviation selection unit 28, and terms in the first language extracted by the term / abbreviation extraction unit 26. It is registered in the vocabulary part 31a of the part 17.

  The dictionary unit 31 stores vocabulary and rules necessary for analyzing the input original text of the first language and translating it into the second language, and is formed in the storage device 17. Vocabulary / rules necessary for translation are stored in the vocabulary part 31a, the morphological analysis processing part 31b, the syntax analysis part 31c, the conversion rule part 31d, and the generation rule part 31e of the dictionary part 31.

  The output unit 32 outputs the result of translation generated by the original text analysis / translation unit 25 to the output device. For example, display is output to the display device 13 as an output device. The output device is not limited to the display device 13 and may be output to the voice output device by voice instead of the output from the display device 13 or in addition to the output from the display device 13 or printed out to the printing device. Alternatively, it may be output to a recording medium via the disk drive 16.

  Next, the flow of translation processing in the machine translation apparatus 11 according to the embodiment of the present invention will be described with reference to FIGS. 3 and 4 by taking Japanese-English translation as an example. Needless to say, the present invention can be applied not only to Japanese-English translation but also to English-Japanese translation and translation between other languages, such as Chinese-Japanese translation and Japanese-Chinese translation.

  FIG. 3 is a flowchart showing the processing contents of the machine translation apparatus according to the embodiment of the present invention. When the user inputs the original text data of the first language to be translated from the input device, first, the input unit 22 receives the original text data (S301). The input original text data is stored in the memory 19 or the storage device 17. Next, the sentence division unit 23 reads the original text data from the memory 19 or the storage device 17 and divides the original sentence into one sentence unit (S302). For each sentence of the divided original sentence, the control unit 24 sets “1” to the variable I (S303), and takes out the I-th original sentence (S304). The I-th original sentence is taken out to the work area 21 of the memory 19.

  Next, the original text analysis / translation unit 25 performs the original text analysis process with reference to the dictionary unit 31 with respect to the extracted I-th original text (S305). That is, the source text analysis / translation unit 25 performs a morpheme analysis process of one original sentence with reference to the vocabulary part 31a and the morpheme analysis rule part 31b, and performs a syntax analysis process with reference to the syntax analysis rule part 31c. Get information on part of speech and syntax of each word that makes up. The parsing process and other techniques used here are not specific to the present invention, but are general techniques related to language processing.

  Subsequently, using the original text analysis result by the original text analysis / translation unit 25, the term / abbreviation extraction unit 26 tries to extract a term / abbreviation candidate that may be related to the term and the abbreviation (S306). It is determined whether there are abbreviation candidates (S307). The term / abbreviation extraction unit 26 stores the term / abbreviation candidate in the term / abbreviation candidate storage area of the memory 19 when it is determined in this abbreviation / candidate presence / absence determination that there is a term / abbreviation candidate. Then, an abbreviation process is performed for the term / abbreviation candidate (S308), and a translation process for the I-th original sentence including the term / abbreviation candidate is performed. Details of the abbreviation processing will be described later.

  On the other hand, if it is determined in step S307 that there is no term / abbreviation candidate, the process for abbreviations in step S308 is omitted, and the process proceeds to step S309. Translation processing is performed (S309).

  Then, it is determined whether or not processing has been completed for all sentences of the original text, that is, whether or not all text processing has been completed (S310). If processing for all sentences of the original text has been completed, the processing is terminated. If the processing for all the original sentences has not been completed, “1” is added to the variable I (I = I + 1) (S311), and the process returns to step S304.

  FIG. 4 is a flowchart showing the contents of the abbreviation process in step S308 of FIG. The source text analysis / translation unit 25 extracts the term / abbreviation candidates extracted by the term / abbreviation extraction unit 26 (S401), and first creates translation candidates for the extracted extracted terms (S402). In other words, the source text analysis / translation unit 25 takes out the term / abbreviation candidates into the work area 21 of the memory 19, refers to the vocabulary unit 31 a and the conversion rule unit 31 d for the extracted terms, and converts the extracted terms into the target language. 2. Obtain translation words and conversion rule candidates, and create translation candidates for the extracted terms.

  Then, for one of the created translation candidates, the control unit 24 sets “1” to the variable J (S403), and takes out the Jth translation candidate (S404). Here, the translation processing technique used in the original text analysis / translation unit 25 is not unique to the present invention, but is a general technique related to machine translation. Specific contents of the translation processing in the original text analysis / translation unit 25 will be described in detail later using an example.

  The abbreviation generation unit 27 generates abbreviation candidates for the translation result of the extracted term created by the original text analysis / translation unit 25 (S405). Specific examples of abbreviation candidate creation will be described in detail later using examples.

  Subsequently, the generation abbreviation selection unit 28 compares the extracted abbreviations in the original text extracted by the term / abbreviation extraction unit 26 with the generation abbreviations generated by the abbreviation generation unit 27 (S406). Then, the generation abbreviation selection unit 28 calculates the degree of coincidence Aj obtained for the Jth translation result and is equal to or higher than the degree of coincidence Aj-1 obtained for the previous J-1th translation result. It is determined whether or not (S407). The generation abbreviation selection unit 28 stores the maximum matching score Amax in the maximum matching score candidate storage area in the memory 19 when the matching score Aj for the J-th translation result is equal to or higher (S408). On the other hand, if the degree of coincidence with the J-th translation result is less than Aj, the process proceeds to step S409. When there are a plurality of generated abbreviation candidates, the degree of coincidence Aj with the abbreviation extracted by the term / abbreviation extraction unit 26 is calculated for each abbreviation candidate, so that the maximum degree of coincidence Amax is stored.

  Subsequently, the control unit 24 determines whether or not the processing of all the translation candidates has been completed (S409), and if there is an unprocessed translation candidate, returns to step S404 (S410). If there is no unprocessed translation candidate, the generation abbreviation selection unit 28 determines whether or not the maximum matching degree Amax is 100% (S411), and when the maximum matching degree Amax is 100%, the matching degree Select a production abbreviation that is 100%. The source text analysis / translation unit 25 sets the translation candidate of the generated abbreviation having a matching degree of 100% as the final translation result (S413). That is, when a generation abbreviation with 100% coincidence is selected by the generation abbreviation selection unit 28, the source text analysis / translation unit 25 translates a translation candidate of the second language that is generated by the abbreviation generation unit 27 as a translation result. Used for.

  On the other hand, the generation abbreviation selection unit 28 activates the translation preliminary candidate generation unit 29 when the maximum matching degree Amax is less than 100%. The preliminary translation candidate generation unit 29 generates abbreviations extracted from the combined abbreviations by combining the words of the terms that are the source of the generated abbreviations generated by the abbreviation generation unit 27 having a matching degree of less than 100%. The term that is the source of the abbreviation with the highest degree of coincidence is generated as a preliminary translation candidate and stored in the preliminary translation candidate storage area in the memory 19 (S412).

  The generation abbreviation selection unit 28 determines whether or not there is an abbreviation having a matching degree of 100% among the abbreviations of the translation preliminary candidate generated by the translation preliminary candidate generating part 29 (S414), and there is an abbreviation having a matching degree of 100%. At that time, the source text analysis / translation unit 25 sets the translation candidate as the final translation result (S413). On the other hand, if it is determined in step S414 that there is no abbreviation having a matching degree of 100%, a message indicating that there is no abbreviation having a matching degree of 100% is displayed on the display device 13 (S415). For example, the abbreviations with the maximum matching degree among the abbreviations with a matching degree of less than 100% are displayed, all the abbreviations with a matching degree of less than 100% are displayed as a list, or abbreviations with a predetermined ratio or more of abbreviations with a matching degree of less than 100% Or display a list.

  Next, the original text analysis in step S305 to the abbreviation processing in step S308 will be described in detail with reference to the following original text 1.

  Original 1: This document describes issues with Solaris installation update (ITU) or driver update (DU).

  FIG. 5 is an explanatory diagram of a syntax tree structure obtained by analyzing the original sentence 1 in the original sentence analyzing process in step S305 of FIG. The term / abbreviation extraction unit 26 uses the original text analysis result shown in FIG. 5 by the original text analysis / translation unit 25 to extract words that may be related to the abbreviation (S306).

  Whether or not there is a relationship between a term and its abbreviation is determined based on the following conditions. For example, “a character string in which a noun phrase and a modification phrase immediately following it include a character string indicating insertion, and characters constituting the modification phrase excluding the character string indicating insertion represent an abbreviation of the language to be translated It is a condition of “consisting of”. The character string indicating insertion is a character or a phrase such as various parentheses, dashes, commas, colons, or “that is”.

  In the example of the original text 1, “(ITU)” immediately after the noun phrase “update at installation” includes parentheses, and characters other than the parentheses are composed only of alphabetic characters constituting English as a translation target language. And consist only of uppercase letters representing English abbreviations. The noun phrases “driver update” and “(DU)” satisfy the same condition. From these points, “update at installation” and “(ITU)”, “driver update” and “(DU)” are extracted as terms that may be related to terms and their abbreviations.

  In addition, about the range judged as a noun phrase, there can be a plurality of candidates such as a continuous part of a noun, an adjective + a noun. In the example of the original sentence 1, there may be a candidate for determining “update at the time of installing Solaris” as a noun phrase. Although description of the translation candidate creation method regarding this candidate is omitted, the degree of coincidence with the extracted abbreviation is smaller than that of other noun phrases. This is because there is a word “Solaris” at the beginning of the noun phrase, and therefore, an alphabet “S” that is not in the extracted abbreviation is included in the head part of the generated abbreviation.

  The term / abbreviation extraction unit 26 attempts to extract a term / abbreviation candidate that may be related to a term and the abbreviation (S306), and then determines whether there is a term / abbreviation candidate (S307). When there are term / abbreviation candidates, the process proceeds to the abbreviation process in step S308 in FIG.

  In the abbreviation processing, the source text analysis / translation unit 25 extracts the term / abbreviation candidate extracted by the term / abbreviation extraction unit 26 (S401), and creates a translation candidate for the extracted extracted term (S402). That is, the source text analysis / translation unit 25 refers to the vocabulary unit 31a and the conversion rule unit 31d to create translation candidates for “update at installation” and “driver update” (S402).

  In the following, the translation candidate creation process in the original sentence analysis / translation unit 25 will be described in detail by taking the translation of “update at installation” as an example. It is assumed that the translation candidates and conversion rules relating to “update at installation” stored in the vocabulary part 31a and the conversion rule part 31d are as follows.

Installation = installation, install, installing
At the time of, time,
Conversion rule 1: Translation order of noun “time”: at the time of noun Conversion rule 2: Translation order of noun “time”: Noun time
Update = renewal, update, updating,
When these translation word candidates and conversion rules are combined in a brute force manner, 3 * 2 * 3 = 18 translation candidates are obtained as follows (S402).
1-1-1. Renewal at the time of installation
1-1-2.renewal at the time of install
1-1-3.renewal at the time of installing
2-1-1.update at the time of installation
2-1-2.update at the time of install
2-1-3.update at the time of installing
3-1-1. Updating at the time of installation
3-1-2. Updating at the time of install
3-1-3. Updating at the time of installing
1-2-1. Installation time renewal
1-2-2.install time renewal
1-2-3.installing time renewal
2-2-1. Installation time update
2-2-2.install time update
2-2-3.installing time update
3-2-1. Installation time updating
3-2-2.install time updating
3-2-3. Installing time updating
Abbreviations generated from these translation candidates are as follows (S405).
1-1-1. RTI
1-1-2. RTI
1-1-3. RTI
2-1-1. UTI
2-1-2. UTI
2-1-3. UTI
3-1-1. UTI
3-1-2. UTI
3-1-3. UTI
1-2-1. ITR
1-2-2. ITR
1-2-3. ITR
2-2-1. ITU
2-2-2. ITU
2-2-3. ITU
3-2-1. ITU
3-2-2. ITU
3-2-3. ITU
That is, four types of abbreviations are obtained: RTI, UTI, ITR, and ITU.

  Here, a general method for generating abbreviations will be described. In the case of a compound word as described above, an abbreviation is basically configured by capitalizing the initial letters of independent words excluding articles, prepositions, conjunctions, etc. among the words constituting the term.

  When there are a plurality of terms having the same initial letter, the second character may be part of the abbreviation for one word constituting the compound word. For example, in the case of doctor of dental surgery and doctor of dental science, the former abbreviation is DDS and the latter is DDSc.

  In addition, when the word starts with “ex”, “X” may be used instead of the initial “E”. For example, the abbreviation for Extensible Markup Language is XML. In addition, there are exceptional rules such as using X for words starting with “trans” or “cross”. In the case of a single word composed of one word, it is generally composed of the first few letters of each syllable constituting the word. When the abbreviation generation unit 27 generates an abbreviation, various rules other than the above are considered.

When the degree of coincidence between these generated abbreviation candidates and the abbreviation ITU extracted from the original text is calculated, the ITUs of 2-2-1 to 3-2-3 become 100% coincidence, and the degree of coincidence is the highest. Become. In calculating the degree of coincidence, the number of characters, the character type, the order of appearance, and the like may be taken into consideration in addition to the proportion of the character strings constituting the abbreviations. Based on the calculation result of the degree of coincidence, translation candidates for “update at installation” are determined as follows.
2-2-1. Installation time update
2-2-2.install time update
2-2-3.installing time update
3-2-1. Installation time updating
3-2-2.install time updating
3-2-3. Installing time updating
When there are a plurality of candidates as in this case, the translation candidates are narrowed down to one based on the translation rule of each word constituting the translated word, the order registered in the dictionary, and the like. In this example, since there is no particular translation rule that can be used, the translation result is determined as “2-2-1. Installation time update” based on the order registered in the dictionary.

  As described above, when the machine translation apparatus according to the embodiment of the present invention is used, the translation of “update at installation” is determined as “install time update”, and the translation result of “update at installation (ITU)” is It becomes “install time update (ITU)”, and a translation result in which the translation result of the term matches the abbreviation can be obtained.

  Next, when the degree of coincidence between each generated abbreviation candidate and the abbreviation extracted from the original text is less than 100%, the translation preliminary candidate generation unit 29 generates a source of a generated abbreviation having a degree of coincidence of less than 100%. The abbreviations extracted from the abbreviations having the highest matching score are generated as preliminary translation candidates. Then, the generation abbreviation selection unit 28 selects an abbreviation having a matching degree of 100% from the abbreviations of the translation preliminary candidate generated by the translation preliminary candidate generation unit 29.

For example, consider the case where the original text 2 is as follows. In this case, the terms and abbreviations extracted from the original sentence 2 are two terms 1 and 2 and two abbreviations 1 and 2.
Original 2: Dentist (DDS), Dentist (DDSc)
Term 1 extracted from source text 1: Dentist Abbreviation extracted from source text 1: DDS
Term 2: extracted from source 2: Doctor of Dentistry Abbreviation: 2, extracted from source 2: DDSc
The abbreviation 1 with the highest degree of coincidence generated from the term 1 by the abbreviation generator 27 is DS (Dental Surgeon), DS (Dentistry Surgeon), DD (Dental Doctor), DD (Dentistry Doctor), DD (Doctor of Dentistry The degree of coincidence between the generated abbreviations DS and DD and the abbreviation 1 “DDS” extracted from the original sentence 2 is 66%. The most abbreviated abbreviation 2 generated from term 2 is DDSc (Doctor of Dental Science), and the abbreviation 2 “DDSc” extracted from original sentence 2 is 100% coincident. is there.

  Since the generation abbreviations DS and DD in Example 1 of the original sentence 2 have a degree of coincidence of 66%, the translation preliminary candidate generation unit 29 determines the source of the generation abbreviations DS and DD generated by the abbreviation generation unit 27. The abbreviation “DDS” extracted from a combination of the words {DS (Dental Surgeon), DS (Dentistry Surgeon), DD (Dental Doctor), DD (Dentistry Doctor), DD (Doctor of Dentistry)} Select combinations that can generate.

  Accordingly, “Doctor of Dental Surgeon”, “Doctor of Dentistry Surgeon”, and “Dentistry Doctor Surgeon” are generated as translation preliminary candidates. When there are a plurality of combinations, it is generated based on whether or not a combination registered in the dictionary unit 31 as a composite word is included, the order registered in the dictionary unit 31, the number of search results by Internet search, etc. Candidate priorities can also be determined.

  In this way, even when the degree of coincidence is less than 100%, if there is a generated translation preliminary candidate, it can be provided to the user as a translation preliminary candidate. In this case, it is possible to provide a user interface that notifies the user that there is a translation preliminary candidate, for example, by displaying a specific symbol at the end of the translation.

  According to an embodiment of the present invention, when an abbreviation of a term is written together with a specific term in the original text, for example, with respect to the notation of “update at installation (ITU)”, a plurality of translations of the term Select the translation candidate “install time update” where the generated abbreviation (RTI, UTI, ITR, ITU) generated from the candidate matches the extracted abbreviation (ITU) extracted from the original text. A translation result that matches the abbreviation (ITU) written together with “update” can be obtained.

  Also, if the generated abbreviations generated from a plurality of translation candidates for terms and the extracted abbreviations extracted from the original text do not match, the word of the term that is the source of the generated abbreviation with a matching degree of less than 100% Since the abbreviations extracted from the combination are generated in combination, even if the abbreviations in the sentence are difficult to match with the abbreviation candidates generated using normal translation knowledge, translation is performed using knowledge different from normal translation knowledge Candidates can be created.

  Furthermore, since the dictionary registration unit 30 registers the term extracted from the original text and the translation candidate that is the source of the generated abbreviation in the dictionary unit 31, if the same term appears in subsequent documents, the translation word is unified. It can be easily achieved.

DESCRIPTION OF SYMBOLS 11 ... Machine translation apparatus, 12 ... Calculation control apparatus, 13 ... Display apparatus, 14 ... Mouse, 15 ... Keyboard, 16 ... Disk drive, 17 ... Auxiliary storage device, 18 ... Processor, 19 ... Memory, 20 ... Program, 21 ... Work area 22 ... Input unit 23 ... Sentence division unit 24 ... Control unit 25 ... Original text analysis / translation unit 26 ... Term / abbreviation extraction unit 27 ... Abbreviation generation unit 28 ... Generation abbreviation selection unit 29 ... Preliminary translation candidate generation unit 30 ... Dictionary registration unit 31 ... Dictionary unit 32 ... Dictionary registration unit

Claims (5)

A storage device storing a machine translation program and a dictionary unit, an input device for inputting original text data of a first language to be translated, an output device for outputting translated text data of a second language after translation, and the machine translation program In a machine translation device provided with an arithmetic control device for performing arithmetic operations,
A dictionary unit storing vocabulary and rules for translating and analyzing the first language into the second language;
An input unit that accepts original text data in a first language input from the input device and stores the original data in the storage device;
A source text analysis / translation unit that performs language analysis and translation of the source text data received by the input unit with reference to the dictionary unit;
Terms that extract abbreviations and corresponding expanded notation terms from the original text with reference to the language analysis results of the original text data by the original text analysis / translation unit, and translate the extracted expanded notation terms in the original text analysis / translation unit An abbreviation extraction unit;
An abbreviation generator for generating abbreviations from second language translation candidates of the expanded notation term extracted by the term / abbreviation extraction unit and translated by the source text analysis / translation unit;
A generation abbreviation selection unit that compares a generation abbreviation generated by the abbreviation generation unit with an extraction abbreviation extracted from original text data by the term / abbreviation extraction unit and selects a generation abbreviation having a degree of coincidence of 100%;
When a generation abbreviation with 100% coincidence is selected by the generation abbreviation selection unit, the source text analysis / translation unit adopts a translation candidate of a second language that is generated by the abbreviation generation unit as a translation result. A machine translation apparatus characterized by:   When the coincidence between the generated abbreviation generated by the abbreviation generation unit and the extracted abbreviation extracted from the original text data by the term / abbreviation extraction unit is not 100%, it is generated by the abbreviation generation unit having a coincidence degree of less than 100% The second language translation candidate words that are the source of the generated abbreviations are combined in a brute force manner to generate abbreviations extracted from them, and the term that is the source of the abbreviation with the highest matching score is translated. A preliminary translation candidate generation unit for generating as a preliminary candidate is provided, and the generation abbreviation selection unit selects an abbreviation having a matching degree of 100% from the abbreviations of the preliminary translation candidate generated by the preliminary translation candidate generation unit. The machine translation apparatus according to claim 1, wherein the machine translation apparatus is a machine translation device.   A dictionary registration unit is provided for registering the expanded notation term extracted from the original text data by the term / abbreviation extraction unit and the translation candidate that is the source of the generated abbreviation selected by the generated translation selection unit in the dictionary unit. The machine translation apparatus according to claim 1, wherein the machine translation apparatus is a machine translation apparatus. The original text data of the first language to be translated is input from the input device, the machine translation program stored in the storage device is calculated and executed by the calculation control device, and the translation processing is performed with reference to the dictionary unit stored in the storage device. In a machine translation method for outputting translated data of a second language after translation from an output device,
Preliminarily storing vocabulary and rules for translating and analyzing the first language into the second language in the dictionary unit,
Receiving the original text data of the first language input from the input device and storing it in the storage device;
Language analysis and translation of the original text data received by the input unit with reference to the dictionary unit,
Extracting abbreviations and corresponding expanded notation terms from the original text with reference to the language analysis results of the original text data,
Translate the extracted expanded terminology,
An abbreviation is generated from a translation candidate in the second language of the translated term of the translated notation,
Compare the generated generation abbreviation with the extracted abbreviation extracted from the original text data, select a generation abbreviation with 100% match,
A machine translation method characterized in that, when a generation abbreviation having a degree of coincidence of 100% is selected, a translation candidate of a second language as a generation source is adopted as a translation result. A storage device storing a machine translation program and a dictionary unit, an input device for inputting original text data of a first language to be translated, an output device for outputting translated text data of a second language after translation, and the machine translation program In a computer for functioning as a machine translation device provided with an arithmetic control device for performing arithmetic operations,
Preliminarily storing vocabulary and rules for translating and analyzing the first language into the second language in the dictionary unit,
A procedure for accepting original text data in a first language input from the input device to a computer and storing it in the storage device;
A procedure for performing language analysis and translation of the original text data received by the input unit with reference to the dictionary unit,
A procedure for extracting abbreviations and corresponding expansion notation terms from the original text with reference to the language analysis results of the original text data;
A procedure for translating the extracted expanded terminology,
Generating abbreviations from translation candidates in the second language of the translated terms of the expression,
Comparing the generated generation abbreviation with the extracted abbreviation extracted from the original text data and selecting a generation abbreviation with a degree of match of 100%;
A program for executing, when a generation abbreviation having a coincidence degree of 100% is selected, a procedure for adopting a translation candidate of a second language as a generation source as a translation result.

Images