GB2241359A

GB2241359A - Translation machine

Info

Publication number: GB2241359A
Application number: GB9101804A
Authority: GB
Inventors: Hitoshi Suzuki; Yoji Fukumochi; Shuzo Kugimiya; Ichiko Sata; Tokuyuki Hirai
Original assignee: Sharp Corp
Current assignee: Sharp Corp
Priority date: 1990-01-26
Filing date: 1991-01-28
Publication date: 1991-08-28
Also published as: GB9101804D0

Abstract

A translation machine which is capable of setting a field of a translated equivalent in precedence and capable of changing a precedence rank of the translated equivalent by referring a translated equivalent finally selected with respect to a word in an input sentence having a plurality of translated equivalents after completing a translation, the translation machine includes a unit for adding an information of a post-learning to the translated equivalent selected finally so that precedence rank of the translated equivalent is set as a next rank to a rank of a translated equivalent output at first in a case of performing a learning of a dictionary, and a unit for detecting a field to be set so that the translated equivalent added the information of the post-learning is consulted with the dictionary in precedence. The machine allows selection between candidate translations of an input word according to a field specified by the user. <IMAGE>

Description

TITLE OF THE INVENTION TRANSLATION MACHINE BACKGROUND OF THE INVENTION 1. Field of the Invention The technical field to which the present invention relates concerns a translation machine which is capable of performing an analysis, a conversion of an input sentence and a generation of a translation equivalent of the input sentence for outputting the translated equivalent. More particularly, the present invention relates to the translation machine which is capable of providing a learning function for changing the ranks of translated equivalents of a word included in a sentence described in a target language and capable of setting a field of the translated word.

2. Description of the Related Art The inventors of the present invention know that there has been proposed a translation machine which allows a user to specify a field of the translated word desired in advance, and in a case that one original word has two or more translated equivalents, the translation machine is adapted to output a desired word on a basis of an information of the field specified.

The above-mentioned translation machine also provides a learning function for ranking up a translated equivalent selected in the latest translation.

However, the above-mentioned translation machine gives a priority to the latest-selected translated equivalent without regarding whether a user specifies the field or not. When the input sentence has the same structure as the latest processed sentence, the abovementioned translation machine outputs a translated word in the latest translation whatever the field a user may specify. If the new input sentence needs a different translated word for an original word from the latest processed sentence, that is, the new input sentence belongs to a different field from the latest processed sentence, the above-mentioned translation machine cannot translate the new input sentence correctly. As a consequence, the technical (ie. practical) problem which arises is that it results in a lowering of the rate of the correct answer.

Further, an interactive translation machine selects the proper translated word on the basis of the grammar rule and the context of the input word in a case that the input word has two or more translated equivalents. The interactive translation machine allows a user to select the proper translated word from the translated words stored in the dictionary.

However, the above-mentioned interactive translation machine often needs to specify the field and the style of an input sentence as the selecting standards. In this case, the proper grammar rule and the context of the input word are not sufficient to select the proper translated word.

In translating from English to Japanese, for example, a word "book" is generally translated into

(hon)". In an economic field, it should be translated into

(choubo: meaning an account book)". As another example, a word "perform" may be translated into

(suikou~suru: matching to an English word of Latin derivation like "accomplish") or

(okonau: matching to a word of Anglo-Saxon origin like "carry out" or "do") on the basis of the user's taste and sentence style. The above-mentioned interactive translation machine cannot cope with such delicate selection of the translated word.

The above-mentioned interactive translation machine is designed so that the user may specify the translated word in accordance with the user's taste for obtaining the desired translated word. However, it results in the technical tie. practical) problem of lowering of the operating efficiency.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide a translation machine which is capable of performing a precedence process in accordance with a learning in a specified field only if a field is set at a time of processing a translation in a case of performing a learning process of a dictionary, as well as capable of outputting a translated word which is learned in the specified field with a priority at a time when a field is set in a case of processing a translation.

The object of the present invention can be achieved by a translation machine which is capable of setting a field of a translated equivalent in precedence and capable of changing a precedence rank of the translated equivalent by referring a translated equivalent finally selected with respect to a word in an input sentence having a plurality of translated equivalents after completing a translation, the translation machine includes a unit for adding an information of a postlearning to the translated equivalent selected finally so that a precedence rank of the translated equivalent is set as a next rank to a rank of a translated equivalent output at first in a case of performing a learning of a dictionary, and a unit for detecting a field to be set so that the translated equivalent added the information of the post-learning is consulted with the dictionary in precedence.

In operation of the translation machine, the adding unit serves to add the learning information to a translated word selected in the latest translation and to rank up the information-added translated word next to the translated word output at first if the field is prespecified in executing the translation. In consulting dictionaries, the detecting unit sets the field and consults the translated word added the specified-field and the post-learning information with the dictionaries at first.

As a consequence, the translation machine enables to perform the learning in each field in a case that a user is allowed to specify the field and to learn the translated word resulting in outputting a more suitable translated sentence to each field.

Preferably, the translation machine further includes a unit for storing the adding unit and the detecting unit.

More preferably, the storing unit is a central processing unit.

Preferably, the translation machine further includes a memory table for storing the dictionary, grammar rules and tree structure converting rules.

More preferably, the dictionary includes a basic dictionary for storing all the translated equivalent.

The translation machine further includes a translation module for translating the input sentence described in the source language into a sentence described in the target language, preferably.

The dictionary further includes a learning dictionary for storing the learning processed equivalents, preferably.

The translation module preferably includes a dictionary consulting and morphologic element analyzing unit which is adapted to consult the dictionaries and for obtaining grammatical and translating information for each word so that a tense, a person, a numeral and so forth of the each word is analyzed.

Furthermore preferably, the translation module further includes a syntax analyzing unit which is adapted to convert a sentence structure of the input sentence described in the source language into a sentence structure of a sentence described in the target language on a basis of the tree structure converting rules.

Preferably, the translation module further includes a sentence generating unit which is adapted to generate a translated word on a basis of the sentence structure of the sentence described in the target language and the information obtained from the dictionaries and to output the translated word.

The source language is an English and the target language is a Japanese, preferably.

It is another object of the present invention to provide a translation machine which is capable of specifying a field and a text style with classification codes in accordance with the user's instruction so that the translated word selected on a basis of the user's taste is output in precedence of any other words.

Another object of the present invention can be achieved by a translation machine which is capable of performing a sentence structure analysis, a sentence structure transformation and a translated word generation, the translation machine being capable of outputting translated equivalents, the translation machine includes a unit for adding classification codes to each of translated words, a unit for specifying a classification code during a translation process, and a unit for outputting a translated word having added a classification code which is the same as the specified classification code in precedence in a case that the word of the input sentence has a plurality of translated word candidates.

In operation of the translation machine, the adding unit enables to pre-specify the operator's desired classification code added to each translated word stored in the dictionary for the purpose of selecting the desired field and text style.

The specifying unit enables to pre-specify the user's desired two or more classification codes and to add ranks to the classification codes. In a case that a word included in the input sentence has two or#more translated equivalent candidates, the outputting unit enables to output a translated word with the same classification code as the specified classification word in precedence of any other translated word.

As a consequence, the translation machine enables to output a translated word with the specified classification code having a higher rank in precedence of any other translated word in a case that the word included in the input sentence has two or more translated equivalent candidates.

Preferably, the specifying unit includes a main memory having a classification code buffer,a word buffer, and a translated word buffer.

More preferably, the classification code buffer includes a classification code table.

The translation machine further includes a unit for specifying a plurality of classification codes with a priority during a translation process, and a unit for outputting a translated word having a classification code which is the same as the classification code having a higher priority among the plurality of the specified classification codes in a case that there are a plurality of translated word candidates for a word in the input sentence, preferably.

Furthermore preferably, the classification code represents a field or a style of each of the translated equivalent.

Further objects and advantages of the present invention will be apparent from the following description of the preferred embodiments of the invention as illustrated in the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram showing an arrangement of a translation machine according to a first embodiment of the present invention; Fig. 2 is a block diagram showing an arrangement of a translation module included in the translation machine of Fig. 1; Figs. 3 and 4 are a flowchart showing the learning process performed in the translation machine of Fig. 1; Figs. 5 and 6 are a flowchart showing the dictionary-consulting process performed in the translation machine of Fig. 1; Fig. 7 is a block diagram showing arrangement of a translation machine according to a second embodiment of the present invention; Fig. 8 is a flowchart showing the translating process performed in the translation machine of Fig. 5; Fig. 9 is a view showing the analyzed syntax in a tree-structure manner.

DESCRIPTION OF THE PREFERRED EMBODIMENTS The translation machine according to a first embodiment of the present invention will be described with reference to Figs. 1 to 6.

Fig. 1 is a block diagram showing an arrangement of the translation machine according to the first embodiment.

As shown in Fig. 1, 1 denotes a central processing unit (CPU) which includes an adding unit and a detecting unit, 2 denotes a main memory, 3 denotes a cathode-ray tube (CRT) for displaying a translated sentence, 4 denotes a keyboard from which a user enables to give various indications to the translation machine, 5 denotes a translation module which translates a sentence described in a source language into a sentence described in a target language. The translation module 5 is connected to a table 6 which stores dictionaries, grammar rules, and tree-structure converting rules.

Fig. 2 is a block diagram showing an arrangement of the translation module 5.

As shown in Fig. 2, the translation module 5 includes a dictionary consulting and morphologic element analyzing unit 51 (for short, word analyzing unit), a syntax analyzing unit 52, a syntax converting unit 53, and a sentence generating unit 54.

The word analyzing unit 51 serves to consult the dictionaries 6 and to obtain grammatical and translating information for each word in the input sentence for analyzing a tense, a person, a numeral and so forth of each word.

The syntax analyzing unit 52 serves to convert the sentence structure of the input sentence described in the source language into the sentence structure of a target language on the basis of the tree-structure converting rules.

The sentence generating unit 54 serves to generate a translated word on the basis of the sentence structure and the information obtained from the dictionaries and to output the translated word.

In the following, the learning process of the abovementioned translation machine will be described with reference to Figs. 3 and 4. The following processes are controlled by the CPU1 with the adding unit and the detecting unit. The dictionaries 6 include a basic dictionary and a learning dictionary. The basic dictionary stores the translated equivalents of all the words converted thereby. The learning dictionary stores the translated equivalents of the learning-processed words. By replacing a pointer of a word set to the basic dictionary with a pointer of a word set to the learning dictionary, the learning dictionary is made effective for finding out the word.

The program remains in a translation-finishing state at a time when the learning process starts since the learning process is performed immediately after the translating process is terminated.

In this state, the syntax analyzing unit 52 stores in a buffer thereof an information of a dictionary page on which the translated word selected on the analysis is located and an information of a dictionary page on which the translated word given a priority by a user is located. At steps S1 and S2, the information of both dictionary pages are read out of the buffer and are saved in a buffer of the main memory 2. At steps S3 and S4, a flag is reset and a copy page is initialized to 1. Then, at a step S5, it is checked whether or not the dictionary page is a specified page in the analysis. If not, the process goes to a step Sll.

At the step S5, if it is yes, the process goes to a step S6 at which it is checked whether or not a user has specified a field at a time when the translating process is performed. If the field is not specified, the process goes to a step S7 for performing the normal learning process. At the step S7, an information of the translated word on the page selected by the user is copied to the buffer of the main memory 2. At a step S8, the information other than the information of the translated word is copied from the page specified in the analysis to the buffer. At a step S9, a post-learning information is also copied to the buffer.

On the other hand, if the field is selected by a user, then at a step S10 a flag for copying a page matching to the selected field is set and the process goes to the step Sll.

At the step Sll, it is checked whether or not a field information is added to the page specified in the analysis. If it is not, at a step S12, the page specified in the analysis is copied to the buffer. If it is yes, then at a step 513 a minor flag is set to the field information and is copied to the buffer.

The minor flag is used to give a priority to the latest learning in each field at a time of using the dictionaries.

At a step S14, it is checked whether or not a flag for copying the page matching to the field selected by a user is set. If it is not, the process goes to a step S19. On the other hand, if it is yes, the information of the translated word is copied from the page selected by the user to the buffer at the step S15. Then, at the step S16 the information other than the information of the translated word is copied from the page specified in the analysis to the buffer. Then, at the step S17, the field and the post-learning informations set are added to the buffer. At a step S18, a flag for copying the page matching to the field is reset.

At the step S19, it is checked whether or not the current page is a last page of a translated equivalent to the subject word. If it is not, the process goes to a step S22 at which it is set to the next page. Then, the process returns to the step S5 from which the similar operation is repeated. If it is yes, at a step S20, the copied content in the buffer is written in the learning dictionary. At a step S21, a dictionary pointer of the subject word in the basic dictionary is replaced with a pointer of the learning-dictionary.

As described above, the learning dictionary is created by copying the translated word selected by the user immediately after the translated word specified in the analysis if a field is specified in performing the translation, and by copying the translated word selected by the user immediately before the translated word specified in the analysis if a field is not specified.

Therefore, if no field is specified in the next translating operation, in outputting the translated word, the translated word specified in the analysis takes precedence of the translate word selected by the user at a time when the field is set, thereby it makes possible to perform a learning process in each field.

Next, the process of consulting the dictionaries will be described with reference to Figs. 5 and 6.

At a step S51, it is checked whether or not a field is specified. If it is not, the process goes to a step S58 at which the normal dictionary-consulting operation is performed. If it is specified, the process goes to a step S52 at which the dictionary page is initialized to 1. Then, at a step S53, it is checked whether or not the information of the specified field is added to a translated word located on the page. If it is yes, a consultation of the dictionary is performed for the page.

At a step S55, the information (mark) indicating a completion of the dictionary-consulting is added to the translated word. At a step S56, it is checked whether or not the current page is the last page of the translated word. If it is not, the process goes to a step S57 at which the page is incremented and returns to a step S53 from which the similar process is performed.

The informations of the field and the post-learning are checked so that the translated word having both of the informations takes a precedence of any other word for consulting the dictionary.

On the other hand, at the step S56, if the current page is the last page of the translated word, the process goes to a step S58 at which the page is initialized to 1 for consulting the dictionary from the first page to the last page in sequence. If the field information is added to the translated word, at a step S9, it is checked whether or not the dictionary-consulting mark (meaning that the dictionary has been already consulted for the mark-added word) is added one page by one page. This checking is performed in order to prevent repeated operations of the dictionary-consulting. If the mark is added to a page, the marked-page is skipped, that is, the dictionary is not consulted for the page, at a step S62, the dictionary-consulting operation goes to a next page and the process returns to the step S59.If the mark is not added to the page, at the step S60, the dictionaryconsulting operation is performed for the page. At the step S61, it is checked whether or not the current page is the last page of the translated word. If it is not, the process goes to the step S62 and if it is yes, the process is terminated.

The above-mentioned dictionary-consulting process enables to output the translated word which is processed in the learning in the specified field in precedence rather than the translated equivalent since the translated word is given a priority of the dictionaryconsulting comparing with the translated word which is not processed in the learning in the same field in a case that the field is set.

The translation machine according to a second embodiment of the present invention will be described with reference to Figs. 7 to 9. The second embodiment of the present invention concerns with the translation machine which is capable of allowing a user to prespecify the user's desired field and text style with a classification code so that the translation machine is capable of outputting the desired translated word in precedence of any other translated word.

The arrangement of the second embodiment is the same as that of the first embodiment except the main memory 2.

As shown in Fig. 7, the main memory 2 contains the classification-code buffer 21, a word buffer 22, and a translated-equivalent buffer 23. For understanding the arrangement of this embodiment, it is necessary to refer to the description of the first embodiment and Figs. 1 and 7.

Fig. 8 is a flowchart showing the operation of the translation module in this embodiment. Table 1 shows a table which contains the classification codes and the classifications matching to those codes.

The classification code represents the field and the text style of each translated word. The field indicates an economy, a machine, an information and so forth, the text style indicates the Katakana form (mainly including the Japanese word of European derivation, the Hiragana form (mainly including the Japanese words of Japanese origin), the Kanji form (mainly including the Japanese words of Chinese derivation) and so forth. The predetermined classification code to each classification is added to each translated word stored in a dictionary for a translation.

The translation process of the above-mentioned translation machine is shown in Fig. 8. At a step S81 in Fig. 8, a user selects and specifies the classification codes matching to the field of the original text and the user's desired style of the translated text on the basis of the classification code table shown in Table 1.

Table 1 Classification Code Table

Category Code Category 000 001 Economy 002 Machinery 003 Information 100 Katakana Form 101 Hiragana Form 102 Chinese Form 200 User 1 201 User 2 In Table 1, a classification code 000 indicates that no code is selected, a classification code 001 indicates an economy field, a classification code 002 indicates a machinery field, a classification code 003 indicates an information field, and so forth.

A classification code 100 represents a translated word of the Katakana form, a classification code 101 represents a translated word of the Hiragana form, a classification code 102 represents a translated word of the Kanji form, and so forth.

The classification codes with 201 or the later numbers correspond to the classification codes of which the user enables to define its meaning so that the defined codes are added to preferred translated words.

The classification code specified by the user is stored in the classification-code buffer 21 contained in the main memory 2. Assuming that the user selects the classification code 003 representing the information field at first and the classification code 100 representing the translated word of the Katakana form secondly as shown in Table 2.

It results in that classification-code buffer 21 stores the classification code 003 in a first buffer, the classification code 100 in a second buffer, and the classification code 000 for indicating no other code is saved in the buffer 21 in a third buffer.

Table 2 Classification-code Buffer

(1) 003 (2) 100 (3) 000 Herein, it is assumed that a sentence of "The display shows a document" is input.

At a step S82 (see Fig.8), the translation machine reads the input sentence. At a step S83, the word analyzing unit 51 (see Fig. 2) starts to -consult the dictionary for each word included in the sentence. The information of the dictionary-consulting is drawn from the memory 6 so that each part of speech matches to each English word is saved in the word buffer 22 of the main memory 2 (see Fig. 5) as shown in Table 3.

Table 3 Word Buffer

English Word Part of Speech The Article display Noun Verb shows Noun Verb a Article document Noun Verb The translated equivalent to each word and the classification code added to each translated word are saved in the translated-word buffer 23 in such a manner as shown in Tables 4 to 7.

At a step S84 in Fig. 8, the syntax analyzing unit 52 (see Fig. 2) serves to analyze the syntax of the input sentence.

Table 4 display (noun)

Translated Word Category Code (1) R Br1 000 (2) Xggg 003 (3) r d 100 Table 5 show (verb)

Translated Word Category Code (1) gt 000 (2) t s 003 Table 6 document (noun)

Translated Word Category Code (1) t X 000 (2) F 100 Table 7 computer (noun)

Translated Word Category Code (1) Eb 000 (2) a > t -a 100 (3) tgt 200 (4) =gt 201 Table 8 shows a part of the grammar rules which are used for the analysis and are stored in the memory 6.As shown, the grammar rule (a) indicates that a sentence consists of a noun phrase and a verb phrase,the grammar rule (b) indicates that a noun phrase consists of an article and a noun, and the grammar rule (c) indicates that a verb phrase consists of a verb and a noun phrase.

Table 8 Grammar Rules

(a) Sentence -+ Noun phrase + Verb phrase (b) Noun phrase -+ Article + Noun (c) Verb phrase -+ Verb + Noun phrase For the parts of the input sentence "The display" and "a document", the grammar rule (b) (a noun phrase consisting of an article and a noun) is applied. That is, those parts are recognized as noun phrases. For the part of the input sentence "shows a document", the grammar rule (c) (a verb phrase consisting of a verb and a noun phrase) is applied. That is, the part is recognized as a verb phrase.

The syntax-analyzing process results in providing a tree structure of the syntax as shown in Fig. 9.

On the basis of the tree structure, the syntax analyzing unit 52 serves to restrict two or more part- f- speech candidates for each word stored in the word buffer (see Table 3) to one candidate. Table 9 shows the restricted result, in which "display" matches to a noun, "shows" matches to a verb, and "document" matches to a noun.

Table 9 Word Buffer

English Word Part of Speech Translated Word The Article ##i#j#j#if##i#f display Noun t r 4 41 shows Verb a Article document Noun At a step S85 (see Fig. 6), the translated equivalents to English words are saved from the translated-word buffers 23 shown in Tables 4 to 7 to the word buffer 22 shown in Table 9 by referencing the classification-code buffer 21 which stores the classification code specified by the user at the step S81.

If the translated-word buffers shown in Tables 4 to 7 have the translated word with the same classification code as the first priority classification code 003, the translated word is removed at first to the translatedword unit of the word buffer 22 shown in Table 9. In this example, translated words " #### (hyouji.souchi)" corresponds to "display" and

(hyoujisuru)" corresponds to "show" are removed at first because those translated equivalents have a classification code 003.

Then, it is checked whether or not the translatedword buffers shown in Tables 4 to 7 have the translated word with the same classification code as the second priority classification code 100. If it is yes, the translated word is removed to the translated-word unit of the word buffer 22. In this input sentence, such a translated word

(disupurei)" corresponds to the input word "display" and the translated word

(dokyumento)" corresponds to the input word "document" are removed to the translated-word unit. Since the third priority classification code is 000 in Table 2, the translated words left in the translated-word buffers shown in Tables 4 to 7 are removed to the translated-word unit of the word buffer 22.

As a result, the translated-word unit of the word buffer 22, as shown in Table 9, has the translated equivalents

(hyoujisouchi)",

(disupurei)" and

(chinretsu)" ranked in sequence for the input word "display",

(hyoujisuru)" and

(simesu)" ranked in sequence for the input word "show", and

(docyumento)" and

(bunsho)" ranked in sequence for the input word "document".

At a step S86 (see Fig. 8), the sentence generating unit 54 (see Fig. 2) generates the translated sentence.

In generating the sentence, the unit 54 serves to pick up the translated word with a higher rank out of the translated-word unit of the word buffer 22. The translated sentence is thus generated as

(Hyoujisouchi ha dokyumento wo hyouji suru)" as shown in Table 10.

As described above, in a case of generating the translated sentence, the sentence generating unit 54 enables to output a translated word to which classification codes specified by the user, in this case the information 00 3 and the Katakana 100, are added in precedence of any other translated word.

As another example, if the user selects only the classification code 100 representing the Katakana form, as shown in Table 11, the sentence generating unit 53 serves to output the translated sentence of

(Disupurei-ha-docyument-wohyoujisuru)".

If the user selects only the classification code oo3 representing the information field, as shown in Table 12, the sentence generating unit 53 serves to output the translated sentence of

(Hyouji.souchi-ha-bunnsho-wo-hyoujisuru)".

If the user selects no classification code, as shown in Table 13, the sentence generating unit 53 serves to output the translated sentence of

(Chinretsu-ha-bunnsho-wo-hyouj isuru)".

It is to be understood from the above description that this embodiment allows the user to specify any classification code with the corresponding rank from the predetermined classification code table, thereby it enables to generate the translated equivalent matching to the user's purpose.

Further, the user may add a classification code of which the user newly defines for any translated equivalent to any word in addition to the predetermined classification code.

For example, the user may add the codes 200 (user 1) and 201 (user 2) included in the classification code table shown in Table 1 to the third translated word

(Densanki)" and the fourth translated word

(Keisanki)", respectively to the input word "computer" shown in Table 7.

At the step S81 (see Fig. 8), if the user specifies the classification code 200, like another classification code, the translated equivalent with the classification code 200 takes a precedence of any other translated equivalent. It results in

(Densanki)" being output as "computer".

The translation machine is not limited to the abovementioned languages and as the source and the target languages, any other languages may employ in places of English and Japanese.

Table 10 Classification-code Buffer

(1) 003 (2) 100 (3) 000 Translated Sentence

Table 11 Classification-code Buffer

(1) 100 (2) 000 Translated Sentence

Table 12 Classification-code Buffer

(1) 003 (2) 000 Translated Sentence

Table 13 Classification-code Buffer

(1) 000 (2) Translated Sentence

It will be understood from the above description that each of the two embodiments of the invention comprises technical features which afford technical (ie. practical) advantages in a translation machine in providing an improved system for selecting between a number of translation equivalents of an input word for achieving enhanced translation efficiency.

Many widely different embodiments of the present invention may be constructed without departing from the spirit and scope of the present invention. It should be understood that the present invention is not limited to the specific embodiments described in this specification, except as defined in the appended claims.

Claims

CLAIMS:

1. A translation machine which is capable of setting a field of a translated equivalent in precedence and capable of changing a precedence rank of said translated equivalent by referring a translated equivalent finally selected with respect to a word in an input sentence having a plurality of translated equivalents after completing a translation, said translation machine comprising: means for adding an information of a postlearning to said translated equivalent selected finally so that a precedence rank of said translated equivalent is set as a next rank to a rank of a translated equivalent ouput at first in a case of performing a learning of a dictionary; means for detecting a field to be set so that said translated equivalent added said information of said post-learning is consulted with said dictionary in precedence.

2. A translation machine according to claim 1, wherein said translation machine further comprises a means for storing said adding means and said detecting means.

3. A translation machine according to claim 2, wherein said storing means is a central processing unit.

4. A translation machine according to claim 1, wherein said translation machine further comprises a memory table for storing said dictionary, grammer rules and tree structure converting rules.

5. A translation machine according to claim 4, wherein said dictionary includes a basic dictionary for storing all said translated equivalent.

6. A translation machine according to claim 5, wherein said translation machine further comprises a translation module for translating said input sentence described in said source language into a sentence described in said target language.

7. A translation machine according to claim 6, wherein said dictionary further includes a learning dictionary for storing said learning processed equivalents.

8. A translation machine according to claim 7, wherein said translation module includes a dictionary consulting and morphologic element analyzing means which is adapted to consult said dictionaries and for obtaining grammatical and translating information for each word so that a tense, a person, a numeral and so forth of said each word is analyzed.

9. A translation machine according to claim 8, wherein said translation module further includes a syntax analyzing means which is adapted to convert a sentence structure of said input sentence described in said source language into a sentence structure of a sentence described in said target language on a basis of said tree structure converting rules.

10. A translation machine according to claim 9, wherein said translation module further includes a sentence generating means which is adapted to generate a translated word on a basis of said sentence structure of said sentence described in said target language and said information obtained from said dictionaries and to output said translated word.

11. A translation machine according to claim 1, wherein said source language is an English and said target language is a Japanese.

12. A translation machine which is capable of performing a sentence structure analysis, a sentence structure transformation and a tranlated word generation, the translation machine being capable of outputting translated equivalents, said translation machine comprising: means for adding classification codes to each of translated words; means for specifying a classification code during a translation process; and means for outputting a translated word having added a classification code which is the same as said specified classification code in precedence in a case that said word of said input sentence has a plurality of translated word candidates.

13. A translation machine according to claim 12, wherein said specifying means inclues a main memory having a classification code buffer, a word buffer, and a translated word buffer.

14. A translation machine according to claim 13, wherein said classification code buffer includes a classification code table.

15. A translation machine according to claim 12, wherein said translation machine further comprises a means for specifying a plurality of classification codes with a priority during a translation process, and a means for outputting a translated word having a classification code which is the same as said classification code having a higher priority among said plurality of said specified classification codes in a case that there are a plurality of translated word candidates for a word in said input sentence.

16. A translation machine according to claim 13, wherein said classification code represents a field or a style of each of said translated equivalent.

17. A translation machine with the facility for accumulating a learning dictionary of user-preferred translation equivalents; characterised by means for choosing the latest registered equivalent automatically only if the user-specified subject and matter field corresponds to field information associated with that equivalent.

18. A translation machine substantially as hereinbefore described with reference to Figures 1 to 6 of the accompanying drawings.

19. A translation machine substantially as hereinbefore described with reference to Figures 7 to 9 of the accompanying drawings.