JP3999771B2 - Translation support program, translation support apparatus, and translation support method - Google Patents

Description

  The present invention relates to a translation support program, a translation support apparatus, and a translation support method for translating sentences in one language into sentences in another language, for example.

  There is a machine translation device that automatically translates a document in one language (source language) into a document in another language (target language).

  As a conventional machine translation device, referring to a translation memory which is a database storing a sentence (original sentence) written in a certain language and a sentence (translation) expressing the sentence in another language, A translation support apparatus that supports translation work is known (see, for example, Patent Document 1).

The flow of translation processing in a conventional translation support apparatus is as follows.
1. A previously translated bilingual sentence is registered in the translation memory in advance, and if there is a sentence similar to the translation target sentence in the translation memory, it is referred to and used for translation.
2. If there is no sentence similar to the translation target sentence in the translation memory, machine translation is executed to create a translation sentence (subordinate translation), and the translation is completed by appropriately modifying the subtranslation manually.
3. The translation target sentence and the completed translation are registered in the translation memory and reused when a new translation target document is translated.
JP-A-6-68142

  In the conventional translation support device, when the translation result of machine translation is manually corrected and registered in the translation memory, the correction of the original sentence (pre-editing) is easy even if the correction of the translation (post-editing) can be easily performed. There was a problem that it could not be done.

  This is because if the original text registered in the translation memory is changed from the original original text, it is difficult to reuse it when translating another document.

  For example, it is assumed that an html document and a rich text document having the same contents are created from a certain Japanese text document. Then, it is assumed that English translation documents are created based on the Japanese text documents and these are registered in the translation memory. At that time, it is assumed that the original text registered in the translation memory has been changed from the original text.

  It is assumed that the contents of the translation memory are used to translate an html document and a rich text document to create an English version html document and a rich text document.

  However, since the original text different from the original text is registered in the translation memory, the similarity rate with the data in the translation memory is low. Further, when the original sentence is corrected so that it can be easily translated by the machine translation device, it may become an artificial sentence that does not actually exist. In such a case, the possibility of not matching with the actual sentence is further increased.

  Therefore, when the original text is corrected, it is necessary to save the original text before correction separately and replace it with the corrected text when registering the translation memory.

  However, even in this case, there is a problem that the replacement work with the original text becomes very complicated especially when the original text is divided and combined. In addition, it is not uncommon to need to split and combine text. For example, if the original Japanese sentence is composed of a plurality of sentences connected by “ga”, the result of translation is better if the sentence is divided, or in the case of layout information in a document with format information such as an html file. If the sentences are divided and need to be combined for correct translation, there is no limit.

  In this way, considering the registration in the translation memory, the correction of the original text cannot be easily performed. Therefore, in reality, the correction of the machine translation result is mostly dependent on the correction of the translation.

  Users who can transcribe English by themselves from the beginning do not make full use of the machine translation function, and do a great deal of effort by overwriting and editing the translation. Even though we knew that there was a problem, there were many cases where we had to accept a machine-translated translation without being able to figure out how to fix it.

  The present invention has been made to solve such a problem, and can perform editing processing so that even if the original text is corrected, it can be restored to the state before editing, and the machine translation function and the translation assets in the translation memory are effectively used. The purpose is to provide a translation support program, a translation support apparatus, and a translation support method that can be utilized.

In order to achieve the above-described object, a translation support program according to the present invention includes a translation dictionary, a translation memory for storing a translation result, a computer having an editing history storage unit for storing a translation sentence and an editing history for the original sentence. in the translation support program which processes a textual other language translation, the computer, the original text, the translation means for generating a translation by machine translation on the basis of the translation dictionary, and the original text by said translation means said means in association translations be stored in the translation memory, when said string in at least one sentence of the translation memory stored original and the translated sentence has been edited, in that position, the An editing means for generating a text with a special symbol by adding a special symbol indicating an editing part to the text string of the text,
The original with the generated special symbols by the editing means, characterized in that to function as a means for storing in said edit history storing section as an editing history of the original textual.
In the above-mentioned translation support program, the computer translates the original text with the special symbol again according to the special symbol to generate a new translated text, and a new translated text generated by the re-translating means. May be associated with the original text with the special symbol as a translation result document and stored in the translation memory.

Translation supporting apparatus of the present invention, the textual one language, a translation means for generating a translation of the other language by machine translation based on the translation dictionary stored in advance, the translation results by the translation means a translation memory that stores in association with the translation and the original text, and editing history storage unit for editing history editing at least one of the translated translation and original text by said translation means is stored, the original text and the translation Editing means for generating an original sentence with a special symbol by adding a special symbol indicating an editing portion to the original character string at that position when a character string in at least one sentence of the sentence is edited And means for storing the original text with the special symbol generated by the editing means in the editing history storage section as an editing history of the original text.
In the translation support apparatus, a retranslation means for re-translating the original sentence with the special symbol in accordance with the special symbol to generate a new translation sentence, and a new translation sentence generated by the re-translation means as a document of the translation result And means for storing in the translation memory in association with the original text with the special symbol.

The translation support method of the present invention is a translation support method for translating an original sentence of a certain language into another language by a computer having a translation means, a translation dictionary, a translation memory, an editing means, and an edit history storage unit. but the original text of a language, and Luz step to generate a translation in other languages by machine translation based on year translation dictionary stored in advance, said translation means, said translation and the original text and storing in the translation memory in association with, if the string in the at least one sentence of said original and said translation is edited, in that position, the editing portion for strings of the original generating a textual with special symbols added is the editing means of special symbols indicating the original text with the generated special symbols, the edit history storing unit said editing means as the edit history of the original textual Characterized by a step of 憶. In the translation support method, the step of re-translating the original sentence with the special symbol by the re-translation means according to the special symbol to generate a new translation sentence, and the new translation sentence generated by re-translation and the editing means There may have a step of storing the translation memory in association with the original text with the special symbols as a document of the translation result.

  In the present invention, an original sentence with a special symbol is generated by adding a special symbol to an editing position for performing at least one of a division position, a combining position, and a single sentence editing of the selected character example, and the original editing history is generated. Is stored in the editing history storage unit, and by detecting the special symbol from the original text with the special symbol with reference to the editing history storage unit, the original text with the special symbol can be returned to the original text before editing.

  In addition, since a new translation sentence obtained by re-translating the original sentence with the special symbol is stored in the translation memory in association with the original sentence with the special symbol as a translation result document, the translation result with high accuracy can be reused.

  As described above, according to the present invention, it is possible to perform an editing process so that even if the original text is corrected, it is possible to return to the state before editing, and the function of machine translation and the translation assets in the translation memory can be effectively utilized.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of a translation support apparatus according to an embodiment of the present invention.
The translation support apparatus according to this embodiment is roughly divided into four parts: a sentence division processing unit 1, a translation processing unit 2 as a translation unit, an original / translation editing unit 3 as an editing unit, and a document output unit 4.

  The sentence division processing unit 1 includes an original document input unit 5 written in a source language to be translated, an original document automatic sentence division unit 6 that automatically divides the original document into one sentence according to a predetermined sentence division rule, And a sentence division rule table 7 in which predetermined sentence division rules referred to at the time of automatic sentence division are stored. The original document input unit 5 is configured by, for example, a keyboard, a mouse, a graphic user interface screen, and the like, and functions as an operation unit that performs a character string editing operation on at least one of the original sentence and the translated sentence. The predetermined sentence division rule stored in the sentence division rule table 7 may be stored in the form of a file in the memory, for example, and is described in advance, that is, set in the processing program of the sentence division processing unit 1. May be.

  The translation processing unit 2 includes a translation processing control unit 8 that controls the translation processing, a sentence (translation target sentence: first document) written in a source language (for example, Japanese language) to be translated, Translation in which a sentence (translation result document: second document (hereinafter referred to as a translated sentence)) written in another language, that is, a target language (such as English) is stored in association with a pair (one set) A memory 9, a translation memory search unit 10 for searching whether or not there is a sentence (similar sentence) similar to the original sentence composed of the original language divided into one sentence in the translation memory 9, and the original sentence automatically In particular, a machine translation processing unit 11 that translates into a sentence in a target language and a translation dictionary 12 as a translation dictionary that is referred to by the machine translation processing unit 11 are provided.

The translation processing unit 2 functions as a translation unit that generates a translation that is a target language document by machine-translating a certain original text based on a translation dictionary 12 stored in advance.
The translation dictionary 12 includes dictionary information for language translation (dictionary data such as Japanese → English dictionary, English → Japanese dictionary), morphological analysis, syntax analysis, semantic analysis, analysis rules for language conversion, conversion Rules are stored.

  The translation memory 9 stores a pre-translation document and a translation result document obtained by translating the pre-translation document in association with each other for the purpose of reusing the translation result by the translation processing unit 2. The translation memory 9 is used as part of a function for accumulating in the form of a database and improving the efficiency of translation work in order to reuse past translation results.

The machine translation processing unit 11 includes a morphological analysis, a syntax analysis, a semantic analysis, a language conversion processing unit 13 and a translation generation unit 14.
The original / translation editing unit 3 functions as a unit that generates a third document (see FIG. 6) by adding a special symbol to a position corresponding to a character string editing operation performed on the original by the operation unit. .

The original sentence / translation editing section 3 has an editing history storage section 15 in which an original history and an editing history of editing the translation are stored. The editing history storage unit 15 is realized by a memory, a storage area provided in the hard disk device, or the like. The original / translation editing unit 3 functions as means for storing the generated third document in the editing history storage unit 15 as the editing history of the first document. That is, the edit history storage unit 15 stores an edit history in which at least one of the original sentence and the translated sentence is edited.
The original / translation editing unit 3 divides the character string selected from the original using the parallel translation editing screen, which is an operation unit, based on the translation with special symbols stored in the editing history storage unit 15. It functions as means for estimating a position where at least one of combination and single sentence editing is performed, and adding a special symbol to the position to generate an original sentence with a special symbol. The original / translation editing unit 3 functions as a unit that stores the generated original with special symbols in the editing history storage unit 15 as an editing history of the original.

  The translation generation unit 14 converts the morpheme analysis, syntax analysis, semantic analysis, and language conversion processing unit 13 into morpheme analysis, syntax analysis, semantic analysis, and language conversion into meaningful phrases (character strings). Process to arrange in.

  The document output unit 4 outputs an original sentence / translation sentence in the form of a sentence in the form of a document, and includes a print driver, a display driver, etc. that output at least one of the original sentence, the translated sentence, and each corrected sentence. And hardware such as a printer and a display device.

  The hardware of this translation support apparatus is a computer including a CPU, a memory, a hard disk device, and the like, and a display device and a printing device connected to the computer. An operating system (hereinafter referred to as OS) for operating the entire computer system and a control program (hereinafter referred to as translation support program) for executing machine translation are installed in the hard disk device. These translation support program, OS, CPU, The memory and the like cooperate to realize processing operations of the sentence division processing unit 1, the translation processing unit 2, the original / translation editing unit 3, the document output unit 4, and the like.

  The operation of the translation support apparatus according to this embodiment will be described below with reference to FIG. FIG. 2 is a flowchart showing the entire processing of the translation support apparatus of this embodiment.

  In this translation support apparatus, when an original document document is input to the original document input unit 5 (step S201 in FIG. 2), the original document automatic sentence dividing unit 6 divides the original document into units of one sentence (step S202). In the original document automatic sentence dividing unit 6, in the case of a Japanese sentence, one sentence is divided with a punctuation mark as a delimiter. In the case of English sentences, one sentence is divided with a period, a colon, a semicolon, etc. as a delimiter. In addition, for example, in a document with format information such as parentheses, line feed symbols, and html, one sentence is divided with a line feed symbol, parentheses, etc. as a delimiter. In English, there are words with a period, such as “Dr.”, so the words that should be considered individually are compiled into a dictionary and referred to as appropriate. The delimiter information is stored (set) in advance in a memory or program.

  The original sentence divided into one sentence unit by the original document automatic sentence dividing unit 6 is sent to the translation processing unit 2 one sentence at a time until it becomes the last sentence (step S203), and the translation processing unit 2 performs the translation processing. Performed (step S204). The detailed contents of the translation process in step S204 will be described later.

  The translated text that has undergone translation processing is sent to the original text / translated text editing section 3 together with the original text. The original / translation editing unit 3 performs a translation and / or original text editing process (step S205). Details of the editing process in step S205 will be described later.

After the editing process of step S205 is performed, when the edited original sentence needs to be translated again, the original sentence / translation editing part 3 returns the sentence to be translated to the translation processing part 2. Then, the translation processing unit 2 determines the presence / absence of a document (step S203), and translation processing is performed (step S204).
Even if there are a plurality of sentences to be translated, the translation process is executed one sentence at a time for every sentence to be translated through the determination process in step S203.

  If the translation process is not necessary again (No in step S206), the original / translation editing unit 3 sends the original and the translation result to the document output unit 4 (step S207), and ends the processing operation as the original / translation editing unit 3. To do.

<Operation of translation processing unit 2>
Here, the translation process shown in step 204 of FIG. 2 will be described with reference to FIG. FIG. 3 is a flowchart showing the translation processing shown at step 204 in FIG.

  When the translation target sentence divided into sentence units by the sentence division processing unit 1 is input to the translation processing unit 2 (step S301), in the translation processing unit 2, the translation memory search unit 10 receives from the translation processing control unit 8. By searching the translation memory 9 using the translation target sentence as a key (step S302), it is determined whether a sentence similar to the translation target sentence exists in the translation memory 9 (step S303). The detailed contents of the translation memory search process by the translation memory search unit 10 in step 302 will be described later.

As a result of the search, if there is a sentence similar to the translation target sentence in the translation memory 9 (Yes in step S303), the translation memory search unit 10 determines that the similar sentence is a translation sentence by translation processing, and the translation result Is output to the display screen of the display device (step S304), and the translation processing for this translation target sentence is terminated.
When there is no sentence similar to the translation target sentence (No in step S303), the translation memory search unit 10 sends the translation target sentence to the machine translation processing unit 11 to execute the machine translation process.
The machine translation processing unit 11 performs source sentence analysis processing (step S305) including various processes such as morphological analysis, syntax analysis, semantic analysis, and language conversion on the translation target sentence received from the translation memory search unit 10, and translation generation. The machine translation process is performed by performing the process (step S306).
When the machine translation process ends, the machine translation processing unit 11 outputs the translation result to the original / translation editing unit 3 (step S304).

<Operation of Original / Translation Editor 3>
Next, the details of the processing of the original / translation editing unit 3 shown in step 205 of FIG. 2 will be described with reference to the flowchart of FIG.

  When the original text and the translation processing result are sent from the translation processing section 2 to the original text / translation text editing section 3, the original text / translation text editing section 3 stores the same contents of the original text / translation text in the editing history storage section 15 as initial values. At the same time, the original text / translation text is displayed on the display screen of the display device. The original sentence / translation editing process is executed on a sentence selected by the user from among the original sentences / translation sentences displayed on the display screen. This display screen is a parallel translation editing screen of the original sentence and the translated sentence, and functions as an operating means for performing an operation of selecting a character string to be edited with respect to at least one of the original sentence and the translated sentence.

  FIG. 4 shows a flow from when a user selects a sentence to be edited to when one editing operation is completed. The processing of FIG. 4 is performed as necessary until all editing operations are completed. Repeated.

  When the editing target character string selected by the user, that is, the editing target sentence is input to the original / translation editing unit 3 (step S401 in FIG. 4), the original / translation editing unit 3 determines whether the editing target sentence is a translation or an original sentence. Is determined (step S402). As a method of this determination, there is a method of determining where the input source (the translated text is input from the machine translation processing unit 11 and the original text is input from the translation processing control unit 8), that is, from where the input source is sent, and a user operation There is a method of judging on the screen or the sentence itself (because the original sentence and the translated sentence are separately displayed on the two-divided screen).

As a result of the determination, if the edit target sentence is a translated sentence (Yes in step S402), the original / translated sentence editing unit 3 performs translation editing (step S403), and outputs the edited result to the document output unit 4 (step S412). The editing process is terminated.
If the edit target sentence is an original sentence (No in step S402), the original sentence / translation sentence editing unit 3 determines whether or not to divide one sentence according to the editing content (step S404) and whether to combine a plurality of sentences. It is determined whether or not (step S407), and processing is performed according to the determination result.

  For example, when dividing a sentence (Yes in step S404), the original / translation editing unit 3 executes a sentence dividing process (step S405), and when combining a plurality of sentences, executes a sentence combining process. (Step S408) In cases other than these two cases, that is, when the editing content is only changed within one sentence (Yes in Step S404), the original / translation editing unit 3 executes single sentence editing processing ( Step S409). A special symbol is attached to the edited sentence by each processing such as sentence division, sentence combination, and single sentence editing. The contents of each process will be described in detail later using the example of FIG. If the selected sentence is long, splitting and combining may be performed at the same time. In this case, the processes of steps S405 and S408 are executed simultaneously. A combination with other single sentence editing processing is also conceivable.

  After these processes, the original / translation editing unit 3 performs an update process of the content of the original text editing history in the editing history storage unit 15 (step S406), and the processing content is stored in the editing history storage unit 15 in time series. Is done.

  After completing the editing of the original text, the user determines whether retranslation is necessary. The user who determines that retranslation is necessary operates the translation button on the display screen, and the user who determines that retranslation is not necessary operates buttons or key operations other than the translation button on the display screen. The editing unit 3 changes the processing content according to the user's operation after editing the original text (step S410).

  For example, when the translation button on the display screen is operated by the user, the original / translation editing unit 3 determines that retranslation is necessary (Yes in step S410). In this case, the edited original is converted into the translation processing unit 2. And a second translation process is executed (step S411).

  If the user performs another operation, the original / translation editing unit 3 determines that re-translation is unnecessary (No in step S410), skips the translation process (step S411), and does the translation need to be edited? A determination process of whether or not is performed (step S412).

  In the determination processing in step S412, when the next sentence (original sentence or translation) is designated as an editing target by the user, the original sentence / translation editing section 3 determines that translation editing is necessary (Yes in step S412), The translation editing is executed for the designated editing target sentence (step S403), and then the translation result sentence is output to the translation memory 9 and the document output unit 4 (step S413), and the editing process is terminated. The translation result sentence output from the original sentence / translation sentence editing unit 3 is stored (registered) in the translation memory 9 in association with the edited original sentence to which the special symbol is added.

  If the user does not specify the next sentence (original sentence or translated sentence) as an editing target and the translated sentence is unnecessary (No in step S412), the original sentence / translated sentence editing unit 3 skips the translated sentence editing process (step S403). Then, the edited sentence is output (step S413), and the editing process is terminated.

<Example of operation of original / translation editing unit 3 and registration to translation memory 9>
FIG. 5 shows the contents of the original text / translation text and the editing history storage section 15 before the editing process sent from the translation processing section 2 to the original text / translation text editing section 3. Using these examples, the specific operation of the original / translation editing unit 3 and the contents registered in the translation memory 9 will be described. When the original text / translated text is not edited, the contents (original text / translated text) in the editing history storage unit 15 are exactly the same as the original text / translated text.

In these examples, the translation result output from the translation process cannot be said to sufficiently reflect the contents of the original text and needs to be corrected.
For example, sentence 51 is an example in which “Tanaka” is not recognized as a person's name because it is written in hiragana. In sentence 52, “I am Tanaka,” but “ga,” is used in the meaning of light connection in Japanese, but in the translation result, it is interpreted as the meaning of reverse connection, and the reverse connection conjunction “although” is output. ing. Sentence 53 and sentence 54 are examples in which one sentence is divided into two parts for convenience of layout and is not correctly translated.

  In consideration of the above points, the user performs a sentence selection operation, and the original / translation editing unit 3 performs sentence division processing, sentence combination processing, and one sentence editing processing in accordance with the selection operation, and results of correcting the original sentence Is shown in FIG.

Sentence 61 is obtained by correcting hiragana notation to kanji notation. By using Kanji notation, “Tanaka” is interpreted as a personal name, and the correct translation result is output. In many cases, correct translation results can be obtained by changing the notation and expression so that the meaning is uniquely determined.
A sentence 62 and a sentence 63 are obtained by dividing an original sentence connected by “ga” into two sentences and completing each sentence as a sentence. If these sentences are translated again, of course, the conjunctive conjunction else does not appear in the translation. When such sentence division processing is performed, the editing history storage unit 15 stores the editing history in which the special symbol “@ number @” is inserted at the end of the divided first half character string and at the beginning of the second half character string. The The number between “@” and “@” is the ID number of the special symbol for identifying the insertion of the special symbol at this position. The originals with the same ID number are connected before editing. Indicates that it was. The numbers are normally given as serial numbers by the original / translation editing unit 3 when a sentence clause or sentence combination is detected.

  The sentence 64 is a combination of the sentences 53 and 54. The retranslation result of the combined sentence is meaningful. When such a sentence merging process is performed, the editing history storage unit 15 stores an editing history in which “@ number @” is inserted in the coupling part between sentences.

In the case of the combining process, there are not a plurality of original sentences having the same ID number, but there are an original sentence and a translated sentence having the same ID. Since the machine translation processing unit 11 re-translates the combined original sentence, the combined part in the translated sentence should not originally exist, but the translation result before combining stored in the editing history storage unit 15 (translated sentence) And the retranslation result are compared to estimate (identify) the translation portion corresponding to the original portion, and perform translation. For example, the sentence 64 is estimated (specified) to be cut after “This processing”. This is because the previous character string “This processing” stored in the editing history storage unit 15 and the re-translation result “This processing” are completely matched, so it is assumed that the break of the translated sentence is after processing. This is because the degree of coincidence between the two translation results is the highest.
That is, the translation processing control unit 8 applies a sentence break to the translation (the most translated result) stored in the editing history storage unit 15 based on the translation with edit symbols stored in the editing history storage unit 15. presume.

FIG. 7 shows an example in which the original sentence is corrected and the translation is re-translated by the machine translation processing unit 11 and further corrected.
This example is an example in which a modification that changes the translated word “processing” to “process” for “processing” in the sentence 74 is added. As described above, when a translation result that is syntactically correct is obtained by correcting the original sentence, a translation at a certain level can be obtained by only correcting a part of the translation result.

Therefore, a user who can transcribe himself / herself from the beginning does not need to create a translation from the beginning, and the translation effort can be greatly reduced.
For users who have difficulty in creating English sentences from the beginning, correct English sentences can be obtained with a slight correction, so that even if the translation result of machine translation is wrong, the correction can be made relatively easily.

FIG. 8 is a diagram showing the contents registered in the translation memory 9 from the above translation result.
The original text registered in the translation memory 9 is not the corrected original text but the contents of the editing history storage section 15, and a special symbol indicating a sentence division part or a sentence combining part is added to the original text string. (A third document form).

  In the translation support apparatus of this embodiment, such a registration method to the translation memory 9 is adopted. However, only when the original text is divided, the restored original text is registered in the translation memory 9. A method is also possible. According to this method, even if the translation memory search unit 10 does not have a special function, if the sentence to be translated matches the original original text before editing, it is paired with the original original text registered in the translation memory 9. The translated text can be output.

Hereinafter, the case of registering in the translation memory 9 with the more general content shown in FIG. 8 will be described.
<Translation memory search operation during new document translation>
The translation memory search operation when the new document shown in FIG. 9 is translated using the translation memory 9 shown in FIG. 8 will be described. The translation memory search is the process shown in step S302 of FIG.

  In this case, the translation memory search unit 10 performs a comparison process with one data (referred to as memory data 01) in the translation memory 9 for one sentence (referred to as translation sentence 01) to be translated, After the comparison process is completed, a comparison process with the next data in the translation memory 9 is performed. When this is repeated and the comparison process between the translated sentence 01 and all the data in the translation memory 9 is completed, the search process for the next translated sentence 02 is started.

  The reason why such a comparison method is used is to explain the operation of the present embodiment in an easy-to-understand manner. Even if another search method such as index creation is used for speeding up the inspection, the scope of this patent It does not deviate from. Depending on the contents of the comparison process, a plurality of translated sentences and a plurality of memory data may be compared together.

  10 and 11 are flowcharts showing a comparison process between the translation sentence 01 and one translation memory data item (referred to as memory data 01). FIG. 10 shows an exact match search process, and FIG. A similar sentence search process is shown.

  The translation memory search unit 10 first executes the complete sentence search process of FIG. 10, and then executes the similar sentence search process of FIG. 11 for the translation sentence 01 for which no memory data 01 having a complete match has been detected. To do.

In the following, first, the operation of the search process will be described using the flowchart of FIG. 10, and a description using a specific example will be added to the case where the example sentence of FIG. 9 applies.
<Translation memory search operation-exact search>
When a translation target sentence (translation sentence 01) is input to the translation memory search unit 10 (step S501), the translation memory search unit 10 first compares the translation sentence 01 with the memory data 01 to completely match each other. Whether or not (step S502). Here, the complete match means that the translated sentence 01 and the memory data 01, that is, the translated sentence stored in the translation memory 9 are not different one by one.

  As a result of the comparison, when the translation sentence 01 and the memory data 01 completely match, the translation memory search unit 10 determines that a matching sentence exists in the translation memory 9 (Yes in step S502), and selects the matching sentence. The memory data is 01 (step S503).

  The case where the result of step S503 is obtained corresponds to the case where the translated sentence 01 is the sentence 90 in FIG. In other words, even when the original text is edited so that it is written in kanji at the time of the first translation, the translation memory 9 stores the original hiragana text as shown in FIG. Therefore, the memory data 01 matches the translated sentence 01 in the same hiragana notation as the original, and the correct English sentence “I am Tanaka.” Is taken as the translation result. The above is the case where the data in the translation memory 9 is matched by performing a single sentence editing process.

  In the determination process of step S502, when the translation sentence 01 and the memory data 01 do not completely match (No in step S502), the translation memory search unit 10 precedes “@” in the original sentence of the translation sentence 01 and the memory data 01. A certain character string is compared to determine whether or not they completely match (step S504).

  As a result of this comparison and determination, if the translated text 01 and the character string preceding “@” in the original text of the memory data 01 do not completely match (No in step S504), the translation memory search unit 10 determines that the translated text 01 and It is determined that there is no matching sentence, and the search process is terminated.

  On the other hand, when the translated text 01 and the character string before “@” in the original text of the memory data 01 completely match (Yes in step S504), the translation memory search unit 10 indicates that “@” is the memory data 01. Is determined at the end of the sentence (step S505).

  As a result of this determination, if “@” is present at the end of the sentence of the memory data 01 (Yes in step S505), the translation memory search unit 10 further determines a character string in the translated sentence that does not match (this part is regarded as an unmatched part). Whether or not the matching part is called the matching part) matches the data in the translation memory 9 (referred to as memory data 02) having the @ symbol with the same ID number at the head part. Determination is made (step S506).

As a result of this determination, if the unmatched part and the memory data 02 are completely matched, the translation memory searching part 10 determines that there is a matching sentence in the translation memory 9 (Yes in step S506), and about the matching sentence Deletes the special symbol “@” and combines the memory data 02 after the memory data 01 to generate a combined statement (memory data 01 + 02) (S507).
The case resulting in S507 corresponds to the case where the translated sentence 01 is the sentence 91 in FIG.

When the translation memory search unit 10 searches the translation memory 9, first, “I am Tanaka,” is the original text 82 of the memory data “I am Tanaka, @ 1 @” “@ 1 @” Matches the previous part (step S504). Note that the number “1” of @ 1 @ is a serial number that is mechanically added in order when a new phrase or coupling position is detected.
In the investigation processing in the next step S506, the part “He is Nakata” matches the character string after @ 1 @ of the memory data against the original text 83 “@ 1 @ He is Nakata.”
Based on the result of the investigation, in the process of S507, an English translation result “I am Tanaka. He is Nakada.” That does not include the reverse conjunctive conjunction is generated.

  Thus, when creating the contents of the translation memory 9, when the original text is divided and registered in the translation memory 9, special symbols (@ 1 @, etc.) indicating that the divided data are linked to each other. ) Is added, so that the memory data matches a translation that is not divided as in the original. The above is a case where translation memory data registered by performing sentence division processing matches.

When the “unmatched part” does not completely match the memory data 02 in the determination process of step S506 (Yes in step S506), the translation memory search unit 10 uses a preset mathematical expression (function formula) to match the fuzzy match The rate is calculated, and the numerical value of the calculation result is compared with a preset reference value to determine whether or not the numerical value of the calculation result exceeds the reference value (step S508).
That is, when the similarity is equal to or greater than a certain value (Yes in step S508), the translation memory search unit 10 regards that the “matching part” of the translation sentence 01 matches the memory data 01, and translates As a result, a translation of the memory data 01 is output, and it is determined that the “unmatched portion” does not match the memory data (step S509).
If the similarity is less than a certain value (No in step S508), the translation memory search unit 10 determines that both the “matching part” and the “unmatching part” do not match the memory data, and is a perfect match. The search process is terminated.
This is because it may be better to reconsider the meaning of the sentence as a whole when the sentence part matches but the remaining part has a very low degree of coincidence. However, according to the user's will, the setting is changed to skip the determination processing in step S508, and the “matching part” is determined to match the memory data, and a mode for displaying the translation of the memory data is introduced. May be.

The case where the result as in step S509 is obtained corresponds to the case where the translated sentence is the sentence 92 in FIG. The first half of sentence 92 “I am Tanaka,” is the same as “I am Tanaka, but” the previous part of memory data vs. original sentence 82 “@”, but the second half “Born in Chiba” is memory Only the particle “ha” and the auxiliary verb “is” coincide with the original text 83 of the data, and the degree of coincidence is very low (3 out of 8 words). For this reason, the “matching part” is determined not to match the memory data together with the “unmatching part”.
In the determination process of step S505, when “@” is not at the end of the memory data (No in step S505), the translation memory search unit 10 determines that the unmatched part of the translation memory 9 in the process of step S504 is the next translation. It is determined whether or not it completely matches the target sentence (referred to as translated sentence 02) (step S510).
If the unmatched part matches the translated sentence 02 (Yes in step S510), the translation memory search unit 10 determines that there is a matched sentence, and the matched sentence is “@” in the memory data 01 for the translated sentence 01. For the previous part and the translated sentence 02, a joining process is performed with the part after "@" in the memory data 01 (step S511), and a translated sentence is generated.

  The case where the result as in step S511 is obtained corresponds to the case where the translated sentences are the sentence 93 and the sentence 94 in FIG.

  That is, when the translation memory search unit 10 searches the translation memory 9, first, the sentence 93 “This process” in FIG. 9 is “@ 2 @” of “This process is @ 2 @” of the original text 84 of the memory data. Matches the part before "" (step S504).

  Next, the portion of the sentence 94 “do as follows” in FIG. 9 matches the portion after “@ 2 @” (step S510). In step S511, a translation (translation result) “This process @ 2 @ is performed as follows” is generated based on the search result.

  As described above, when the data of the translation memory 9 is created, even if the original text is joined and registered in the translation memory 9, a special symbol (@ 2 @) indicating the joining history is inserted into the joining portion. As with the original, the memory data matches the translated text that is not combined.

  In addition, since a special symbol (@ 2 @) is inserted even in the translation part of the joint part corresponding to the joint part of the original sentence, the corresponding translation sentence can be output for each part of the original sentence that is not joined. . The above is a case where the translation memory data registered by performing sentence combination processing matches.

  In the process of step S510, when the “unmatched part” in the translation memory 9 does not completely match the translated sentence 02 (No in step S510), the translation memory search part 10 calculates the matching rate of fuzzy matching, and in step S508 Similar to the determination process, it is determined whether or not the similarity exceeds a certain value (step S512).

  As a result of the determination, if the similarity is equal to or higher than a certain value (Yes in step S512), the translation memory search unit 10 determines that the “matching part” in the translation memory 9 matches the translation sentence 01, and the translation result The portion before the “@” in the translation of the memory data is output, and it is determined that the “unmatched portion” did not match the memory data (step S513).

If the degree of similarity is less than a certain value (No in step S512), the translation memory search unit 10 determines that neither the translation sentence 01 or 02 matches the memory data, and a complete match search process. Ends. In addition, as in the case of the comparison determination process with the reference value in step S508, a mode for displaying the translation of the memory data of the “matching part” without performing the comparison determination in step S512 may be set. .
The case where the result as in step S513 is obtained corresponds to the case where the translated sentences are the sentence 95 and the sentence 96 in FIG.
That is, the sentence 95 matches the character string before “@” in the translation memory data 84, but the sentence 96 matches only the character string after “@” only in the verb “do”. The utilization is different and the degree of agreement is very low. Therefore, the translated sentence 01 is determined to be inconsistent with the memory data together with the translated sentence 02.

<Translation Memory Search Operation-Similar Sentence Match Search>
The similar sentence search in FIG. 11 is performed on the translation target sentence determined as having no matching sentence in the complete match search in FIG. Hereinafter, the similar sentence search process will be described with reference to the flowchart of FIG. A case where the memory data subjected to sentence division processing and sentence combination processing hit as a similar sentence will be described with reference to a specific example of FIG.
If the translation memory search unit 10 determines in the complete match search process shown in FIG. 10 that there is no sentence matching the translation target sentence, the translation memory search unit 10 performs a similar sentence search process.

  In this case, as shown in FIG. 11, when a translation target sentence without a matching sentence is input to the translation memory search unit 10 (step S601), the translation memory search unit 10 first searches the translation memory 9 to translate It is determined whether there is a similar sentence in the memory 9 (step S602).

Here, the presence or absence of a similar sentence is determined based on whether or not the similarity between the translation target sentence and the memory data 01 is a certain value or more.
In the case of Japanese, the degree of coincidence in the similar sentence search is obtained by calculating the ratio of the number of characters such as how many characters are matched in the character string. In the case of English, the ratio of the number of words indicating how many words are matched in the character string is obtained by calculation.

  Words with different utilization forms are determined to be “match” after being multiplied by a certain coefficient. If the similarity obtained as a result of the calculation is greater than or equal to the value set in advance by the user (Yes in step S602), the translation memory search unit 10 determines that a similar sentence exists in the translation memory 9. The similar sentence is set as memory data 01 (step S603).

If the similarity is less than the value preset by the user (No in step S602), the translation memory search unit 10 determines that there is no similar sentence in the translation memory 9, and then translates The degree of similarity between the sentence and the character string preceding “@” in the original text of the memory data 01 is determined (step S604).
Even when the similarity between the sentence to be translated and the character string preceding “@” is determined, the similarity is less than a certain value (No in step S604), the translation memory search unit 10 has no similar sentence. And the similar sentence search process is terminated.

  On the other hand, if it is determined that there is a similar sentence (Yes in step S604), the translation memory search unit 10 determines whether “@” is at the end of the sentence of the memory data 01 (step SS605).

  As a result of this determination, if “@” is at the end of the sentence of the memory data 01 (Yes in step S605), the translation memory search unit 10 further matches the unmatched part in the translated sentence that did not match in the comparison determination process in step S604. The degree of similarity with the memory data 02 is determined (step S606).

As a result of this determination, if the similarity is less than a certain value (No in step S606), the translation memory search unit 10 determines that there is no similar sentence, and ends the process.
If the degree of similarity is greater than or equal to a certain value (Yes in step S606), the translation memory search unit 10 determines that there is a similar sentence, deletes the special symbol @ number @ in the searched similar sentence, and stores the memory data. A statement is generated by combining the memory data 02 after 01 (step S607).

The case resulting in step S607 corresponds to the case where the translated sentence is the sentence 97 in FIG.
That is, sentence 97 in FIG. 9 “I am Tanaka,” and translation memory 9 vs. original sentence 82 “I am Tanaka,” and sentence 98 “Hare is Nakata.” The original text 82 “He is Nakata” is identical except that “I” and “he” are different.

  In the comparison determination process in step S605, when “@” is not at the end of the sentence of the memory data 01 (No in step S605), the translation memory search unit 10 further performs an unmatched part of the translation memory 9 in the comparison determination process in step S604. And the degree of similarity with the translated sentence 02 is determined (step S608).

If the similarity is less than a certain value (No in step S608), the translation memory search unit 10 determines that there is no similar sentence and ends the process.
On the other hand, if the degree of similarity is equal to or greater than a certain value (Yes in step S608), the translation memory search unit 10 determines that there is a similar sentence, and the similar sentence is “@” in the memory data 01 for the translated sentence 01. For the "previous part, translated sentence 02", a sentence with the part after "@" in the memory data 01 is generated (step S609).
The case resulting from step S609 corresponds to the case where the translated sentences are the sentence 98 and the sentence 99 in FIG.

  That is, “that process” in the sentence 98 of FIG. 9 is different from the memory data vs. the previous part of “@” in the original sentence 84 (“this” and “that”), and the sentence 99 is a pair of memory data. This completely matches the part after the “@” in the original text 84.

As described above, according to the translation support device of this embodiment, when the original text and the translation result are registered in the translation memory 9, if the original text is edited, the state of the original text before editing can be reproduced. Then, a document (third document) in which a special symbol (@ number @) is added to the position of the character example obtained by editing the original text is generated and stored in the editing history storage unit 15 as the editing history of the first document.
In addition, since the translation result document (fourth document) obtained by retranslating the generated third document is stored (registered) in the translation memory 9 in association with the third document, the accuracy of machine translation can be improved. Referring to the editing history storage unit 15 makes machine translation easier to translate.
Furthermore, the special text added to the third document can be restored to the original text state by deleting it according to the rules at the time of addition, so that the original text itself can be easily edited. Furthermore, editing such as sentence division, sentence combination, term replacement, and information addition can be easily performed on the original sentence.

  As a result, the ability of machine translation can be fully utilized to improve the efficiency of translation work.

1 is a block diagram showing the overall configuration of a translation support apparatus according to an embodiment of the present invention. The flowchart showing the flow of the whole process of the translation assistance apparatus of FIG. The flowchart showing the flow of the translation process of the translation assistance apparatus of FIG. The flowchart showing the flow of the edit process of the translation assistance apparatus of FIG. The figure which shows the content of the original sentence and translation before an edit process, and an edit history memory | storage part. The figure which shows the content of the original text after an original text edit, the retranslation result, and an edit history memory | storage part. The figure which shows the content of the original sentence after translation edit, a retranslation result, and an edit history memory | storage part. The figure which shows the registration content of a translation memory. Sentence to be translated in a new document. 3 is a flowchart showing translation memory search processing at the time of an exact match search in the translation support apparatus of FIG. 1. The flowchart which shows the translation memory search process in the case of a similar sentence search in the translation assistance apparatus of FIG.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Sentence division process part, 2 ... Translation process part, 3 ... Original sentence / translation edit part, 4 ... Document output part, 5 ... Original text document input part, 6 ... Original text automatic division part, 7 ... Sentence division rule table, 8 ... translation processing control unit, 9 ... translation memory, 10 ... translation memory search unit, 11 ... machine translation processing unit, 12 ... translation dictionary, 13 ... morphological analysis, syntax analysis, semantic analysis, language conversion processing unit, 14 ... translation Generation unit, 15... Editing history storage unit.

Claims (9)

In a translation support program for translating an original sentence in one language into another language by a computer having a translation dictionary, a translation memory for storing a translation result, an edit history storage section for storing an edit history for the translated sentence and the original sentence ,
The computer,
The original text, the translation means for generating a translation by the machine translation on the basis of the translation dictionary,
Means for storing in said translation memories in association with the translation and the original text by said translation means,
If the string in at least one sentence of said translation memory stored original and the translated sentence has been edited, in that position, by adding a special symbol indicating the edit portion on a string of the original Editing means for generating original text with special symbols,
A translation support program for causing an original text with a special symbol generated by the editing means to function as means for storing the original text as an editing history in the editing history storage unit. The translation support program according to claim 1,
The computer,
Re-translation means for re-translating the original sentence with the special symbol according to the special symbol to generate a new translation sentence;
A translation support program that functions as means for storing a new translated sentence generated by the retranslating means in a translation result document in association with the original sentence with the special symbol in the translation memory. The translation support program according to claim 1,
The special symbol is composed of a number between @ and @, and the number is an ID number of a special symbol for identifying that the special symbol is inserted at this position. Support program. In the translation support program according to claim 3,
A translation support program characterized by showing that originals with the same ID number among originals with special symbols are connected before editing. In the translation support program according to claim 3 or 4,
The translation support program according to claim 1, wherein the number is given by a serial number by the editing means when a sentence clause or sentence combination is detected. The original of a language, a translation means for generating a translation of the other language by machine translation based on the translation dictionary stored in advance,
A translation memory to store the translation results by the translation means in association with the translation and the original text,
An edit history storage unit for storing an edit history of editing at least one of a translated sentence and an original sentence translated by the translation unit;
If the string in at least one sentence of said original and said translation is edited, in that position, textual with special symbols by adding special symbol indicating the edit portion on a string of the original Editing means for generating
A translation support apparatus, comprising: an original history with a special symbol generated by the editing means; and means for storing the original history as an editing history of the original text in the editing history storage unit. The translation support apparatus according to claim 6,
Re-translation means for re-translating the original sentence with the special symbol according to the special symbol to generate a new translation sentence;
A translation support apparatus, comprising: means for storing a new translated sentence generated by the retranslating means in the translation memory in association with the original sentence with the special symbol as a translation result document. In a translation support method for translating an original text in one language into another language by a computer having a translation means, a translation dictionary, a translation memory, an editing means, and an editing history storage unit ,
It said translation means, and away step to generate a translation in other languages by machine translation based textual one language, the year translation dictionary stored in advance,
The translation means associates the original sentence with the translated sentence and stores them in a translation memory;
If the string in at least one sentence of said original and said translation is edited, in that position, a special symbol indicating an editing portion for strings of the original special added said editing means Generating a text with a symbol;
A translation support method comprising: a step of storing the generated original text with a special symbol in an editing history storage unit as an editing history of the original text. 9. The translation support method according to claim 8, wherein said computer further comprises re-translation means,
Re-translation means re-translates the original sentence with the special symbol according to the special symbol, and generates a new translation sentence;
A translation support method, comprising: a step of storing, in a translation memory, a new translated sentence generated by re-translating the editing means in association with the original sentence with the special symbol as a translation result document.

Images