JP2009116584A - Machine translation device and machine translation program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To display a different part between an original text and an original translation example in association with words and phrases in the translated translation example in a different manner from the other words and phrases, and to easily edit the different part in the translated translation example. <P>SOLUTION: In a machine translation device, a translation part 24 translates the original text in a first language by using a dictionary part 21 of a storage part 19, and retrieves the translation example similar to the original text from the example dictionary part 22 for the original text designated as a retrieval key. An example processing part 25 associates a different part between the original translation example and the original text, and associates the words and phrases in the translated translation example which should be edited for the different section with the words and phrases in the original text. A control part 23 displays, on a display part 17, in a different manner from the other words and phrases, every different part between the original translation example and the original text according to the operation instruction of a user. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention has a function of imitating the relationship between a first language sentence and a second language sentence obtained by translating the first language sentence to create a translation sentence into the second language sentence for the input original sentence of the first language. The present invention relates to a machine translation apparatus and a machine translation program.

  There is a machine translation device having a function of translating an original sentence of a first language to be translated into a sentence of a second language by a computer. In such a machine translation apparatus, a plurality of translation examples consisting of a parallel translation of a first language sentence and a second language sentence are stored in advance in a translation example database, and a translation example similar to the input translation target original sentence is searched. Then, it is presented to the user together with the original text to be translated, and the translation is supported. Since the original text of the translation example often does not exactly match the original text to be translated, it is rarely necessary to use the translated text of the translation example as a translation corresponding to the original text to be translated without editing. It will be used after editing the translation.

  Therefore, in order to reduce the editing of the translated text, the original text in the translation example and the corresponding phrase in the translated text are presented to the user, or the elements in the translation example that are common to the constituent elements of the translation target text are highlighted. Have been proposed (see, for example, Patent Document 1 and Patent Document 2).

  Also, contrary to presenting the corresponding words to the user, among the words in the translation of the translation example, the words other than the words corresponding to the words in the original text to be translated are displayed to the user in reverse video. There are some (see, for example, Patent Document 3).

Further, a plurality of pairs of parallel example sentences including information on correspondence between the first language example sentences and the phrases of the second language example sentences are stored, and a plurality of first language example sentences similar to the original text of the first language to be translated are stored. From the bilingual example sentence of the first sentence, the difference between the original sentence and the first example sentence of the searched parallel example sentence is determined in units of words, and the phrase of the second language example sentence corresponding to the difference part in the first language example sentence is determined. Create a compound sentence that replaces the relevant phrase with the phrase in the source text, and creates a translated sentence that replaces the phrase in the original sentence with the translation corresponding to the phrase. Output of at least one of one language example sentence, second language example sentence corresponding to the first language example sentence, synthesized sentence and translated sentence, and can perform natural and high-quality translation based on the example database (For example, see Patent Document 4) ).
JP 2003-330924 A JP 2005-339087 A JP-A-9-244504 JP 2006-11842 A

  However, according to the related art, the user needs to determine which part of the translated sentence of the translation example corresponds to the difference between the original sentence to be translated and the original sentence of the translation example by comparing the original sentence and the translated sentence. For this reason, in the case of a long sentence or a translation example having a plurality of different portions, it takes time and effort to specify the location in the translation of the translation example corresponding to the focused difference.

  The object of the present invention is to display the difference between the original text to be translated and the original text of the translation example in correspondence with the words in the translation of the translation example and distinguish them from other words and phrases, and to the portion corresponding to the difference in the translation of the translation example. It is to provide a machine translation device and a machine translation program that can easily edit the above.

  The machine translation apparatus according to the present invention accumulates a plurality of translation examples including a machine translation program, a dictionary unit that accumulates knowledge and rules necessary for translation, and a parallel translation of a first language sentence and a second language sentence registered in advance. A storage device storing the example dictionary unit, an input unit for inputting a source text of a first language to be translated and information necessary for operation, a source language of the first language to be translated and a second language after translation In a machine translation device comprising a display unit for displaying a translation of the above and a CPU for calculating and executing the machine translation program, the input original sentence to be translated in the first language is translated using the dictionary unit of the storage unit A translation unit that retrieves a translation example similar to the translation target text from the example dictionary unit of the storage device with respect to the translation target text using the translation target source text as a search key, and a translation example source text searched by the translation unit When An example processing unit for associating a difference with the translation target text and associating a phrase in the translation of the translation example that needs to be edited with respect to the difference, and according to a user operation instruction And a control unit that displays a portion corresponding to each difference between the original text of the translation example and the original text to be translated from other words and phrases on the display unit.

  According to the present invention, it is possible to display a difference between the original text to be translated and the original text of the translation example in correspondence with a word or phrase in the translation text of the translation example to distinguish it from other words and phrases, and to a portion corresponding to a difference in the translation text of the translation example. Can be easily edited.

  FIG. 1 is a configuration diagram of a machine translation apparatus according to an embodiment of the present invention. The machine translation apparatus 11 translates and outputs a first language into a second language. For example, a software program such as a machine translation program is installed in a general computer, and the software program is executed by the CPU 12. It is realized by. In the machine translation apparatus 11, a CPU 12, a ROM (Read Only Memory) 13, and a RAM (Random Access Memory) 14 are connected via a bus 15. An input unit 16, a display unit 17, and a storage device 19 are connected to the bus 15.

  The storage device 19 stores a machine translation program 20 and a dictionary unit 21 and an example dictionary unit 22. The machine translation program 20 includes a control unit 23, a translation unit 24, and an example processing unit 25. The control unit 23 includes a correspondence display control unit 23a, the translation unit 24 includes a translation example search unit 24a, a phrase translation unit 24b, and a translation reconstructing unit 24c. The example processing unit 25 includes a common phrase association. Means 25a and difference equivalent phrase association means 25b are provided. The dictionary unit 21 stores a vocabulary unit 21a, a morpheme analysis rule 21b, a syntax analysis rule 21c, a conversion rule 21d, a syntax generation rule 21e, a morpheme generation rule 21f, a connection rule 21g, and the like.

  Based on the input signal from the input unit 16, the CPU 12 reads out and executes a boot program for starting the machine translation device 11 from the ROM 13, and further reads out an operating system (not shown) stored in the storage device 19. The CPU 12 controls each device based on the input signal of the input unit 16, reads out the data of the machine translation program 20 and the dictionary unit 21 stored in the storage device 19 and the like, loads the data into the RAM 14, and reads out from the RAM 14. Based on the command of the issued program, a series of processes to be described later such as data calculation or processing is realized.

  The input unit 16 is an input means for inputting data and commands such as character data and file data of the original text to be translated in the first language, and is usually a pointing device such as a keyboard or a mouse / touch panel, a voice recognition or character recognition function, Alternatively, it is realized by an external storage medium reading device such as a CD drive, a network input device, or the like. That is, the input unit 16 inputs a command to be translated, such as a sentence to be translated, and various commands such as a translation instruction and a dictionary registration instruction. In addition, for input of translation target sentence, input such as reading from OCR, flexible disk, magnetic tape, magnetic disk, the Internet, or reading a sentence sentence taken with a microphone and converted into a natural language character string by a dictation device, etc. Is also included. Voice input using a microphone can also be used to input various commands.

  The display unit 17 is an output unit for data and translation results input from the input unit 16, and displays the screen and translation results on a display device such as a CRT or a liquid crystal display, or saves the translation results or the like in a file. and so on. That is, the display unit 17 controls various commands to the control unit 23 such as an instruction command to output a translation result that is an output of the translation unit 24 or to search and display a translation example for the input sentence from the example dictionary unit 22. The response from the unit 23 is displayed and output. The display unit 17 is usually a display unit such as various displays, but the translation result is output to a printing unit such as a printing machine, or a unit that directly outputs to a flexible disk, magnetic tape, magnetic disk, or other media. Transmission means for transmitting may be used. Moreover, the audio | voice output to a speaker may be sufficient.

  The dictionary unit 21 includes a vocabulary unit 21a, a morpheme analysis rule 21b, a syntax analysis rule 21c, a conversion rule 21d, a syntax generation rule 21e, and a morpheme generation rule 21f. The dictionary unit 21 is referred to when a translation process is performed by the translation unit 24. It consists of various information and knowledge.

  The example dictionary unit 22 stores a translation example including a parallel translation of a first language sentence and a second language sentence. In the embodiment of the present invention, an example having an example for translation between two languages, the first language and the second language, will be described, but it can be used for translation between three or more languages. A multilingual translation example dictionary that stores a plurality of examples as a set may be used.

  The translation unit 24 generates a translation for the input sentence based on the knowledge / information in the dictionary unit 21 and the translation example in the example dictionary unit 22 and presents it to the user. A translation example search means 24a for detecting a translation example using as a search key, a phrase translation means 24b for translating an arbitrary selected phrase in the input sentence, and a translation result obtained by the phrase translation means 24b A translation reconstruction means 24c is provided for reconstructing the translation by converting it into a form suitable for the corresponding part.

  The example processing unit 25 includes a common phrase association unit 25a and a difference equivalent phrase association unit 25b. The common phrase association unit 25a uses the syntax analysis unit (not shown) of the translation unit 24 and the knowledge / information of the dictionary unit 21, and the first language sentence (original text) and the second language in the translation example. Corresponding words / phrases in the sentence (translation) are associated with each other, and among the inputted original sentences to be translated, the words / phrases that are common to the original sentence of the translation example are also associated. Further, the difference equivalent phrase association unit 25b detects a portion corresponding to a different portion different from the original sentence of the translation example from the input sentence (translation target original sentence) from the translated sentence of the translation example, and the difference portion in the input sentence. Is associated.

  The control unit 23 has correspondence display control means 23a. Corresponding display control means 23a selects from the translation of the translation example when the user gives an instruction to select any word or phrase in the input sentence with the mouse, based on the result associated with the example processing unit 25. The part corresponding to the entered word is displayed in a different manner from other words, for example, by highlighting the part.

  Next, the control unit 23 that controls the overall movement will be described. FIG. 2 is a flowchart showing the contents of interactive processing with the user in the control unit 23. First, the control unit 23 determines whether there is an input of an original text to be translated from the input unit 16 (S1), and when there is an input of the original text to be translated, instructs the display unit 17 to display the original text to be translated. (S2).

  Next, the control unit 23 determines whether there is a translation instruction command (S3). When the translation command is sent, the control unit 23 instructs the translation unit 24 to start translation processing (S4). As a result, the translation unit 24 performs a translation process on the input source text to be translated into the second language translation using the knowledge and rules of the dictionary unit 21. When the translation process ends, the translation result is sent to the display unit 17 through the control unit 23 and presented to the user.

  Next, the control unit 23 determines whether or not a specific sentence is specified as a search key sentence in the original text to be translated, and there is a translation example search instruction command for starting the translation example search process (S5). When there is an instruction command, the search command of the example for translation is sent together with the information of the search key sentence instructed to the translation unit 24 (S6). When receiving the translation example search instruction command, the translation unit 24 searches the translation example registered in the example dictionary unit 22 by the translation example search unit 24a to search for a translation example having an original sentence similar to the search key sentence. The search result is returned to the control unit 23. The search result sent to the control unit 23 is sent to the display unit 17 and presented to the user.

  Here, in the example search for translation, when a plurality of examples having a predetermined similarity or higher are searched, the search is performed in order according to the similarity. Alternatively, only one example with the highest similarity may be searched. In the similarity determination, some criteria such as a character string base and a morpheme base are adopted. Furthermore, information on synonyms and synonyms may be used for similarity calculation.

  FIG. 3 is an explanatory diagram illustrating an example of a search result display screen in step S6 of FIG. As shown in FIG. 3, the upper part of the screen is composed of a translation target text area and its translation area, and the lower part of the screen is composed of a source text area and a translation text area of a translation example for the translation target text that is the search key sentence.

  The translation target original text input from the input unit is displayed in the translation target source text area at the top of the screen, and the result translated by the translation unit 24 using the translation instruction command is displayed in the translation target translation area at the top of the screen. Then, in the lower part of the screen, the original text and the translated text of the translation example for the translation target text that is the search key text are displayed as a pair.

  FIG. 3 shows the case where the results of the searched translation example are two. The parts colored in the original translation example are “The general name of our laboratory is the KM Lab.” Is the difference part. In this embodiment, the original text to be translated is colored immediately after the search for the translation example (the state in FIG. 3) as “the official name of our laboratory is the Knowledge Management Laboratory.” This is done for the part corresponding to the difference in the original text “The official name of the product is KnowledgeMeister.” Or the translated text “This product is called the Knowledge Meister. Is selected, the display is switched to display the difference from the selected translation example.

  The original text to be translated and the original text of the first translation example searched are color-coded in the same color, assuming that the following parts are the corresponding different parts.

・ Original “general” for translation and “official” for translation examples
・ Translated source text “KM Lab” and translation examples “Knowledge Management Laboratory”
When the user selects the second translation example, the color-coded display is switched so as to indicate the correspondence between the differences in the following parts.

・ Original “general” for translation and “official” for translation examples
・ "Our laboratory" for the original text to be translated and "the product" for the original text for translation
The original text to be translated "the KM Lab" and the original text for translation examples "KnowledgeMeister"
Next, in step S7 of FIG. 2, the user selects a specific translation example in a state where the search result of the translation example as shown in FIG. 3 is displayed, and the translation of the translation example is applied to the translation target original sentence. It is determined whether or not a translation example translation output command for output as a translation has been input (S7). This translation example translation output command is selected by, for example, clicking the specific translation example in FIG. 3 with a mouse. If there is an example translation output command for translation, as shown in FIG. 4, the translation of the selected translation example is displayed in the translation area corresponding to the translation target original text as a search key (S8). FIG. 4 shows the case where the user clicks on the translation of the translation example displayed at the top with the mouse, the translation area of the translation example is highlighted, and the translation of the translation example is displayed in the translation area for the translation target original sentence. Is done. That is, the selected translation example is displayed in the translated text area of the fifth translation target original text. In this example, the translation of the translation example as a search result is clicked, but the original sentence of the translation example may be selected.

  Subsequently, the control unit 23 selects a word / phrase displayed in the original text area or the translated text area of the original text to be translated, and the command operation of a word / phrase correspondence display instruction command for distinguishing and displaying both corresponding words / phrases from the input section 16. It is determined whether or not an input has been made (S9). When the instruction operation of the phrase correspondence display instruction command is input, the selected phrase is determined, and the display control for distinguishing from the other phrases is performed, such as highlighting the phrase (S10). For example, if the fifth “laboratory” in the source text region of the source text to be translated is double-clicked with the mouse, display control such as highlighting “laboratory” in the target text region corresponding to this phrase is performed. Details of this processing will be described later.

  Next, the control unit 23 determines whether or not a phrase translation instruction command for obtaining a translation result for the phrase in the original text region of the original text to be translated is input from the input unit 16 (S11). When the command operation is input, phrase translation processing is performed (S12). That is, a phrase translation command for the selected phrase is sent to the translation unit 24, the translation result is obtained from the translation unit 24, and the translation result is displayed on the display unit 17. Details of this processing will also be described later.

  In the state where the phrase translation result is displayed, whether or not a translation restructuring instruction command for reconstructing the translation is input by reflecting the translation selected from one of the translation candidates for the difference phrase. When the determination is made (S13) and the translation restructuring instruction command is input, the selected translation candidate, the translation source text and the translation information are sent again to the translation unit 24 to reconstruct the translation (S14). The reconstructed translation is displayed as a translation in the translation area for the translation target original. Details of this processing will also be described later.

  In addition, the control unit 23 determines whether or not an instruction to end the entire process has been input from the input unit 16 (S15). When the instruction to end the entire process has been input, all the processes are ended. Further, it is determined whether or not any other command is input (S16). When another command is input, processing corresponding to the input is performed (S17). In this way, the control unit 23 performs processing in response to various commands from the input unit 16, and ends the processing based on the overall processing end instruction.

  Next, the flow of processing from when the original text to be translated is input to the translation section 24 until the target language translation is generated will be described. FIG. 5 is a flowchart showing the processing contents of the translation unit 24 in step S4 of FIG.

  The dictionary unit 21 is composed of various information and knowledge that are referred to when executing each step of the translation process. The translation unit 24 generates a translation for the input sentence based on the knowledge / information in the dictionary unit 21 and the translation example in the example dictionary unit 22 and presents it to the user. As an overall flow, if there is a useful translation example in the example dictionary unit 22, a translation is generated based on that, and if there is no useful translation example, only the knowledge / information of the dictionary unit 21 is used to translate the translation. Generate.

  First, when an input sentence, which is a source sentence to be translated, is sent to the translation unit 24, a dictionary lookup process / morpheme analysis is performed on the input sentence using the vocabulary part 21a and the morpheme analysis rule 21b in the dictionary part 21. To obtain various information necessary for the translation process, such as the part of speech and the type of utilization, the translation, and the like (S51). As a result, the part of speech of words constituting the input sentence and candidates for utilization are output.

  Subsequently, an example search is performed on the example dictionary unit 22. That is, it is detected whether translation examples having the same original sentence as the input sentence are accumulated (S52). At this stage, morphological analysis results and various information for each word in the original text are acquired, so information on katakana and kana notation, kanji and hiragana notation, and presence / absence of kana are also available. Has been acquired. With these pieces of information, even if the translation example does not completely match the input sentence with one character, it is possible to perform a vibration absorption process that does not recognize the difference regarding the fluctuation of these notations as a difference. On the other hand, a configuration in which the search is successful only when there is an example of a completely matching translation that does not absorb the fluctuation of the notation is also possible. Moreover, even if there is a difference between characters and words, a configuration may be adopted in which the number of different characters regarded as the same value or the lower limit value of the ratio of the number of different words is provided to control the possibility of search. This makes it possible to extract translation examples that are similar in how words / characters are used. In such a case, it is not necessary to set the timing of the example search after the dictionary lookup processing / morpheme analysis, and the configuration may be performed at the beginning of the processing of FIG. If the result of the example search in step S52 is successful (S53), the translation of the searched translation example is generated as the translation of the input sentence, and the process is terminated.

  On the other hand, when the result of the example search is unsuccessful, that is, when there is no translation example that can be used as it is in the translated sentence, the syntax analysis process is continued (S54). In this parsing process, a sequence of words, parts of speech, and usage is input, and the syntax of the input sentence is analyzed using the parsing rules 21c to construct the syntax structure of the source language.

  When the syntax structure of the input sentence is obtained, conversion processing from the source language syntax structure to the target language structure is performed using the conversion rule 21d (S55). At this stage, it is determined in what syntax the target language sentence is translated and what kind of translation is output. Then, final generation of the target language sentence is performed using the syntax generation rule 21e and the morpheme generation rule 21f (S56), and the translation process for one input sentence is terminated.

  In the state where the generation of the translation in step S56 is completed, the part of speech, utilization, attribute information, etc. of the word string constituting the translation have been obtained, and which phrase in the translation corresponds to the constituent element in the original sentence. Information is also available.

  FIG. 6 is an explanatory diagram of an example of the data structure of the original sentence information and the data structure of the translated sentence information obtained by analyzing the structure of the input sentence in the translation unit 24. In FIG. 6, the data structure of the original sentence information of the following example sentence is shown in FIG. 6A, and the data structure of the translated sentence information is shown in FIG. 6B.

Original: The official name of our project is Unexplored Software Creation Project.
Translation: The official name of our project is the Unexplored Software Creation Project.

  Each component of the original sentence and the translated sentence has part-of-speech information, other attribute information, and an array number indicating which array of the other data structure corresponds to each other. An element containing “−1” indicates that there is no corresponding phrase.

  Next, an arbitrary word or phrase in the original sentence displayed in the original sentence area or the translated sentence displayed in the translated sentence area is selected, and an operation command for distinguishing and displaying the corresponding phrase in the other area is displayed on the control unit 23. Processing when input is described. Here, the case where the distinction display is reversed is described.

  FIG. 7 shows the flow of processing from when there is a corresponding phrase display display command in the control unit 23 until the translated sentence or the phrase of the original sentence corresponding to the selected phrase from the original sentence or the translated sentence is highlighted. It is a flowchart to show. The contents of this process differ depending on the type of sentence containing the selected word / phrase. First, it is determined what type of sentence (target sentence) includes the selected word / phrase (S21). If the sentence is an original sentence or a translation sentence that has been subjected to translation processing, since data as shown in FIG. 6 has already been obtained, the phrase corresponding to the selected phrase is obtained from this data, and the step Proceeding to S27, a reverse display command for the corresponding word is sent to the display unit 17. Inversion is not performed for words to which "-1" is assigned as corresponding word information in the data. This is the end of the corresponding phrase display process.

  When the sentence with the selected word / phrase is a sentence using the translation of the translation example similar to the original sentence, detected as the input key, as the translation sentence of the translation result, as detected in the fifth sentence in FIG. Then, the process flow from step S22 is followed. First, through the translation unit 24, using the knowledge and information of the dictionary unit 21, the original text to be translated (input original text), the original text of the translation example used (first language sentence) ("The official name of FIG. equivalent to our laboratory is the Knowledge Management Laboratory.), a translation of the translation example (second language sentence) (corresponding to "This laboratory is formally called the Knowledge Management Laboratory" in Figure 4) Dictionary lookup and morphological analysis are performed (S22), and then syntax analysis of the original / translated sentence of the translation example is performed (S23). This is because the original text data of the original text information and the data structure of the translated text information as shown in FIG. 6 are not obtained in the case of a sentence using a translation example.

  FIG. 8 is an explanatory diagram of an example of an analysis result after the original English sentence of the sentence using the translation example is subjected to the syntax analysis. Here, only the modification relationship between elements is shown and the relationship name is omitted, but in reality, an analysis structure is obtained together with the relationship name. The same analysis is performed on the translated Japanese sentence.

  When the analysis result is obtained, a process of converting the English sentence (first language sentence) of the translation example into the structure of the second language sentence is performed, and the candidate words of the second language corresponding to the constituent elements in the English sentence Get the structure. In addition, in the conversion process to the first language sentence for the Japanese sentence analysis result structure of the translation sentence (second language sentence) of the translation example, the candidate for the first language phrase corresponding to the constituent element in the Japanese sentence. (S24).

  FIG. 9 is an explanatory diagram showing an example of the conversion result of the original text shown in FIG. In the figure, “SW” indicates which word in the English sentence the Japanese word in the figure is derived from. What is indicated by “TW” is a list of translation candidates that exist as Japanese translation candidates. Multiple items are listed separated by commas.

  On the other hand, FIG. 10 is an explanatory diagram of an example of a result obtained by assigning the corresponding translation of the original sentence to the translation of the sentence using the translation example without changing the sentence structure of the translation. That is, an explanation of an example of the result of adding the corresponding English translation as “TW” information in the Japanese structure without completely converting the result of parsing the Japanese sentence in the second language into English as the first language FIG.

  In this embodiment, for the sentence in the second language, after the syntactic analysis, the conversion process is not performed, and the process until the corresponding translation candidate for the first language is obtained without changing the structure of the second language. For the second language sentence, it is possible to obtain a translation of the corresponding first language directly after dictionary lookup / morpheme analysis without performing syntax analysis. Since the syntactic information can be used for obtaining the corresponding translated word, more appropriate candidate words can be obtained.

  When the translation information for each component of the source sentence (first language sentence) and the translation sentence (second language sentence) in the translation example is obtained, the common phrase matching unit 25a determines which element and which element in the source sentence and the target sentence. Is checked (corresponding phrase determination), and a pair of corresponding elements is extracted from the analysis / conversion result of the first language sentence and the analysis result of the second language sentence (S25). The outline of the processing flow is shown below.

  [1] The following processing is performed for each component of the analysis result of the first language.

  [1-1] The first language sentence headword (obtained from the analysis result equivalent to FIG. 8), the second language sentence analysis result (figure enclosed in ellipses such as “the”, “name”, “of”) 10), an element included in the TW candidate is searched, and a translation corresponding to the headword is extracted as a corresponding candidate. If there are a plurality of pairs, a plurality of candidate pairs are extracted.

  [1-2] Check whether one of the translation candidates in the first language (obtained from the conversion result corresponding to FIG. 9) matches the component in the analysis result of the second language sentence. Extract as a candidate for correspondence. If there are a plurality of pairs, a plurality of candidate pairs are extracted.

  [2] As a result of the above [1], if there is no overlap in the correspondence and the correspondence is uniquely determined, it is determined as a corresponding phrase. In addition, even when the number of elements does not match, such as when one element in one structure matches two or more continuous elements in the other structure, it is extracted as a corresponding phrase.

  [3] As a result of the above [1], when the correspondence is ambiguous, for example, when the same word is used twice, it is directly modified in the analysis result with other corresponding phrases Further, based on the presence / absence of crossing of the modification relationship and the coexistence relationship with other corresponding word / phrase candidates in the partial structure, a more likely correspondence relationship is selected to eliminate the ambiguity.

  [4] When the corresponding phrase candidates are extracted by the above [3], the analysis result structure of the first language sentence (corresponding to FIG. 8) and the analysis result structure of the second language sentence (corresponding to FIG. 10) of these corresponding phrase candidates ), A structurally continuous portion is detected.

  In the examples of FIGS. 8 to 10, it is determined that the following words / phrases correspond as a result of [2] [3].

(English) official = (Japan) official (English) name = (Japan) name (English) of = (Japan) (English) our = (Japan) laboratory = (Japan) laboratory = (Japan) Institute (English) management = (Japan) Management (English) knowledge = (Japan) Knowledge As shown in Fig. 9, there are two "laboratory" and "laboratory", but according to [3], direct modification relationship with other components Using this information, the corresponding elements can be identified one-on-one.

  Subsequently, as a result of [4], the following correspondence is extracted. The second and third are noun phrases that form part of the first sequence. Such partial phrases are also detected.

(English) official / name / of / our / laboratory = (Japan) official name of our laboratory (English) official / name = (Japan) official name (UK) our / laboratory = (Japan) our laboratory (UK) knowledge / management / laboratory = (day) Knowledge Management Laboratory As a result of the above processing, the original text and the translated text of the translation example are represented as a data structure similar to the data structure of the original text and the translated text after the translation processing of FIG.

  FIG. 11 is an explanatory diagram of an example of the data structure for the translation example, and shows an example of the data structure for the following translation example. FIG. 11A shows the data structure of the original sentence, and FIG. 11B shows the data structure of the translated sentence.

Original: The official name of our laboratory is the Knowledge Management Laboratory.
Translation: The Institute is formally called the Knowledge Management Institute.

  Next, when the process for determining the correspondence between the original sentence and the translated sentence in the translation example is completed in step S25 of FIG. 7, the translation target original sentence (input original sentence) and the translated sentence of the translation example are subsequently converted by the difference equivalent phrase association unit 25b. A matching process for matching is performed (S26). For the original text to be translated, a dictionary lookup / morpheme analysis result is obtained in step S22. Further, in the process of searching for the initial translation example, a difference word / phrase between the original text to be translated and the original text of the translation example is determined. That is, the words are colored and displayed in FIGS.

  Therefore, referring to the data structure for the original text of the example for translation as shown in FIG. 11, corresponding phrase information of the data structure of the original text to be translated is given as follows.

  [1] For common elements that are not different between the original text of the translation example and the original text to be translated, information on the corresponding words and phrases between the original text of the translation example and the translated text (the rightmost of the individual elements of the array in FIG. 11A) (Numeric information in the column) is copied to the data structure of the original text to be translated.

  [2] For the elements that are different, the following processing is performed using the information of the elements of the differences in the corresponding translation examples.

  [2-1] When the number of words of the constituent elements of the difference portion is 1: 1, information on the corresponding phrase with the translation of the difference element in the original sentence of the corresponding translation example (individual elements of the array in FIG. 11A) Copy the numerical information in the rightmost column of the element.

  [2-2] If the number of constituent elements in the difference part is other than 1: 1, the translation example and the source text to be translated are connected with continuous elements and the data structure is changed one-to-one. Information of the corresponding phrase with the translation of the difference element in the original sentence (number information in the rightmost column of each element of the array in FIG. 11A) is copied.

  FIG. 12 is an explanatory diagram of an example in which corresponding phrase information is added to the data structure for the translation example of FIG. 12A shows the original text data after the change of the translation example, FIG. 12B shows the translated text data after the change of the translation example, and FIG. 12C shows the original text data after the change.

  As shown in FIG. 4, in the original text data, the “general” of the original text to be translated, which is the first difference between the original text of the translation example and the original text to be translated, has a one-to-one difference word number from the original text of the translation example. Therefore, the corresponding phrase information of the original “official” in the translation example is copied as its corresponding phrase information. The second difference part “KM lab” has a different number of words from the original difference part “Knowledge Management Laboratory” in the translation example, so the difference part “Knowledge Management Laboratory” in the original example translation and the correspondence in the translation The “Knowledge Management Laboratory” part of the phrase part is integrated into one element, and the input source text side also integrates “KM lab” into one element, and the data structure is changed so that each has a one-to-one correspondence. Then, after resetting the corresponding phrase information of the translation example, the information is copied as the corresponding phrase information of the original text data. As described above, when the corresponding phrase determination process is completed, the process proceeds to a process of highlighting the corresponding phrase of the phrase selected by the user (S27).

  In the present case where the translation of the translation example is used as the translation of the input source text, the data in FIG. 12B is treated as the translation data, and the display is the same as the processing in the sentence subjected to the translation processing described above. It can be performed.

  FIG. 13 is an explanatory diagram showing an example of the case where a word / phrase in the original text is selected on the display screen of the translation example in step S27 of FIG. 7, in which “general” in the original text is selected and instructions for displaying the corresponding word / phrase are shown. In this example, “formally” is highlighted as a corresponding word.

  FIG. 14 is an explanatory diagram showing another example when a phrase in the original text is selected on the display screen of the translation example. Corresponding by selecting “KM lab” or “KM”, “lab” in the original text In this example, “KM lab” and “Knowledge Management Laboratory” are highlighted as corresponding ranges when a phrase display instruction is given. When a word is selected for a word in the translation area instead of the word in the original text, the corresponding word in the original text area can be highlighted in the same manner as in the original text.

  Next, in the determination in step S21, when the translation process is not the original text or the translation text, and the translation example is not used, the case where a completely matching translation example is used as a result of the translation process, This corresponds to the case where the translated text is input. In this case, the process from step S28 to S31 is performed. Since the process from step S28 to S31 is the same as the process performed for the translation example in steps S22 to S25, the description thereof is omitted.

  As described above, it is possible to know the phrase of the translation example to be edited by instructing the corresponding phrase display for the difference between the translation examples. Although this example is a short sentence, it takes time for humans to determine the correspondence when it becomes a long sentence. However, according to the embodiment of the present invention, the corresponding phrase display instruction is given to the difference in the translation example, so that it is edited. It is possible to reduce the burden for knowing the words and phrases in the power translation example.

  Next, processing for selecting an arbitrary phrase in the original text, obtaining the translation result, and reconstructing the translation based on the translation result will be described. FIG. 15 is a flowchart showing the flow of phrase translation processing in the phrase translation unit 24b of the translation unit 24. The phrase processing means 24b of the translation unit 24 obtains a translation result for the selected word / phrase.

  When a phrase composed of an arbitrary number of words in the original text is selected from the input unit 16 and a phrase translation instruction is given, the control unit 23 sends the information on the original text and the information on the range of the selected word to the translation unit 24. (S31). When the original text information and the selected word / phrase range information are input to the translation unit 24, the phrase translation unit 24b causes the selected word / phrase to be translated (S32). The phrase translation unit 24b translates the entire original sentence and extracts a partial word string in the translated sentence corresponding to the selected phrase. The extraction result also includes information on possible translation candidates for each individual word constituting the partial word string. The whole source text is translated in order to deal with the case where the translation is determined by the relationship with other words, but a simple embodiment is also possible in which each word is simply looked up in a dictionary and its translation candidate is found. is there.

  FIG. 16 is an explanatory diagram showing an example of the phrase translation result display in step S33 of FIG. 15, showing a display example of the phrase translation result when “general” in the translation target text shown in FIG. 13 is selected. Yes. FIG. 17 is an explanatory diagram showing another example of the phrase translation result display in step S33 of FIG. 15, and shows a display example of the phrase translation result when “KM lab” is selected.

  As shown in FIGS. 16 and 17, when there are a plurality of translation candidates, other candidates can be referred to or selected from the pull-down menu. In FIG. 16 and FIG. 17, the inverted translation is the translation to be selected. When the “Reflect to translation” button is clicked with the translation selected in this way, the translation reconstruction process is started. The

  FIG. 18 is a flowchart showing the flow of translation reconstruction processing in the translation reconstruction means 24c of the translation unit 24. First, the range (replacement phrase range) of the phrase in the translation corresponding to the phrase selected from the translation target text is acquired (S41), and the attribute and other information of the phrase in the replacement translation are taken into account. Arrange (S42). Then, the phrase replacement is reflected in the translation (S43).

  That is, in the translation reconstruction process in the translation reconstruction unit 24c of the translation unit 24, when the range of a phrase in the translation is acquired, the translation reconstruction unit 24c refers to the attribute and other information of the phrase to be replaced, and the phrase Considering whether the translation result can be replaced as it is, or if it is inflected or if it is a Japanese verb, it is necessary to add an auxiliary word, etc., and conversion necessary for completing the translation is performed.

  For example, in the case of an English noun, various arrangements within the range used in the morpheme generation process at the time of normal translation are possible, such as the use of plural forms, the change of ownership, the use of Japanese idioms, and the conversion of states. In FIG. 16, “general” is selected, but based on the connection rule 21 g of the dictionary unit 21, the “formally” attribute information of “formally” in the translated sentence data is referred to as “continuous form + {ha}”. To "". The same applies to the case where “general” is selected, and “ha” is added and converted into “general” after being used in the continuous form.

  FIG. 19 is an explanatory diagram showing an example of a screen display after the phrase translation result in step S43 of FIG. 18 is reflected in the translated sentence and the translated sentence is reconstructed. The phrase translation result for “general” in the original sentence is shown. The example of a screen display after reflecting in a translation and reconstructing a translation is shown. FIG. 20 is an explanatory diagram showing another example of a screen display after the translation of the phrase is reflected by reflecting the phrase translation result in step S43 of FIG. 18, and the phrase translation for “KM lab” The screen display example after reflecting a result in a translation and reorganizing a translation is shown.

  In the above description, the source text “general” is selected and multiple candidate words are displayed, but conversely, the translation words in the source text area of the input source text are displayed in reverse video by selecting the translation words in the translation example. You can do it. FIG. 21 is an explanatory diagram showing an example of a display example of phrase translation results in step S33 of FIG. 15 when a translation word of a translation example is selected.

  When you select `` formally '' in the translation example translation in the translation display area, it is not semantically compatible, but the difference in the translation corresponding to the difference between the source text to be translated and the translation example source text A translation candidate (obtained as a phrase translation result) corresponding to the phrase “general” is presented in a pull-down menu. An appropriate word / phrase can be selected from the translation words presented in this pull-down menu and reflected in the translation. At that time, if the dictionary unit has a learning dictionary, the selected translation word can be stored in the learning dictionary. Thus, when “formally” in the translation example is selected, it corresponds to the phrase “general” in the original text as a difference part in the translated text that does not correspond semantically but corresponds to the difference in the original text. It is characterized by the appearance of translation candidates.

  Moreover, you may comprise so that it may refer not only to the attribute of the phrase in the translation to replace but to the information of the phrase before and after that. For example, when it is necessary to change the usage depending on the word that comes later, such as in Japanese, it is necessary to re-determine and reconnect the connectivity with the immediately following word (example: “Because it is important”, “So” is connected to the adjective verb form of “Important”, but “Large” cannot be connected and it is necessary to say “Because it is large”).

  Conversely, in the translation reconstruction process, it is also effective to adjust the connection such as the utilization form of the translation element of the part that is not the difference. For example, in the case of an English translation, this corresponds to a case where the use of the predicate verb is changed when the difference corresponds to the subject part. In the case of a Japanese translation, this corresponds to connection adjustment assuming a case where “even” is “even” but “even” is “even”. At this stage, the result of the syntax analysis as shown in FIG. 10 and the data of the word string as shown in FIG. 12 are obtained, so that it is possible to carry out a certain check and adjustment of the translation data by referring to it.

As a rule for connection adjustment at the time of reflection in the translation, many connection rules 21g as shown in Table 1 are stored in the dictionary unit 21. It is checked whether or not the boundary of the replacement part in the translation matches this, and if it matches, the translation reconstruction means 24c of the translation unit 24 uses the morpheme generation rule 21f to convert the phrase usage form and the like, Is replaced.

  The connection rule 21g in Table 1 includes a left side “definition of morpheme sequence before conversion” and a right side “definition of a morpheme sequence after conversion” with “=” in between. If it matches the pattern on the left side of the connection rule, it converts the phrase usage form to the pattern on the right side. The number “Continuous Form 2” is entered to indicate which type is utilized because the continuous form has two uses. The morpheme generation rule 21f defines how to use it.

  For example, “Because it is important” in connection rule (1), “Because it works and beautiful” in connection rule (2), “Read and take off” in connection rule (3), and “Important” in connection rule (4) In the connection rule (5), the connection is adjusted as “write and run”, and in the connection rule (6), “beautiful”. This connection adjustment will generate a grammatical translation even if the words in the translation example translation can be converted to a non-grammatic translation simply by replacing the translation with the translation corresponding to the difference between the original and the translation example. Can do.

  According to the embodiment of the present invention, the difference between the original text to be translated and the original text of the translation example is associated with the words and phrases in the translation of the translation example that needs to be edited corresponding to the different part. In response to the operation instruction, a portion corresponding to each difference portion is displayed separately from other words and phrases, so that a word or phrase in the translation of the translation example corresponding to each difference portion designated by the user can be shown.

  In other words, the difference between the translation example searched for using the translation target text as a search key and the translation target text reaches multiple places, or which part of the translation example of the translation example presented in a long sentence is the difference part of the translation target text Even if it is difficult to understand whether the part needs to be edited, it is possible to easily show the phrase in the translation of the translation example corresponding to each difference part designated by the user. Therefore, it is possible to reduce time and effort required for the user to edit the translated sentence and complete the translated sentence for the translation target original sentence.

  In addition, it is possible to easily obtain an appropriate translation result for the difference between the original text to be translated and the original text in the translation example, and to easily edit the portion corresponding to the difference in the translation text in the translation example.

The block diagram of the machine translation apparatus concerning embodiment of this invention. The flowchart which shows the processing content of the interactive process with the user in the control part in embodiment of this invention. FIG. 3 is an explanatory diagram illustrating an example of a search result display screen in step S6 of FIG. 2; Explanatory drawing of an example of the display screen of the example for translation in FIG.2 S8. The flowchart which shows the processing content in the translation part after step S6 of FIG. Explanatory drawing of an example of the data structure of the original sentence information which analyzed the structure of the input sentence in the translation part in embodiment of this invention, and the data structure of translated sentence information. The flowchart which shows the flow of a process from the display instruction | indication of a corresponding phrase display in the control part in embodiment of this invention until it reverse-displays the phrase corresponding to the selected phrase. FIG. 8 is an explanatory diagram of an example of an analysis result after the original English sentence using the translation example in step S23 of FIG. FIG. 9 is an explanatory diagram illustrating an example of a conversion result of the original text illustrated in FIG. 8. Explanatory drawing of an example of the result of having provided the translation of the original sentence corresponding to the translation of the sentence using the example for translation with a translation structure. Explanatory drawing of an example of the data structure with respect to the example for translation. FIG. 12 is an explanatory diagram of an example in which corresponding phrase information is added to the data structure for the translation example of FIG. 11. Explanatory drawing which shows an example at the time of selecting the phrase in the original text on the display screen of the example for translation. Explanatory drawing which shows another example at the time of selecting the phrase in the original text on the display screen of the example for translation. The flowchart which shows the flow of the phrase translation process in the phrase translation means of the translation part in embodiment of this invention. Explanatory drawing which shows an example of the display of the phrase translation result in step S33 of FIG. Explanatory drawing which shows another example of the display of the phrase translation result in step S33 of FIG. The flowchart which shows the flow of the reconstruction process of the translation in the translation reconstruction part of the translation part in embodiment of this invention. Explanatory drawing which shows an example of the screen display after reflecting the phrase translation result in step S43 of FIG. Explanatory drawing which shows another example of the screen display after reflecting the phrase translation result in step S43 of FIG. Explanatory drawing which shows an example of the example of a display of the phrase translation result in FIG.15 S33 when the translation word of the example for translation is selected.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 ... Machine translation apparatus, 12 ... CPU, 13 ... ROM, 14 ... RAM, 15 ... Bus, 16 ... Input part, 17 ... Display part, 19 ... Memory | storage device, 20 ... Machine translation program, 21 ... Dictionary part, 21a ... Vocabulary part, 21b ... morphological analysis rule, 21c ... syntax analysis rule, 21d ... conversion rule, 21e ... syntax generation rule, 21f ... morpheme generation rule, 21g ... connection rule, 22 ... example dictionary part, 23 ... control part, 23a ... Corresponding display control means, 24 ... translation section, 24a ... translation example search means, 24b ... phrase translation means, 24c ... translation reconstruction means, 25 ... example processing section, 25a ... common word / phrase correspondence means, 25b ... difference equivalent phrase correspondence Means

Claims (6)

  A memory storing a machine translation program, a dictionary unit storing knowledge and rules necessary for translation, and an example dictionary unit storing a plurality of translation examples composed of parallel translations of first language sentences and second language sentences registered in advance An input unit for inputting an original sentence in a first language to be translated and information necessary for operation; a display part for displaying an original sentence in a first language to be translated and a translated sentence in a second language after translation; In a machine translation apparatus comprising a CPU for calculating and executing the machine translation program, the input translation target original text in the first language is translated using the dictionary section of the storage section and the translation target text is used as a search key. A translation unit that searches a translation example similar to the translation target original from the example dictionary unit of the storage device with respect to the translation target original, and a difference between the source text of the translation example searched by the translation unit and the translation target original An example processing unit for associating a phrase in a translation example of a translation example that needs to be edited with respect to the difference portion and a phrase in a source text to be translated, and according to a user operation instruction, A machine translation apparatus comprising: a control unit configured to distinguish a portion corresponding to each difference from the original text to be translated from other words and displayed on the display unit.   The translation unit extracts the translation example that is the same as the translation source text or a phrase / character using the translation target text as a search key, and outputs the translation of the translation example as a translation for the translation target text The machine translation apparatus according to claim 1, wherein:   The translation unit searches a translation example similar to the translation target original specified as a search key by the user's operation, and selects the translation of the translation example selected by the user's operation from the search result for the translation target original The machine translation apparatus according to claim 1, wherein the machine translation apparatus outputs the translated sentence.   The translation unit obtains a translation result for the phrase when an arbitrary phrase constituting the original text to be translated is selected by the user, and associates the obtained translation result with the phrase in the translation of the translation example. 4. The machine translation apparatus according to claim 3, wherein the translated sentence is reconfigured by replacing the translated phrase.   The translation unit has an attribute of a replaced phrase or a context before and after the replaced phrase from a plurality of translation result candidates for the phrase when an arbitrary phrase constituting the original text to be translated is selected by the user. 5. The machine translation according to claim 4, wherein a non-translated word is selected, and a part of a translated element other than the replaced phrase is changed based on the attribute of the replaced phrase or the context before and after the replaced phrase. apparatus.   A memory storing a machine translation program, a dictionary unit storing knowledge and rules necessary for translation, and an example dictionary unit storing a plurality of translation examples composed of parallel translations of first language sentences and second language sentences registered in advance An input unit for inputting an original sentence in a first language to be translated and information necessary for operation; a display part for displaying an original sentence in a first language to be translated and a translated sentence in a second language after translation; In a machine translation apparatus comprising a CPU for calculating and executing the machine translation program, the translation target original text in the first language input to the computer is translated using the dictionary section of the storage section, and the translation target text is A function for searching a translation example similar to the translation target original from the example dictionary unit of the storage device as a search key with respect to the translation target original, and a difference between the searched translation example original and the translation target original A function for associating a part and a phrase in the translation of the translation example that needs to be edited with respect to the difference part, and a phrase in the translation target original and the translation example original and the translation target according to a user operation instruction A machine translation program for realizing a function of distinguishing a portion corresponding to each difference from the original text from other words and displaying on the display unit.

Images