JP5106431B2 - Machine translation apparatus, program and method - Google Patents

Description

The present invention relates to a machine translation apparatus , a program, and a method for translating a first language into a second language.

  In a machine translation device that translates a first language into a second language, a translation example in which a sentence in the first language and a sentence in the second language are paired is prepared in advance in a bilingual database, and translated using the translation example. Some have a function. This compares the sentence in the first language with the sentence in the first language in the parallel translation database, and outputs an example sentence in the second language in the parallel translation database when there is a complete match.

  In addition, even when there is no perfect match, in order to output an example sentence that can be reused at high speed when a user generates a translated sentence, word analysis is not performed, and the degree of similarity (similarity) ) Is calculated, and an approximate example sentence that is equal to or higher than the similarity specified by the user is output. In this case, the user sets the similarity high and uses it to output an example sentence that can be reused more. On the other hand, in order to calculate the degree of similarity of how close it is, there is a method of weighting the degree of similarity using a thesaurus (see, for example, Patent Document 1).

JP-A-3-276367

  However, in Japanese Patent Application Laid-Open No. H10-228867, it is impossible to output an example sentence approximated at high speed because word analysis is performed and a thesaurus is used. Further, when the total number of characters and words is small, even if the number of mismatched characters and words is small, the mismatch rate is high, and the similarity is low. For this reason, there are cases where the user can easily use even if the degree of similarity is low, and if the degree of similarity is set high, example sentences in such a case cannot be output.

An object of the present invention is to automatically correct a lower limit value of a designated similarity level internally even when the number of characters and words is small, and it is easy for a user to use even when a designated similarity level is high. A machine translation apparatus , a program, and a method capable of outputting a translation example.

The present invention includes a bilingual database and example sentences in the first language and the example sentence in the second language is stored in advance as a translation example in translation in the storage device, similar original and translation examples of the first language inputted from the input device An input processing unit that reads the degree condition, and a ratio of the number of matching words between the original text in the first language read by the input processing unit and the first language example sentence in the parallel translation database. Similarity calculation means for calculating the similarity with an example sentence, and a correction value for the similarity is defined in advance based on the number of words in the original text of the first language input from the input device, and stored in the storage device A similarity condition correction table, a similarity condition correction unit that corrects a similarity condition based on the number of words of the original text in the first language read by the input processing unit and the content of the similarity condition correction table, and the similarity Condition correction means A bilingual database search means for searching an example sentence in a first language satisfying the corrected similarity degree and a bilingual example sentence in the parallel translation from the parallel translation database; an example sentence in the first language searched by the bilingual database search means; And an output processing unit that outputs an example sentence in the second language of the parallel translation to an output device.

  According to the present invention, even when the total number of characters and words is small, the lower limit value of the specified similarity is automatically corrected internally, and the user can easily use it even when the specified similarity is high. A translation example can be output.

The functional block diagram of the machine translation apparatus which concerns on embodiment of this invention. The block block diagram which shows the hardware constitutions of the machine translation apparatus which concerns on embodiment of this invention. The flowchart which shows operation | movement of the machine translation apparatus which concerns on embodiment of this invention.

  FIG. 1 is a functional block diagram of a machine translation apparatus according to an embodiment of the present invention, and FIG. 2 is a block configuration diagram showing a hardware configuration of the machine translation apparatus according to the embodiment of the present invention. In FIG. 2, the machine translation device 11 is realized by installing a software program such as a machine translation program in a general computer and executing the software program in the processor 13 of the arithmetic control device 12.

  The arithmetic control device 12 performs various arithmetic operations related to machine translation. The arithmetic control device 12 has a processor 13 and a memory 14. A memory translation program 15 is stored in the memory 14, and the processor 13 performs processing. When executed, the work area 16 is used. Calculation results and the like of the calculation control device 12 are displayed and output on the display device 18 that is the output device 17 and also output to the communication network via the communication control device 19. As the display device 18, for example, a CRT display, a plasma display, a liquid crystal display, or the like can be used.

  The input device 20 is used to input information to the arithmetic control device 12, and includes, for example, a mouse 21, a keyboard 22, a disk drive 23, and a communication control device 19. For example, the mouse 21 and keyboard 22 are connected via the display device 18. Then, various commands are input to the arithmetic and control unit 12, and the keyboard 22, the disk drive 23, and the communication control unit 19 input a document to be translated and information necessary for translation.

  That is, the disk drive 23 inputs / outputs a file of a document to be translated to / from a storage medium, and the communication control device 19 connects the machine translation device 11 to a communication network such as the Internet or a LAN. The communication control device 19 is a device such as a LAN card or a modem, and data transmitted / received to / from the communication network via the communication control device 19 is transmitted / received to / from the arithmetic control device 12 as an input signal or an output signal. Further, a hard disk drive (HDD) 24 is provided for storing a calculation dictionary of the calculation control device 12 and a translation dictionary storing knowledge and rules necessary for translation.

  FIG. 1 is a functional block diagram of a machine translation apparatus 11 according to an embodiment of the present invention. Each functional block in the arithmetic and control unit 12 shown in FIG. 1 corresponds to each program constituting the machine translation program 15 described above. That is, when the processor 13 executes each program constituting the machine translation program 15, the arithmetic control device 12 functions as each functional block. Each block of the storage device 25 corresponds to a storage area of the memory 14 and the hard disk drive 24 in the arithmetic control device 12.

  The storage device 25 of the machine translation device 11 includes an input data storage unit 37 that stores input data input from the input device 20 and read by the input processing unit 31, and a variable data storage unit 38 that stores variables used in arithmetic processing. A translation dictionary unit 26 storing various translation knowledge used in translation processing, a bilingual database 27 in which example sentences in the first language and example sentences in the second language are stored in parallel as translation examples, and stored in the translation database 27 A similarity condition correction table 28 in which a correction value for the similarity between the original sentence in the first language and the example sentence in the first language is defined in advance based on the number of words in the example sentence in the first language; A similar example list 29 for storing translation examples (examples in the first language and example sentences in the second language in parallel) searched together with the example sentences in the first language together with the original sentence in the first language;

  The arithmetic control device 12 of the machine translation apparatus 11 includes a control unit 30 that controls the entire apparatus, an input processing unit 31 that performs an input interface with the outside, and an output processing unit 32 that performs an output interface with the outside. A translation unit 33 for translating the input document read via the input processing unit 31, and a first language source text input from the input device 20 and read by the input processing unit 31 and the first language of the bilingual database Similarity calculation means 34 for calculating the ratio of the number of words that match the example sentence as similarity, and the number of words in the original language of the first language read from the input processing unit 31 and read by the input processing unit 31 and the similarity condition correction table 28 The similarity condition correction means 35 for correcting the similarity condition based on the contents of the first sentence, the first language example sentence that satisfies the similarity condition corrected by the similarity condition correction means 35, and And a translation database search means 36 for searching the sentences in the language from the bilingual database 27. Here, the similarity condition is a threshold value for searching for similar example sentences from the parallel translation database 27, and is specified by, for example, a threshold value representing the similarity as a percentage.

  The dictionary unit 26 stores various kinds of knowledge and information necessary for the translation processing in the translation unit 33, such as a vocabulary unit, morpheme analysis rules, syntax / semantic analysis rules, conversion rules, syntax generation rules, morpheme analysis rules, and the like. Consists of. The vocabulary part records the first language words and idioms and the second language translations in association with each other, and also records various information such as part-of-speech information, utilization information, and concept information.

The bilingual database 27 stores in advance a pair of example sentences in the first language and example sentences in the second language as translation examples. Table 1 shows an example.

  As shown in Table 1, the parallel translation database 27 is composed of pairs of example sentences in the first language and example sentences in the second language. For example, the first example sentence in the first language of the parallel translation database 27 is “This is a pen.”, And the example sentence in the second language paired therewith is “This is a pen.” Similarly, the second example sentence in the first language is “Documents should not be removed from the head office premises.” The second example sentence in the second language is “Please take documents outside the head office. It must not be. " In the same manner, the nth example sentence in the first language and the example sentence in the second language are paired.

The similarity condition correction table 28 is a similarity specified by the user from the input device 20 based on attributes such as the number of characters and words of the original text or translation example in the first language, the number of characters between the original text and the translation example, and the difference in the number of words. The data for correcting the value of the degree condition is stored. Table 2 shows an example of the similarity condition correction table 28 when a correction value based on the number of words in the original text in the first language is defined as an attribute.

  As shown in Table 2, the correction value is −20% when the number of original language words in the first language is “4”, and the correction value when the number of original language words in the first language is “5”. Is −15%, the correction value is −10% when the number of original language words in the first language is “6-9”, and the correction value is when the number of original language words in the first language is “10”. Is 0%, and when the number of words in the original text in the first language is “11-15”, the correction value is set to + 10%. In this way, when the total number of words is small, the correction value is negative. When the total number of words is small, even if the number of mismatched characters or words is small, the mismatch rate is high and the similarity is low. Therefore, when the total number of words is small, it is possible to search for example sentences even if the degree of similarity is low. Thereby, even if the similarity is low, it is possible to search for an example that can be easily used by the user.

In Table 2, when the correction value is determined according to the number of words in the original text of the first language, the correction value is determined as a percentage, and the corrected similarity condition is obtained by adding to or subtracting from the similarity condition specified by the user. . On the other hand, as shown in Table 3, the correction value may be determined by a correction coefficient, and the similarity condition corrected by multiplying the similarity condition specified by the user by the correction value may be obtained. Table 3 shows another example of the similarity condition correction table 28 when a correction value based on the number of words in the original text in the first language is defined as an attribute.

  As shown in Table 3, the correction value is 0.8 when the number of original language words in the first language is “4”, and the correction value when the number of original language words in the first language is “5”. Is 0.9, the correction value is 0.95 when the number of original language words in the first language is “6-9”, and the correction value is when the number of original language words in the first language is “10”. If the number of words in the original text in the first language is “11-15”, the correction value is set to 1.1. Thus, the reason why the similarity value is set to be small when the total number of words is small is to make it possible to search for an example that is easy for the user to use even when the total number of words is small. .

Further, as shown in Table 4, the correction value may be determined according to the difference between the number of words in the original text in the first language and the number of words in the first language example in the parallel translation database 28.

  As shown in Table 4, when the difference in the number of words between the first language original sentence and the first language example sentence is “0”, the correction value is 0%, the first language original sentence and the first language example sentence. The correction value is −5% when the difference in the number of words is “1”, and the correction value when the difference in the number of words between the original text in the first language and the example sentence in the first language is “2”. Is −10%, and if the difference in the number of words between the first language original sentence and the first language example sentence is “3”, the correction value is −15%, and the first language original sentence and the first language example sentence When the difference in the number of words is “4”, the correction value is set to −20%.

  In this way, when the difference in the number of words between the original text in the first language and the example sentence in the first language is large, the similarity value is set to a correction value that reduces the similarity due to the difference in the number of words. It is to do. For example, if the original text in the first language is “This is a heavy book.”, The number of words in the sentences to be compared is compared with the first language example sentence “This is a pen.” This is because a difference for one word is always generated because of differences, but in this case, the example sentence “This is a pen.” In the first language of the parallel translation database can also be searched.

  The similar example list 29 stores translation examples (examples in the first language and example sentences in the second language of the parallel translation) searched by the parallel translation database search means 36 together with the original text in the first language. That is, a translation example that matches the similarity condition corrected by the similarity condition correction unit 35 is searched, and the searched translation example is stored together with the original text in the first language. For example, when the first translation example in Table 1 is searched for the original text “This is a book.” In the first language, the example sentence “This is a pen.” In the first language and the second language The original sentence “This is a book” in the first language is also stored in the similar example list 29 together with the example sentence “This is a pen.”

  Next, the input processing unit 31 uses a communication control device 19 such as the Internet, which is the input device 20, a keyboard 22, and the like, from the original text of the first language to be translated and the translation examples stored in the parallel translation database 27. A similarity condition for searching for translation examples and other various commands are received and processed. The input data read by the input processing unit 31, the original text in the first language, and the similarity condition are stored in the input data storage unit 37.

  The output processing unit 32 outputs responses from the control unit 30 to various commands to the control unit 30 to the output device 17. For example, the output processing unit 32 is searched by the translation result in the translation unit 33 or the parallel translation database search means 36. An example sentence in the first language and an example sentence in the second language of the translation are displayed on the display device 18.

  The control unit 30 controls the movement of the entire apparatus. For example, the text data and similarity condition of the original text of the first language to be translated read by the input processing unit 31 are extracted from the input data storage unit 37 and translated. It controls the overall control such as sending to the output unit 32, the similarity calculation means 34, the similarity condition correction means 35, and the translation result in the translation part 33 and the search result in the parallel translation database search means 36. It is.

  The translation unit 33 performs a translation process on the text data in the original text of the first language sent from the control unit 30 while referring to a dictionary or the like stored in the dictionary unit 26.

  The similarity calculation means 34 calculates the ratio of the number of matching words between the original text in the first language read by the input processing unit 31 and the example sentences in the first language in the parallel translation database 27 as the similarity, and the example sentences (similarity similar) 100% example sentence) is returned to the control unit 30 as a translated sentence, or word analysis is performed using the data of the dictionary unit 26 for example sentences that do not completely match, and the ratio of the number of matching words is calculated as the similarity To do.

  The similarity condition correction unit 35 corrects the similarity condition based on the number of words in the original text in the first language read by the input processing unit 31 and the content of the similarity condition correction table 28. Then, the parallel translation database search means 36 searches the parallel translation database 27 for the first language example sentence that satisfies the similarity condition corrected by the similarity condition correction means 35 and the second language example sentence of the parallel translation. The original text in one language is stored in the similar example list 29.

Next, the operation of the machine translation apparatus according to the embodiment of the present invention will be described. FIG. 3 is a flowchart showing the operation of the machine translation apparatus according to the embodiment of the present invention. To make the actual flow easier to understand, the first language is explained as English and the second language is explained as Japanese. The case where the correction values in the similarity condition correction table 28 are those shown in Table 2 will be described.

  First, the user inputs the original text of the first language to be translated and the similarity condition from the input device 20 (S1). As described above, the similarity condition is a threshold for searching for similar example sentences from the bilingual database 27, and is specified by, for example, a threshold representing the similarity as a percentage. Assume that the original text of the input first language is “This is a book.” And the similarity condition is 80%.

  When the original text “This is a book.” And the similarity condition 80% are input from the input device 20, the input processing unit 31 inputs the original text “This is a book.” In the first language. The similarity condition 80% is passed to the control unit 30. In the control unit 30, the original text “This is a book.” And the similarity condition 80% received from the input processing unit 31 are stored in the input data storage unit 37 and stored in the input data storage unit 37. The original text “This is a book.” In the first language and the similarity condition 80% are passed to the similarity calculation means 34 and the similarity condition correction means 35.

  The similarity calculation means 34 receives the input original sentence “This is a book.” And the similarity condition 80% in the first language, and substitutes the similarity condition 80% into the variable sim (S2). The variable sim is stored in the variable data storage unit 38 by the similarity calculation unit 34 via the control unit 30. Also, the similarity calculation means 34 initializes the index variable n of the parallel translation database 27 to 1 (S3), searches the parallel translation database 27 and searches for an example sentence “This is a” in the nth (first) first language. pen. "is acquired (S4). The index variable n is stored in the variable data storage unit 38 by the similarity calculation unit 34 via the control unit 30.

  The similarity calculation means 34 compares two sentences (the original text “This is a book.” In the first language and the example sentence “This is a pen.” In the first language), and calculates the similarity (S5). As a method of calculating similarity, any method such as comparison of the number of characters, comparison of the number of words, comparison of the number of phrases, comparison of parts of speech of words, and the like may be used. In the embodiment of the present invention, the similarity calculation unit 34 calculates the similarity by comparing the number of words. The difference between the original text "This is a book." In the first language and the example sentence "This is a pen." In the first language is only "pen" and "book". The similarity calculation means 34 calculates that the similarity is 75%. Then, the similarity calculation means 34 substitutes the calculated similarity for the variable ruiji (S6). The variable ruiji is stored in the variable data storage unit 38 via the control unit 30 by the similarity calculation unit 34.

  Next, the similarity condition correction means 35 counts the number of words in the original text “This is a book.” In the first language (S7), and substitutes the number of words “4” into the variable word (S8). The variable word is stored in the variable data storage unit 38 via the control unit 30 by the similarity condition correction means 35. The number of words may be counted simply by separating them with spaces, or may be counted in units registered in the dictionary using data in the dictionary unit 26.

  Thereafter, the similarity condition correction means 35 extracts the variable word from the variable data storage unit 38, and from the similarity condition correction table 28 of Table 2, the correction value “−20%” for “4” that is the same value as the variable word. (S9), and the correction value is substituted into the variable hosei (S10). The variable hosei is stored in the variable data storage unit 38 via the control unit 30 by the similarity condition correction means 35. Then, the similarity condition correction unit 35 extracts the variable hosei from the variable data storage unit 38, and corrects the variable sim using (sim = sim + hosei) using the variable hosei (S11). The corrected variable sim is updated and stored in the variable data storage unit 38 via the control unit 30 by the similarity condition correction unit 35.

  That is, the similarity condition correction unit 35 extracts the variable sim and the variable hosei from the variable data storage unit 38. Since the variable hosei has a value of −20% and the variable sim has a value of 80%, the corrected variable sim is 80% −20%, which is 60%, and the similarity condition after correction is 60%. . The corrected variable sim is updated and stored in the variable data storage unit 38 via the control unit 30 by the similarity condition correction unit 35.

  When the similarity condition correction table 28 of Table 3 is used instead of the similarity condition correction table 28 of Table 2, x0.8 is acquired for the variable hosei, and the corrected sim is 80%. × 0.8 is 64%, and the similarity condition after correction is 64%.

  The parallel database search means 36 compares the similarity calculated by the similarity calculation means 34 with the value of the similarity condition (variable sim obtained in step S11) corrected by the similarity condition correction means 35 ( S12) If the similarity calculated by the similarity calculation means 34 is greater than the corrected similarity condition (variable sim obtained in step S11), the nth (first) first language The example sentence and the example sentence in the second language are stored in the similar example sentence list 29 (S13). On the other hand, if the similarity calculated by the similarity calculation means 34 is equal to or less than the corrected similarity condition (variable sim obtained in step S11), the parallel translation database search means 36 is nth (first). ) Are not stored in the similar example sentence list 29. In the case of the embodiment of the present invention, the similarity calculated by the similarity calculation means 34 is 75%, and the corrected similarity condition is 60%, so it is stored in the similar example sentence list 29.

  Thereafter, the bilingual database search means 36 determines whether the nth (first) example sentence in the first language is the last (S14), and if not, returns the variable sim to 80% of the initial value ( S15), the index variable n of the parallel translation database 27 is added (S16), and the process returns to step S4.

  As a result, the similarity calculation unit 34 acquires the second example (n = 2) example sentence “Documents should not be removed from the head office premises.” In the parallel translation database 27, and Comparison is made with an example sentence “This is a book” in one language, and the similarity is calculated (S5). In the first language example “Documents should not be removed from the head office premises.”, None of the ten words match the first language example “This is a book.” Calculated. Then, the similarity calculation means 34 substitutes the calculated similarity 0% for the variable ruiji (S6).

  Next, the similarity condition correction means 35 counts the number of words in the original document “Documents should not be removed from the head office premises” in the first language (S7), and substitutes the number of words “10” into the variable word ( S8). Thereafter, the similarity condition correction means 35 acquires the correction value “0%” for the value “10” of the variable word from the similarity condition correction table 28 of Table 2 (S9), and uses the correction value as the variable hosei. (S10). Then, the similarity condition correcting unit 35 corrects the variable sim by using the variable hosei (sim = sim + hosei) (S11). As a result, the variable sim which is the similarity condition after correction is 80%.

  In this case, the similarity calculated by the similarity calculation unit 34 is 0%, and the value of the similarity condition (variable sim obtained in step S11) corrected by the similarity condition correction unit 35 is 80%. Since the parallel database search means 36 has a similarity calculated by the similarity calculation means 34 smaller than the corrected similarity condition (S12), the nth (second) example sentence in the first language and the second Bilingual example sentences are not stored in the similar example sentence list 29 (S13).

  Then, the processes in steps S4 to S16 are repeated until the nth first language example sentence is the last. As a result, the content of the similar example sentence list 29 finally created is returned to the control unit 30 by the parallel translation database search unit 36, and the control unit 30 outputs the contents of the similar example sentence list 29 (S17). That is, the control unit 30 causes the output processing unit 32 to display and output the contents of the similar example sentence list 29 on the display device 18 of the output device 17. As a result, even when the user designates the similarity condition of 80%, a short example sentence with a one-character difference of 75% can be output as an example for translation.

  Here, when the similarity is calculated by the similarity calculation means 34, if the input original is “This is a heavy book.”, The example sentence “This is a pen.” In the first language of the parallel translation database. Compared on a word basis, the number of words in the sentences to be compared is different, so a difference for one word is always generated. Therefore, when the number of words in the sentences to be compared is different, as shown in the similarity condition correction table 28 in Table 4, the number of words in the first language and the first language example sentence in the parallel translation database 28 A correction value is determined according to the difference in the number of words, and the similarity condition is further corrected by the similarity condition correction table 28.

  The definition content of the similarity condition correction table 28 shown in Tables 1 to 4 of the embodiment of the present invention is an example, and the number of levels and numerical values for level division can be freely set and implemented. Further, although the number of words has been described, it is also possible to obtain a similarity correction value based on other attributes such as the number of characters and the number of phrases. Furthermore, a uniform correction factor may be defined without changing the similarity correction value according to attribute values (number of words, number of characters, number of phrases, etc.) divided into levels.

  In the above description, the first language example sentence is obtained from the parallel translation database 27 in step S4, and then the process proceeds to step S5 to calculate the similarity and correct the similarity. Before the search of 27, the similarity condition is set in advance in combination with only the original attribute (number of words, number of characters, number of phrases, etc.) or the average value of example sentence attributes of the first language stored in the parallel translation database 27. It is also effective to perform a search for an example with respect to the parallel translation database 27 together with the corrected similarity condition. In this case, since the search range can be limited (branch of the search range) using the similarity condition, an effect of improving the search speed can be obtained.

  According to the embodiment of the present invention, the similarity condition correction table 28 in which the correction of the similarity degree is defined, and the contents of the similarity condition correction table 28 are similar based on the original language of the first language or the attribute of the translation example. Since the similarity condition correction means 35 for correcting the degree condition is provided, it is possible to adjust the degree of similarity designated according to the original text of the first language and the situation of the translation example, and whether or not the output of the similar translation example can be determined. Can be controlled. That is, since the correction value is determined as the attribute based on the number of words, characters, and phrases of the original text in the first language, it is possible to search for an example that can be easily used by the user even when the total number of words, characters, and phrases is small. As a result, it is possible to output a short example sentence that does not satisfy the similarity specified by the one-character difference.

  In addition, the attribute of the original text or translation example in the first language is the difference between the number of characters, the number of words, the number of phrases in the original text in the first language and the translation example, the number of characters, the number of words in the original text in the first language and the translation example, Even when there is a difference in the number of phrases, it is possible to output similar example sentences for translation regardless of the difference.

DESCRIPTION OF SYMBOLS 11 ... Machine translation apparatus, 12 ... Operation control apparatus, 13 ... Processor, 14 ... Memory, 15 ... Program, 16 ... Work area, 17 ... Output device, 18 ... Display apparatus, 19 ... Communication control apparatus, 20 ... Input device, DESCRIPTION OF SYMBOLS 21 ... Mouse, 22 ... Keyboard, 23 ... Disk drive, 24 ... Hard disk drive, 25 ... Storage device, 26 ... Dictionary part, 27 ... Bilingual translation database, 28 ... Similarity condition correction table, 29 ... Similarity example list, 30 ... Control , 31 ... input processing unit, 32 ... output processing unit, 33 ... translation unit, 34 ... similarity calculation means, 35 ... similarity condition correction means, 36 ... parallel database search means, 37 ... input data storage unit, 38 ... Variable data storage

Claims (5)

Read into the storage device and the sentence of the first language and the example sentence in the second language and translation database stored in advance as a translation example in translation, the similarity conditions of the original and translation examples of the first language inputted from the input device The ratio of the number of matching words between the input processing unit and the first language source text read by the input processing unit and the first language example sentence in the parallel translation database is similar to the first language source sentence and the first language example sentence. Similarity calculation means for calculating the degree of similarity, and a similarity condition correction in which a correction value of the similarity is defined in advance based on the number of words in the original text of the first language input from the input device and stored in the storage device A similarity condition correction unit that corrects a similarity condition based on the table, the number of words of the original text in the first language read by the input processing unit and the content of the similarity condition correction table, and the similarity condition correction unit Corrected A bilingual database search means for searching an example sentence in a first language satisfying a similarity degree and a second language example sentence of the parallel translation from the parallel translation database, an example sentence in the first language searched by the parallel translation database search means, and machine translation apparatus and an output processing unit that outputs to an output device sentence of the second language. 2. The machine translation apparatus according to claim 1, wherein the number of words in the first language is replaced with the number of characters or phrases in the first language example sentence or the original sentence. 2. The machine translation device according to claim 1, wherein, instead of the number of words in the original text in the first language, a difference in the number of words, a difference in the number of characters, or a difference in the number of phrases between the example sentence in the first language and the original text. .
Storage peripherals to a sentence of the first language and the example sentence in the second language and translation database stored in advance as a translation example in translation, based on the number of words of the first language of the original input from the input device in advance first the computer that closed the language of the original and a similarity condition correction table correction value of the similarity between the sentence in the first language is defined stored in said storage device, the first language inputted from said input device The procedure for reading the similarity condition of the original text and the translation example, and the ratio of the number of matching words between the read original text in the first language and the first language example sentence in the parallel translation database, the original text in the first language and the example text in the first language A procedure for calculating the similarity condition, a procedure for correcting the similarity condition based on the number of words of the original text in the read first language and the content of the similarity condition correction table, and a first condition that satisfies the corrected similarity condition Example sentences in one language Machine translation for executing the procedure for searching the sentences in the second language of the translation from the translation database, and a procedure for outputting to an output device sentence of the second language sentence and its translation of the retrieved first language program. Based on the bilingual database in which the example sentences of the first language and the example sentences of the second language are stored in parallel as translation examples in the storage device, and based on the number of words in the original language of the first language input from the input device A machine translation method executed by a computer having a similarity condition correction table in which a correction value for similarity between the original text and the first language example sentence is defined and stored in the storage device. A step of reading the similarity condition of the original text in the first language and the example for translation, and the ratio of the number of matching words between the read original text in the first language and the first language example sentence in the parallel translation database Calculating the similarity between the first language example sentence and the first language example sentence, correcting the similarity condition based on the number of words read in the first language source text and the content of the similarity condition correction table, A first language example sentence that satisfies the similarity condition and a second language example sentence translated from the first language example sentence, and a search of the first language example sentence and the second language example sentence translated therefrom are output. A machine translation method comprising: outputting to a device.

Images