JP4417967B2 - Example database and example search system - Google Patents

Description

  The present invention relates to an example database storing an example book for translation and an example search system for searching an example database.

  In creating a translated sentence, it is very important in terms of the quality of the translated sentence to use a correct translation word suitable for the field of the translation document. For example, a medical document contains the sentence “The agents seem to be effective in the treatment of the pathogen.” Speaking of “treatment” in the medical field, the word “treatment” comes to mind first, but it is unclear whether it is appropriate to use the term “treatment” for “pathogen”. There are other translations in the English-Japanese dictionary such as "Handling and Processing". If the translation cannot be selected based on the description of the English-Japanese dictionary for each translation, refer to the translation document and other examples in the same field. Decide. Also, if there is a word that is not listed in the dictionary in the original document, the translation is also examined with reference to an example.

  If there is a bilingual document of the source language (English in the example) and the target language (English in the example) for the purpose of investigating and selecting translations that match the context and words that are not in the dictionary, Convenient. That is, for a search result obtained by specifying “pathogen” and “treatment” as search target words, it is possible to know a translation corresponding to “treatment” from Japanese sentences including “pathogen”. .

  Conventionally, in order to facilitate the search, a database called a translation memory, in which these translation original text document and translation text document are registered in a database and referred to at the time of translation, has been utilized (for example, see Patent Document 1). In this translation memory, the user calculates the degree of coincidence with the original translation in addition to actively searching the database by specifying the search target word, and automatically searches the database for parallel translations with a similarity or higher specified by the user. It can also be displayed, and the convenience of search is achieved.

  On the other hand, even if there is no bilingual document, there are many cases where it can be obtained if it is an example book consisting of a single language. Therefore, conventionally, the following two methods have been performed in order to obtain a target translation using a collection of examples written in a single language.

  The first method is a method of searching for an example by designating a word in the same language as the language in which the example is described as a search target word. For example, if you specify “Pathogen” and “Treatment” as the search target words and execute a search, if a medical example is detected to some extent, “Treat pathogen” is an expression used in the medical field. It can be judged.

The second method is a method using a machine translation system (see, for example, Patent Document 2). In this method, an example collection of the target language is analyzed, statistical information and co-occurrence information of each word constituting the sentence is acquired, and using the acquired information, the translation word in the translation dictionary of the machine translation system is acquired. The priority order is changed to preferentially translate frequently used words and words having a co-occurrence relationship in the example.
JP 2003-203072 A JP 2004-86243 A

  However, the actual situation is that bilingual documents used in the translation memory of Patent Document 1 are difficult to obtain.

  In addition, in the first method of obtaining the target translation using a collection of examples written in a single language, if such examples are not detected, there is no indication that this expression is not used in the medical field. Even so, I can't get the right expression. In such a case, it is possible to perform a search by specifying only “pathogen” as the search target word, and examine the obtained examples one by one, but if the detected examples are enormous, it will be very laborious There was a problem that it took. In addition, there is a method of searching by estimating the correct translation, but the labor and effort of trial and error is enormous.

  On the other hand, in the second method of obtaining a target translation using a collection of examples written in a single language, the statistical information and co-occurrence information of the examples are limited to independent words, and independent words The connection between them is generated by a machine translation system. On the other hand, especially when the translated language is a foreign language, there are cases where it is desired to know how to connect independent words rather than independent words themselves. For example, in the original translation of "XX pathogen micrograph", how "XX pathogen" and "micrograph" are connected, ie "XX pathogen's micrograph" "a micrograph of XX pathogen" "XX pathogen .... You may want to know which "in the micrograph" should be adopted. With respect to such a purpose, the document of Patent Document 2 cannot fulfill the purpose because the example data itself is not presented.

  In this way, when creating a translation, two methods have been used to meet the need to obtain the target translation using examples written in a single language. The first method is a method of searching for an example by specifying a word in the same language as the language in which the example is described as a search target word, and is specified when the search target word is too narrowed to detect the example. Although a method of reducing the number of search target words and expanding the search target can be adopted, there is a problem that it is necessary to search for desired information from a lot of unnecessary search results, which is troublesome. In addition, there is a method of searching by estimating the correct translation, but the labor and effort of trial and error is enormous.

  The second method is a method using a machine translation system, which preferentially translates frequently used words and words having co-occurrence relations by using word statistical information and co-occurrence information obtained by analyzing an example book. However, in this second method, statistical information and co-occurrence information are only reflected in independent words, so this is not an effective method when you want to know how to connect independent words. It was necessary to display the example directly for the purpose.

  An object of the present invention is to solve the above-described problem, and to provide an example database and an example search system that can efficiently search an example book consisting of a single language.

An example database construction apparatus according to the present invention includes an external document input unit that reads an external document described in a first language that is a target language of translation, and a grammar of each word that constitutes a sentence by performing language analysis of the external document. Based on the grammatical information obtained by the external document language analysis unit that obtains information and the external document language analysis unit, a word in the first language and a word in the second language that is the source language are stored in pairs. By referring to the bilingual dictionary registered in the apparatus, a word string composed of one or a plurality of words including the same part-of-speech of the second language corresponding to the word constituting the sentence or a derivative of a different part-of-speech is obtained. And a database construction control unit for registering the sentence and the acquired word string as a pair in an example database .

The example search system of the present invention includes a sentence described in a first language that is a target language of translation and a derivative of the same part of speech or a different part of speech of a second language that is a source language corresponding to the word constituting the sentence. A storage device storing an example database in which a word string composed of one or a plurality of words including a word is stored, and a search character string input unit for reading a search target sentence described in a second language When a database search for searching a search target word designating unit for designating the search target sentence search target word selected by the user, a match in the search term designated by the search target word designating unit from the example database And a section.

  According to the present invention, a sentence described in a first language that is a target language for translation and a derivative of the same part of speech or a different part of speech of a second language that is a source language corresponding to the word constituting the sentence When you want to translate a sentence written in the second language (source language) into the first language (target language), because a pair of word strings consisting of one or more words is stored in the example database Even if there is no parallel translation of the second language and the first language, if there is an example of only the first language (target language), it can be referred to during translation.

  Further, even when a word in the example described in the first language corresponding to the search target word specified in the search target sentence is used in a part of speech or a usage form different from the search target word, it is set as a search target. be able to. As a result, the search target can be effectively expanded, and an efficient search can be performed. In addition, it is possible to greatly reduce the burden of estimating the search target word by trial and error performed on the user side.

  FIG. 1 is a block diagram showing the hardware configuration of an example search system according to an embodiment of the present invention. The example search system 11 includes a calculation control device 12 that performs various calculations related to example search, a display device 13 that displays calculation results of the calculation control device 12, and various commands input to the calculation control device 12 via the display device 13. A mouse 14 and a keyboard 15, a search target document, a disk drive 16 for inputting / outputting a file such as a document to be registered in an example database, a calculation result of the calculation control device 12, an example database, and the like. And a hard disk drive (HDD) for storage. The arithmetic and control unit 12 includes a processor 18 and a memory 19. A program 20 relating to an example search is stored in the memory 19, and a work area 21 is used when processing is executed by the processor 18.

  FIG. 2 is an explanatory diagram of the program 20 of the example search system according to the embodiment of the present invention. The program 20 includes a search unit 1 for searching an example database and an external document, a database building unit 2 for building an example database, and documents written in a second language (source language) in a first language (target language). Machine translation system unit 3 for translation into (language). The search unit 1 includes a search character string input unit 22, a search control unit 23, a search sentence language analysis unit 24, a search target word specification unit 25, a search expression generation unit 26, a database search unit 27, and an output unit 29. The database construction unit 2 includes an external document input unit 30, a database construction control unit 31, a sentence division unit 32, an external document language analysis unit 33, an external document search control unit 34, a bilingual dictionary 35, and an external document search unit 36. The machine translation system unit 3 includes a second language document input unit 37, a machine translation processing control unit 38, a translation processing unit 39, and a machine translation dictionary 40.

  FIG. 3 is a block diagram showing the operation of each component of the program of the example search system according to the embodiment of the present invention. The role of each component of the program 20 will be described below.

  Initially, the component of the search part 1 is demonstrated. The search character string input unit 22 reads a sentence to be searched, and the search control unit 23 has a search sentence language analysis unit 24, a search target word specification unit 25, a search expression generation unit 26, and a database search described later. The processing by the unit 27 and the external document search unit 34 is controlled. The search sentence language analysis unit 24 performs linguistic analysis of the search sentence, and acquires grammatical information of each word constituting the sentence. The search target word specifying unit 25 specifies a word to be searched in the read search target sentence. The search sentence search expression generation unit 26 generates a search expression using grammatical information obtained from the analysis result of the search sentence language analysis part. The database search unit 27 searches the example database 28. The output unit 29 outputs search results from the example database 28.

  Here, the example database 28 stores a plurality of pairs of sentences described in the first language and word strings made up of words in the second language corresponding to the independent words constituting the sentence. Yes. That is, one or a plurality of sentences including a sentence described in the first language that is the target language of translation and a derivative of the same part of speech or a different part of speech of the second language that is the source language corresponding to the word constituting the sentence Are stored in pairs. Details of this will be described later.

  Next, components of the database construction unit 2 will be described. The external document input unit 30 reads a document to be registered in the example database 28, and the database construction control unit 31 controls processing in an external document language analysis unit 33 described later. The sentence division unit 32 divides the read external document into sentence units. The external document language analysis unit 33 performs language analysis on the document read from the external document input unit 30 and acquires grammatical information of each word constituting the sentence. The external document search control unit 34 controls processing in an external document search unit and external document language analysis unit 33 described later. The bilingual dictionary 35 is a term database composed of pairs of words in the first language and words in the second language. The external document search unit 36 searches for a document not registered in the example database 28 from the Web or another existing database.

  Next, components of the machine translation system unit 3 will be described. The second language document input unit 37 reads a document described in the second language, and the machine translation processing control unit 38 controls processing in the sentence division unit 32 and the machine translation processing unit 39. . The machine translation processing unit 39 refers to the machine translation dictionary 40 and translates the second language document divided into one sentence into the first language. The machine translation dictionary 40 is a term database composed of pairs of words in the second language and words in the first language.

  Next, the overall operation will be described. First, the function of each component of the search unit 1 will be described. A search sentence described in the second language is input from the search character string input unit 22 of the search unit 1. The input search sentence described in the second language is sent to the search sentence language analysis unit 24 by the search control unit 23, and is subjected to language analysis and given grammatical information. Subsequently, in the search target word specifying unit 25, the user specifies a search target word. As a search target word, a word in an arbitrary first language can be designated in addition to a word in the second language in the search sentence. Subsequently, a search sentence including the designated search target word is sent to the search expression generation unit 26, and a search expression is automatically generated based on the grammatical information obtained by the search sentence language analysis unit 25. The Then, the search target word and the grammatical information are compared between each sentence data registered in the example database 28 and the search formula. Finally, what is determined to be coincident with a certain criterion is output from the output unit 29.

  In the example database 28, a plurality of sentence data described in the first language is registered in advance by the user. On the other hand, from the words in the second language corresponding to the independent words constituting each sentence. The constructed word string is automatically generated and stored in pairs with each sentence. Details of data registration in the example database 28 will be described later.

  Here, when the search formula is compared with the data in the example database 28, one of the feature points is that the second language is used. The user registered in the example database 28 is a sentence described in the first language, and the word string described in the second language used for the search is automatically generated. That is, the feature of the present invention is that a search using a search expression described in the second language is performed on the example database 28 in which sentences described in the first language are registered. .

  Next, the function of each component of the database construction unit 2 will be described. If the desired search result is not obtained by searching the example database 28, new documents written in the first language are registered in the example database 28 by performing the following two processes, and they are stored in the example database 28. You can perform targeted searches. Whether or not to perform these processes is selected by the user.

  The first process is a process performed when the user selects to register a document described in the first language in the example database 28. The following processes are all controlled by the database construction control unit 31. When a document designated by the user is input from the external document input unit 30, the document is first sent to the sentence dividing unit 32 and divided into units of sentences. Subsequently, it is sent to the external document language analysis unit 33 and subjected to language analysis to obtain grammatical information. Based on the grammatical information obtained here, the original form of independent words constituting the sentence is obtained, the bilingual dictionary 35 is referred to, and a word string composed of words of the second language corresponding to these independent words is obtained. Created. The “original form” refers to a form that is not used in the case of a word that is used by a tense such as a verb or an adjective, and a plural form such as a noun is a singular form. The sentence described in the first language and the generated word string in the second language are registered in the database 28 as a pair. After registration in the example database 28, the same processing as the processing by each component of the search unit described above is performed. It should be noted that the process of registering a document in these example databases 28 can be performed in advance before executing a search.

  The second process is a process performed when the user selects to execute a search of an external document by specifying a word described in the first language. In this case, first, the search character string is sent to the external document search unit 36 by the external document search control unit 34, and a search is performed on the Web described in the first language or another existing database. . Subsequently, the primary information obtained as a result of the search is subjected to the same processing as the first processing described above, and the sentence described in the first language and the generated word string in the second language are paired. It is registered in the example database 28. After registration in the example database 28, the processing by each component of the search unit 1 described above is performed.

  Next, the function of each component of the machine translation system unit 3 will be described. In the example search system 11 of the present invention, in addition to specifying a search target word for a search sentence described in the second language, the translation result described in the first language generated by the machine translation system unit 3 is used. The search target word can be specified. That is, by using the translation candidate of the first language of the machine translation dictionary 40, it is possible to easily narrow down the search results, and thereby the second process in the database construction unit 2 described above is performed. It can be done easily.

  When the source document described in the second language, which is the document to be translated, is read into the second language document input unit 37, the source document is sent to the sentence division unit 32 by the machine translation processing control unit 38, and is sent in units of one sentence. It is divided into. The machine translation processing unit 39 performs language analysis on each divided sentence, and generates a translation in the first language while referring to the machine translation dictionary 40.

  When a sentence in the first language, which is the translation result, is input to the search character string input unit 22 together with a sentence in the second language, which is the original sentence, the same processing as described in the description of the function of each component of the search unit 1 In response, a search expression is generated. In this case, the search target word designated by the user can be designated not only for the sentence in the second language as the original sentence but also for the sentence in the first language as the translation result.

  FIG. 4 is a flowchart showing an outline of processing by the constituent elements of the search unit 1 and the database construction unit 2 of the present invention. In FIG. 4, the part of step S <b> 47, which is a separate process, is the processing content by the component of the database construction unit 2, and the other part is the processing content by the component of the search unit 1. First, the processing content in the search unit 1 will be described, and then the processing content in the database construction unit 2 will be described. The processing contents by the components of the machine translation system unit 3 will be described later.

(Processing in the search unit 1)
In FIG. 4, when a search target sentence is input from the search character string input unit 22 (step S41), the search target sentence is sent to the search sentence language analysis unit 24, subjected to language analysis, and given grammar information ( Step S42). Subsequently, the user specifies a search target word in the search target word specifying unit 25 (step S43), and the search formula generation unit 26 automatically generates a search formula from the search target word (step S44). Next, each data registered in the example database 28 is compared with the search formula in the database search unit 27, and sentence data determined to be the same according to a certain criterion is extracted from the example database 28 (step S45). ). Then, it is determined whether or not a search result is obtained (step S46). If the search result is obtained, it is output from the output unit 29 (step S48). If no search result is obtained, the process proceeds to external document processing in the database construction unit 2 which is a separate process (step S47).

  FIG. 5 shows an example in which the first language (target language) is Japanese and the second language (source language) is English, and a search expression is created from the input search string in steps S41 to S44 of FIG. It is explanatory drawing of an example of the processing content until.

  For example, it is assumed that the search character string 51 “The agents seem to be effective for the treatment of the pathogen.” Is input in step S41. When language analysis is performed in step S42, grammatical information 52 such as plural form and tense is obtained for each word constituting the sentence. The search sentence language analysis performed in step S42 includes syntax analysis and dependency analysis, and these are analysis methods generally used in natural language processing of English sentences.

  Next, it is assumed that the user designates “treatment” and “pathogen” as search target words in the search character string 51 in step S43. Thereby, the search object word command character string 53 is obtained. As a method for specifying a search target word, for example, a user interface such as clicking on a word specified as a search target word and highlighting the selected word can be considered. In addition to the words in the search character string 51, any Japanese word that is the first language can be designated as the search target word. As a designation method, for example, a method of opening a search window and inputting an arbitrary character string can be considered.

  In step S44, a search key is generated using the search target word specified by the user. Here, for example, if “agents” is selected as the search target word, the singular “agent” that is the original form of “agents” is used in the search formula. FIG. 5 shows a search formula 54 when “treatment” and “pathogen” are designated as search target words. The search expression 54 is obtained by connecting specified search target words with a delimiter. As a delimiter, a character string that is not used in a general document, for example, “@@@” is used.

  Using the search formula 54 created in this way, the example database 28 is searched in step S45 of FIG. Details of the character string comparison processing in the search will be described after the registration contents of the example database 28 are described in the following description of processing in the database construction unit 2.

(Processing in the database construction unit 2)
FIG. 6 is a flowchart showing the flow of external document processing in the database construction unit 2 shown in step S47 of FIG. If no search result is obtained in step S46 of FIG. 4, the user is asked whether to register the external document in the example database 28 (step S60). A database using an external document is constructed by the following processing.

  If registration of the external document is selected in step S60, the user inputs the external document described in the first language from the external document input unit 30 (step S61). The external documents here may be documents that can be registered in the database without any problems, such as books that are well-established in the field of search target documents, standards documents, and documents that are sufficiently reliable in terms. It is done. The input document is first divided into sentence units by the sentence dividing unit 32 (step S62). The divided external document is subjected to language analysis by the external document language analysis unit 33 for each sentence and given grammatical information (step S63). Subsequently, referring to the obtained grammatical information and the data in the bilingual dictionary 35 (step S64), a word string composed of words in the second language corresponding to the independent words constituting each sentence is automatically generated. (Step S65). Each sentence in the first language and the generated word string in the second language are registered as a pair in the example database 28 (step S66).

  After the registration is completed, the example database 28 is searched using the search formula generated in step S44 of FIG. 4 (step S67). It is determined whether or not a search result is obtained (step S68). If the search result is obtained, the process ends, and the search result is output in step S48 of FIG. If the search result is not obtained in the determination in step S68, the process returns to step S60, and the user is asked again whether to register the external document in the example database 28.

  In step S60, if the user has not selected registration of the external document, the user is asked whether to search for the external document (step S69). The external document here may be a document whose terminology on the Internet or the like is not reliable enough. If the user does not select execution of the search, the process ends.

  If execution of the search is selected in step S69, the external document search unit 36 searches for the external document (step S70). In this case, the external document is searched only when a word in the first language is included in the search target words specified by the user. The external document search is executed as a primary search using only search target words described in the first language.

  Next, the external document language analysis unit 33 performs language analysis on each sentence that is the collected primary search result, and grammatical information is given (step S71). Subsequently, referring to the obtained grammatical information and the data in the bilingual dictionary 35 (step S72), a second language character string composed of words in the second language corresponding to the independent words constituting each sentence is automatically generated. (Step S73). These second language character strings are paired with the sentence described in the first language as the primary search result, and become the target data of the secondary search using the search expression generated in step S44 of FIG.

  Then, a secondary search is performed (S74), and it is determined whether or not a secondary search result is obtained (step S68). If a secondary search result is obtained, the process ends, and the step of FIG. In S48, the secondary search result is output. If the result of the secondary search is not obtained in the determination in step S68, the process returns to step S60 and the user is asked whether to register the external document in the example database 28 again. When a desired result is obtained by searching an external document, the result can be sent to the database construction control unit 31 and registered in the example database 28.

  FIG. 7 is an explanatory diagram showing an example of language analysis processing and second language character string generation processing in external document processing performed in steps S63 to S65 or S71 to S73 of FIG. In order to be consistent with the example given in the description of the search expression 54 described above, in this example, the second language is described as English and the first language is described as Japanese.

  Now, it is assumed that a sentence 71 “Handling of WW pathogens must be strictly managed” is obtained by inputting or searching related documents described in the first language, and is input to the language analyzing unit 33 for external documents. To do. Then, language analysis is performed in step S63 or S71, and grammatical information 72 for the sentence 71 is obtained. That is, for each word constituting the sentence 71, grammatical information such as part-of-speech, inflection form, and possible connection form is obtained. The analysis performed here is morphological analysis, syntax analysis, and dependency analysis, and these are analysis methods generally used in natural language processing of Japanese sentences. “WW pathogen”, “handling”, “severely”, and “managing” shown underlined in the grammar information 72 are independent words.

  Next, in step S64 or S72, the bilingual dictionary 35 of FIG. 3 is referred to for each of these independent words, and all corresponding English words 73 are acquired. The English words corresponding to each Japanese word are connected via a delimiter (here, a semicolon). As for English words 73, not only words with the same part of speech as Japanese but also derived words having different parts of speech are acquired. For example, in addition to the noun “dealing” for “handling”, the verb “deal” is also acquired, and for “severely”, the adjective “close” is also acquired in addition to the adverb “closely”. .

  Next, an English character string (second language character string) 74 registered in the example database 28 based on the acquired English word 73 is created in step S65 or S73. English words corresponding to different Japanese words are connected via a symbol (here, “@@@”) different from the delimiters of English words corresponding to the same Japanese word. The generated second language character string 74 is registered in the example database 28 in S66 of FIG. 6 as search target data 75 paired with the sentence 71 described in Japanese as the first language. Alternatively, it becomes the target of the secondary search in S74 of FIG.

  Here, the character string comparison contents at the time of search will be described using the following example. These are the search formula 54 shown in FIG. 5 and the search target data (database registration contents) 75 shown in FIG. The data after Jpn of the search target data (database registration contents) represents an example for Japanese, and the data after Eng represents the created English character string. The numbers attached to Jpn and Eng are given for convenience of explanation and mean the first search result example. Other examples will be described later.

Search formula:
"Treatment @@@ pathogen"
Search target data (database registration contents):
Jpn1: The handling of WW pathogens should be strictly controlled
Eng1: WW @@ pathogenic organism; pathogen @@@ dealing; deal; treatment; treat; handling; handle @@@ strictly; strict; closely; close; thoroughly; thorough @@@ administer; administration; control; manage; management
The character string comparison target in the search target data 75 is a second language character string written in English, which is the same language as the search formula 54. First, regarding “treatment” as a part of the search expression 54, it is checked whether there is a matching character string in the character strings after Eng1. In this example, “treatment” is detected, so it is considered a match. Next, “pathogen” is examined in the same manner, and in this case, “pathogen” is detected and regarded as a match.

  In this way, for each search target word in the search formula 54, it is checked whether there is a matching word in the second language character string, and the ratio of the matched word to the number of search target words is obtained as the matching rate. It is done. When the obtained matching rate is equal to or greater than a value separately designated by the user, it is determined that the search target data matches the search formula 54 and is output as a search result.

  Here, it is assumed that the retrieval formula 54 is “treatment @@@ pathogen @@@ dealing”. In this case, “dealing” is also present in the second language character string 74, which corresponds to the same Japanese “handling” as “treatment”. In this case, even if both “treatment” and “dealing” match, the number of matched words is not two but one. In this way, even if it is determined that a plurality of words in the second language corresponding to the same first language match, the search accuracy is improved without affecting the matching rate.

  As described above, in the present invention, with respect to an example described in a language (first language) different from the language (second language) describing the search target sentence, the first corresponding to each word constituting the example. A pair of character strings in two languages is registered in the example database 28, and a search expression generated from a search target sentence described in the second language is compared with the second language character string in the example database 28, and the example database 28 searches are performed. As a result, when a sentence described in the second language (source language) is to be translated into the first language (target language), the first language (target) can be obtained even if there is no parallel translation between the second language and the first language. If there are examples of only (language), these examples can be searched using the second language (source language) and referred to during translation.

  Here, in addition to the search formula 54 described above, other database search result examples will be given to specifically describe the contents and effects of the search processing described so far. Note that the detected search target word portion is underlined.

(Example 2)
When dealing with Jpn2: ZZ pathogens, AA regulations must be observed.
Eng2: ZZ @ pathogenic organism; pathogen @@@ dealing; deal; treatment ; treat; handling; handle @@@ ordinance; regulation @@@ observe; observance; comply; compliance
(Example 3)
Jpn3: Micrographs of XX pathogens are treated with YY reagent ...
Eng3: XX disease @@@ pathogenic organism; pathogen ; @ micrograph; microgram @ YY reagent @ handle; handling; manage; magagement; treat; treatment ;…
(Example 4)
Jpn4: Antibiotic YY is used to treat diseases caused by XX disease pathogens ...
Eng4: XX disease @@@ pathogenic organism; pathogen ; disease; sick; ailment @@@ treat; treatment ; cure; remedy @@@ antibiotic…
In the second example (Jpn2, Eng2), the search target word “treatment” is a noun, while the Japanese sentence uses the verb “handle”. Japanese sentences and English sentences do not always use the same syntax, and English sentences often use nouns, and Japanese sentences often use verbs. Thus, even when a word in a Japanese sentence corresponding to a search target word is used with a part of speech different from that of the search target word, the example search system according to the present invention can detect it.

  This is because, using the language analysis function of the present invention and the bilingual dictionary, derived words having different parts of speech from the English words corresponding to the search target words are added to the English word string used for the search. Accordingly, the search target can be effectively expanded without reducing the search target words, and an efficient search can be performed. In addition, it is possible to greatly reduce the burden of estimating the search target word by trial and error performed on the user side. From this search result, it can be seen that the word “handle” is used as an action against a pathogen.

  In the third example (Jpn3, Eng3), it is understood that “treat” is used for an action performed using a reagent even for an action against a pathogen.

  In the fourth example (Jpn4, Eng4), “treatment” is searched, and it can be seen here that it is used as an action for “disease”.

  The contents of the search described so far will be specifically described using the above example. First, when translating the original medical field written in English into Japanese, “The agents seem to be effective for the treatment of the pathogen.”, I would like to know the translation of “treatment”. In this case, if you search the example database of Japanese sentences by specifying “treatment” and “pathogen” as the search target words, the first example “Handling of WW pathogens should be strictly controlled” is found. Detected. In the medical field, the word “treatment” comes first to the word “treatment”. However, in Example 4, “treatment” is used as an action for a disease, and there is no example that targets “pathogen” (pathogen) itself. Furthermore, considering that there are examples (Examples 2 and 3) that use "handling" and "handling" as words representing actions against pathogens, and the meaning of the original text, the translation of "treatment" “Handling” is considered appropriate. In this way, when you want to know the appropriate Japanese translation of “treatment” in “treatment of the pathogen”, you can specify English words as the search target words, even if there is no Japanese-English translation example. You can search for examples.

  In summary, the effects obtained from the constituent elements of the search unit and the database construction unit are as follows.

(1) When it is desired to translate a sentence described in the second language (source language) into the first language (target language), the first language (target language) can be used even if there is no parallel translation between the second language and the first language. If there are only examples), these examples can be searched using the second language (source language) and referred to during translation. (2) Even if a word in the example described in the first language corresponding to the search target word specified in the search target sentence is used in a part of speech or usage different from the search target word It can be. As a result, the search target can be effectively expanded, and an efficient search can be performed. In addition, it is possible to greatly reduce the burden of estimating the search target word by trial and error performed on the user side.

(Processing in machine translation system part 3)
FIG. 8 is a flowchart showing the processing contents by the components of the machine translation system unit 3 of the present invention.

  When the original document document described in the second language is read into the second language document input unit 37 (step S81), the original document is sent to the sentence division unit 32 by the machine translation processing control unit 38 and is sent in units of one sentence. Divided (step S82). The machine translation processing unit 39 performs language analysis on each divided sentence, and a translation in the first language is generated while referring to the machine translation dictionary 40 (step S83). Each word constituting the generated translation in the first language is associated with the original language in the original text in the second language, and other translation word candidates are also associated with the original language in the original text. Yes.

  Subsequently, when the translation result by the machine translation processing unit 39 is input to the search control unit 23 (step S84), the user uses the search target word designating unit 25 to search for words in the original text described in the second language. In addition, the search target word is also specified from the words in the translation written in the first language (step S85).

  Thus, when specifying a search target word described in the first language, a translation word can be selected using a translation, and the search target word described in the first language can be easily specified. Furthermore, since each translation word in the translation is associated with another translation word candidate, it is possible to specify a search target word from translation word candidates other than the word used in the translation. When a search target word is designated from these translation word candidates, a plurality of translation word candidates corresponding to the same second language word can be designated, and an example including one of the words can be retrieved. This is because the translation of the first language is associated with each word in the original text described in the second language, and can be clearly distinguished from translations associated with other words in the original text. . In this way, the machine translation system unit 3 can efficiently narrow down the search or expand the search target.

  When the search target word is specified by the search target word specifying unit 25 of the search unit 1, the search formula generating unit 26 generates a search formula (step S86). Subsequently, the database search unit 27 compares each sentence data registered in the example database 28 with the search target word, syntax / grammatical information, and other translated word candidates of the search formula 54, according to a certain criterion. Those determined to be the same are extracted as data matching the search expression 54 (step S87).

  Then, it is determined whether or not a search result is obtained (step S88), and when the search result is obtained, it is output from the output unit 29 (step S90). On the other hand, when the search result is not obtained, the process proceeds to external document processing which is another processing (step S89). The content of the separate processing performed in step S89 is basically the same as the content described with reference to FIG.

  The above processing contents will be described using a specific example. For example, for the English expression “treatment of the pathogen”, the Japanese translation generated by machine translation is “pathogen treatment”. Also, suppose that another English word candidate “pathogen” is listed for the English word “pathogen”.

  Here, when “treatment of the pathogen” is searched using the constituent elements of the search unit 1, “treatment” and “pathogen” are designated as search target words, or the first of “pathogen” is specified. When a translation by language is used as a search target, it is necessary to search and specify the translation separately. On the other hand, when using the components of the machine translation system unit 3, it is not necessary to separately check the Japanese word corresponding to “pathogen”, and only the word used in the translation is selected. Can be easily specified. Further, in addition to “pathogen”, another translation word candidate “pathogen” can be added to the search target word and the following search can be performed.

Search target words:
"Treatment""pathogen""pathogen""pathogen"
Search process contents:
Search for examples containing "treatment" and "pathogen" and "pathogen", or search for examples containing "treatment" and "pathogen" and "pathogens" Here, if "treatment" is added to the search terms, Search contents are as follows.

Search target words:
"Treatment""pathogen""pathogen""pathogen""treatment"
Search process contents:
Search for examples that include "treatment" and "pathogen" and "pathogen" and "treatment", or search for examples that include "treatment" and "pathogen" and "pathogens" and "treatment". The case where "pathogen" and "pathogen" are associated with the same original word "pathogen", and "treatment" is "pathogen""Is associated with a different language, and the associated original language can be clearly distinguished.

  As described above, when the components of the machine translation system unit 3 of the present invention are used, the following effects can be obtained. That is, when a translation sentence described in the first language generated by the machine translation system unit 3 is used for a search target sentence described in the second language, the translation word described in the first language is used as the search target word. Can be easily added. In addition, because each word in the search sentence is associated with a translation in the first language, if another translation word candidate is specified for a specific translation, an example including either word is searched. You can specify.

  As described above, in the example search system according to the present invention, when a sentence described in the second language (source language) is to be translated into the first language (target language), there is no parallel translation between the sentences in the second language and the first language. However, if there are examples of only the first language (target language), these examples can be searched using the second language (source language) and referred to during translation. In addition, even when a word in the example described in the first language corresponding to the search target word specified in the search target sentence is used in a part of speech different from the search target word, it can be a search target. . Accordingly, the search target can be effectively expanded without reducing the search target words, and an efficient search can be performed. In addition, it is possible to greatly reduce the burden of estimating the search target word by trial and error performed on the user side.

  Furthermore, when a translation sentence described in the first language generated by the machine translation system unit 3 is used for a search object sentence described in the second language, the translation word described in the first language is used as the search object word. Can be easily added. In addition, because each word in the search sentence is associated with a translation in the first language, if another translation word candidate is specified for a specific translation, an example including either word is searched. You can specify.

The block block diagram which shows the hardware constitutions of the example search system concerning embodiment of this invention. Explanatory drawing of the program of the example search system concerning embodiment of this invention. The block block diagram which showed the function of each program component of the example search system concerning embodiment of this invention. The flowchart showing the outline | summary of the processing content by the search part and database construction part in this invention. It is explanatory drawing of an example of the processing content from the search character string input to creating a search formula in step S41 to step S44 of FIG. 5 is a flowchart showing a flow of external document processing in the database construction unit shown in step S47 of FIG. Explanatory drawing which shows an example of the external document process content performed by step S63-S65 or S71-S73 of FIG. The flowchart showing the processing content by the component of the machine translation system part 3 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Search part, 2 ... Database construction part, 3 ... Machine translation system part, 11 ... Example search system, 12 ... Calculation control apparatus, 13 ... Display apparatus, 14 ... Mouse, 15 ... Keyboard, 16 ... Disk drive, 17 ... Hard disk drive, 18 ... processor, 19 ... memory, 20 ... program, 21 ... work area, 22 ... search character string input unit, 23 ... search control unit, 24 ... language analysis unit for search sentence, 25 ... search target word designation unit , 26 ... Search formula generation unit, 27 ... Database search unit, 28 ... Example database, 29 ... Output unit, 30 ... External document input unit, 31 ... Database construction control unit, 32 ... Sentence division unit, 33 ... Language for external document Analysis unit 34 ... External document search control unit 35 ... Bilingual dictionary 36 ... External document search unit 37 ... Second language document input unit 38 ... Machine translation processing control unit 39 Translation processing unit 40 ... Machine translation dictionary 51 ... Search character string 52 ... Grammar information 53 ... Search target word command string 54 ... Search formula 71 ... Sentence 72 ... Grammar information 73 ... English word 74: Second language character string, 75: Search target data

Claims (5)

An external document input unit for reading an external document described in a first language as a target language of translation ;
A language analysis unit for external documents that performs linguistic analysis of the external document and obtains grammatical information of each word constituting the sentence;
Based on the grammatical information obtained by the language analysis unit for external documents, the bilingual dictionary registered in the storage device is paired with the first language word and the second language word as the source language. , Obtaining a word string composed of one or a plurality of words including the same part of speech or a derivative of a different part of speech corresponding to the word constituting the sentence, and the sentence and the acquired word string; An example database construction device comprising a database construction control unit for registering a pair in the example database . One or more words including a sentence described in the first language that is the target language of translation and a derivative of the same part of speech or a different part of speech of the second language that is the source language corresponding to the word constituting the sentence A storage device storing an example database stored in pairs with a word string composed of:
A search string input unit for reading a search target sentence described in a second language;
A search target word designating unit for designating a search target word in the search target sentence selected by the user ;
Examples retrieval system characterized by comprising a database search unit that searches for a match in the search term designated by the search target word designating unit from the example database. Input wherein a document described in a second language comprising a machine translation system for translating a first language, said machine translation system associated with the candidate word other than the word used in the translation and the translated sentence by is, example search system of claim 2, wherein the language or translation candidates of Oite the transliteration during translation can be specified search target word on the search target word specifying unit. An example database construction device comprising a storage device storing a bilingual dictionary and an example database, an input device for inputting an external document described in a first language as a target language for translation, and a microprocessor for calculating and executing a program In a computer to function,
A procedure for reading into the computer an external document described in a first language as a target language for translation; a procedure for performing linguistic analysis of the external document to obtain grammar information of each word constituting a sentence;
Based on the obtained grammatical information, a procedure for referring to a bilingual dictionary registered in a storage device in which a word in a first language and a word in a second language as a source language are paired;
Obtaining a word string composed of one or more words including the same part of speech of the second language corresponding to the words constituting the sentence or a derivative of a different part of speech;
A program for executing a procedure for registering the sentence and the acquired word string as a pair in an example database. A storage device that stores a bilingual dictionary; an input device that inputs an external document described in a first language that is a target language for translation; and a microprocessor that executes a program. A sentence described in a first language that is a language, and one or a plurality of words including the same part of speech or a derivative of a different part of speech of a second language that is a source language corresponding to the word constituting the sentence In a computer for storing an example database stored in a pair with a word string to function as an example database construction device,
A procedure for reading into the computer a search target sentence written in a second language;
A procedure for specifying a search target word in the search target sentence selected by the user;
A program for executing a procedure for searching an example database that matches a specified search target word.

Images