JP4435144B2 - Data search system and program - Google Patents

Description

  The present invention relates to a data search system and program for searching for information based on a search character string in a natural language.

  In the field of information retrieval services, there is a demand for the construction of an information retrieval system that allows users to perform retrieval processing efficiently. In a general information search system, data including a completely matching character string is output as a search result by matching a keyword input by a user with character information registered in a database to be searched. It has come to be.

  In such an information search system, the user may search for an example in accordance with the context of a specific keyword. However, in such a search process, it takes much time to obtain the search result desired by the user. It may take such a case. That is, in the above-described information retrieval system, the keyword is handled as a simple symbol string, and its contextual role is not considered. For this reason, when enormous search results are obtained, it is necessary for the user to examine the search results one by one.

  For example, it is assumed that an English sentence “The agents seem to be effective in the treatment of the pathogen” is described in a medical document. In the medical field, “treatment” is often translated as “treatment”, but it is unclear whether or not it is appropriate for users to use the term “treatment” for “pathogen”. And In such a case, the user searches medical examples to determine if this expression is appropriate.

  In this case, if the user specifies “pathogen” and “treatment” as keywords and executes a search, if the medical example can be searched to some extent, “treat pathogen” is an expression used in the medical field. I can judge. However, if such an example cannot be retrieved, the correct expression cannot be obtained even though it is assumed that this expression is not used in the medical field. On the other hand, when the user designates “pathogen” as a keyword and executes a search, examples of search results to be obtained must be examined one by one, which is very troublesome.

  The following two search methods are proposed in response to a request to search for a keyword whose syntactic role in a sentence for a specific keyword is known, such as “treatment” for “pathogen” described above. Has been.

  In the first search method, for example, a keyword with a definition specifying a keyword and a part of speech or meaning of a word following the keyword is created, and an example database in which the part of speech or meaning information of the word is assigned in advance is searched. (For example, see Patent Document 1). In such a search method, for example, by specifying {{animal} runs} as a keyword, an example including “XX” can be searched for a search character string “XX runs”. In this search method, it is possible to use a wild card * representing an arbitrary character such as {{animal} * runs}.

In the second search method, a dictionary is created in which nouns that indicate co-occurring verbs and action locations are associated with nouns that are keywords by designating their case relations. Then, morphological analysis and dependency analysis are performed on the search character string to obtain nouns and verbs and their case relations, and from the dictionary, the case relation of “noun-verb” in the search character string is obtained. The associated keywords and their co-occurrence words are detected. In such a search method, for example, it is assumed that a verb “build” and a place “winery” are registered in a keyword “wine” in a dictionary. When the search character string “I want to know where to make wine well” is specified by the user, the “noun-verb” case relationship “wine-make” is acquired. By using this to search the dictionary, it is possible to detect a word representing a place “Winery”. By this method, “XX” can be searched for a search character string “I want to know XX that can make wine well”.
Japanese Patent Laid-Open No. 2-190972 JP 2005-275524 A

  However, in the first search method described above, there are many restrictions due to how the character strings are connected, and there is a problem that it is difficult to cover and narrow down examples. For example, if {{animal} runs} is the keyword, {{animal} runs at a tremendous speed} cannot be searched for an example with an adverb in between. On the other hand, if you use a wildcard such as {{animal} * run}, then an example such as {a hound runs after the {animal} runs away} will hit and the example will work well. Cannot narrow down. In addition, if a keyword is followed by a verb, the user must also specify information such as the verb usage type (five-level usage, etc.) and the usage type (continuous usage type, etc.). The threshold is high. Furthermore, since there are various usage forms of the verbs in the example, it is necessary to prepare keywords that cover variations of the usage forms, which is troublesome.

  On the other hand, in the second search method described above, there is a problem that it is necessary to create a dictionary in which related words and case relationships are associated with keywords in advance, which is troublesome.

  The present invention is a data that can efficiently search a word that does not understand the word itself from an example without knowing the complicated task and knowing the syntactic and semantic role in a sentence for a specific keyword. An object is to provide a search system and a program.

The data search system of the present invention includes a database in which example data including words constituting a sentence and syntax / semantic role information of each word in the sentence is registered, and a search character that receives input of a character string to be searched Specify a search target word for designating a character string specifying the search target word itself as a search target word from the search target character string and the syntax / semantic role information. Search target word designating part for accepting designation of search target words for role information designation for the character string, and character string language analysis for analyzing syntactic and semantic role information of each word constituting the search target character string A search expression generating unit for generating a search expression including a search target word for specifying the character string and role information in syntax and meaning corresponding to the search target word for specifying the role information, and the search expression Using said data A database search unit for searching the example data in the over scan, characterized by comprising an output section for outputting the results for the example data.

  Note that the present invention relating to the apparatus is also realized as a program for causing a computer to realize a function corresponding to the present invention.

  According to the present invention, it is possible to efficiently search a word whose word itself is not understood from an example without knowing complicated work, even if the role of syntax and meaning in a sentence for a specific keyword is known. It becomes possible.

(First embodiment)
FIG. 1 is a block diagram showing the hardware configuration of the data search system according to the first embodiment of the present invention. In the data search system 11 shown in FIG. 1, the calculation control device 12 performs various calculation processes related to data search in the data search system 11. A display device 13, a mouse 14, a keyboard 15, a disk drive 16 and an auxiliary storage device 17 are connected to the arithmetic control device 12. The display device 13 displays the calculation result of the calculation control device 12 and the like. The mouse 14 and the keyboard 15 are used for inputting various commands to the arithmetic control device 12 via the display device 13. The disk drive 16 inputs / outputs a character string to be searched and a file such as a document to be registered in a later-described example database to / from a storage medium. The auxiliary storage device 17 stores a calculation result of the calculation control device 12, an example database, and the like, and includes, for example, a plurality of hard disk drives (HDDs).

  The arithmetic and control unit 12 includes a processor 18 and a memory 19. The memory 19 stores a program 20 related to data retrieval and a work area 21 used when arithmetic processing is executed by the processor 18.

  FIG. 2 is a functional block diagram of the data search system 11 according to the first embodiment. Each functional block in the arithmetic and control unit 12 shown in FIG. 2 corresponds to each program constituting the program 20 described above. That is, when the processor 18 executes each program constituting the program 20, the arithmetic control device 12 functions as each functional block.

  In FIG. 2, a search character string input unit 22 receives an input of a character string to be searched (hereinafter referred to as “search character string”). For example, the search character string is directly input by the user from the keyboard 15 or the like, or is input from the disk drive 16 via the storage medium. The search character string input method is not limited to these, and any input method can be selected.

  The search target word designating unit 23 receives designation of a search target word (hereinafter referred to as “search target word”) in the search character string. For example, the search target word is specified by a drag operation using the mouse 14 or the like. As will be described in detail later, the search target word is used to generate a search expression, and there are two types of search target words depending on the specified contents. One is a search target word for designating a character string for designating the character string itself of the search target word. The other is a search target word for specifying role information for specifying a syntactic and semantic role of the search target word in a sentence.

  The search control unit 24 controls a search character string language analysis unit 25, a search expression generation unit 26, and a database search unit 27, which will be described later. By controlling these, language analysis of the input search character string is performed, a search expression using the specified search target word is generated, and the example data registered in the example database 28 using the search expression Search for.

  The search character string language analysis unit 25 functions as a character string language analysis unit, and performs language analysis of the search character string, thereby acquiring syntactic and semantic role information of each word constituting the search character string. Is. Specifically, by performing morphological analysis, syntactic analysis, dependency analysis, and semantic analysis, which are analysis methods commonly used in natural language processing, the syntactic and semantics of each word constituting the search string Get role information.

  The search formula generation unit 26 generates a search formula using role information on the syntax and meaning of each word constituting the search character string. Specifically, a search expression is generated from a search target word for specifying a character string in the search character string and role information on the syntax and meaning of the search target word for specifying role information.

  The database search unit 27 searches for example data registered in an example database 28 described later. Specifically, the search is performed by comparing the search target word and the role information on the syntax and meaning between the example data registered in the example database 28 and the search formula.

  In the example database 28, a plurality of example data (hereinafter referred to as “sentence data” as appropriate) including words constituting a sentence and syntax / semantic role information of each word in the context are registered. For example, sentence data for the example database 28 is registered via a storage medium inserted in the disk drive 16. Here, in each sentence data, it is assumed that role information on syntax and meaning is associated with each word constituting the sentence in advance.

  The output unit 29 outputs a search result of sentence data registered in the example database 28. For example, the search result is output by displaying on the display device 13. However, the present invention is not limited to this, and the search result may be output by voice instead of the output by the display device 13 or in addition to the output by the display device 13.

  Next, search processing in the data search system 11 according to the first embodiment will be described with reference to FIG. FIG. 3 is a flowchart for explaining search processing in the data search system 11 according to the first embodiment.

  When a search process is performed using the data search system 11 according to the first embodiment, first, a search character string is input from the user, and the search character string input unit 22 receives an input of this search character string. (Step S31). After the search character string is input, the search target word in the search character string is specified from the user this time, and the search target word specifying unit 23 accepts the specification of the search target word (step S32). At this time, the user designates a search target word for designating a character string and a search target word for designating role information. In the search character string, an identifier is assigned to each search target word.

  When the search target word is designated, the search character string language analysis unit 25 performs language analysis of the search character string (step S33). As a result, the syntactic and semantic role information of each word constituting the search character string is acquired, and the syntactic and semantic role information corresponding to each word is given. Then, the search expression generation unit 26 generates a search expression from the search target word for character string specification and the syntactic / semantic role information given to the search target word for role information specification (step S34).

  If the search formula is generated, the database search unit 27 compares the search target word and the role information on the syntax and meaning between the search formula and each sentence data registered in the example database 28. Thus, a search is performed (step S35). If there is sentence data that is determined to be the same according to a certain criterion, the sentence data is extracted from the example database 28 and used as a search result. On the other hand, when there is no sentence data determined to be the same according to the above criteria, the search result indicates that the corresponding sentence data does not exist. When the search by the database search unit 27 is completed, the output unit 29 outputs the search result (step S36). In this way, a series of search processing in the first embodiment is completed.

  Hereinafter, the search process shown in FIG. 3 will be described using a specific example. Here, “the drug seems to be effective for the treatment of the pathogen” is input as a search character string from the user, “pathogen” is specified as a search target word for character string specification, and role information It is assumed that “treatment” is designated as a search target word for designation.

  When the search character string input unit 22 and the search target word specifying unit 23 accept the input of the search character string and the specification of the search target word (steps S31 and S32), the search character string language analyzing unit 25 Then, language analysis of the search character string is performed (step S33). As a final result of such linguistic analysis, the tree structure shown in FIG. 4 is obtained, and the independent words (nodes) constituting the sentence and the syntactic and semantic relationships (links) between the independent words are clarified. In this case, as shown in FIG. 4, “pathogen” is a semantic object of “treatment”, and <pathogen ”is defined as a semantic object of the syntax and semantic role information of“ treatment ”. The word> is acquired.

  Here, the contents of the language analysis in the search character string language analysis unit 25 will be supplemented with reference to FIGS. The search character string language analysis unit 25 first performs morphological analysis of the search character string, subsequently performs syntax analysis or dependency analysis, and finally performs semantic analysis.

  When the morphological analysis is performed on the search character string “the drug seems to be effective for the treatment of the pathogen”, the analysis result shown in FIG. 5 is obtained. In the analysis result shown in FIG. 5, the part of speech, the utilization form, the possible connection form, and the like are associated with each word constituting the search character string. Then, when syntactic analysis or dependency analysis is performed on the analysis result of the morphological analysis shown in FIG. 5, the tree structure shown in FIG. 6 is obtained, and the independent words (nodes) constituting the sentence and the distance between each independent word are obtained. The case relationship (link) becomes clear. Furthermore, when semantic analysis is performed on the tree structure shown in FIG. 6, a tree structure in which the contents of the link are replaced with syntactic and semantic role information is obtained as shown in FIG.

  When the linguistic analysis of such a search character string is completed, the search expression generation unit 26 uses the search target word “pathogen” for specifying the character string specified in step S32 and the syntax of “treatment” acquired in step S33. A search expression is generated from the semantic role information <a word having “pathogen” as a semantic object> (step S34). In this case, the generated search expression is represented by, for example, a tree structure shown in FIG. 7A or a character string shown in FIG.

  When such a search expression is generated, the database search unit 27 searches the sentence data registered in the example database 28 using the search expression (step S35). Specifically, sentence data registered in the example database 28 is searched by matching nodes and links in a tree structure. Note that such a matching technique is an existing technique disclosed in Japanese Patent Laid-Open No. 2005-208825 by the same applicant, and the description thereof is omitted. When the search by the database search unit 27 is completed, the output unit 29 outputs the search result (step S36).

  In the actual example database 28, data is registered on a character basis. Assuming that the above search character string “the drug seems to be effective for the treatment of the pathogen” and the above search formula are registered in the example database 28, these are registered in the following format, for example. .

Search string ： I think (innyo_effective (taisho_treatment (imi_mokuteki_pathogen (shiji_part))) syugo_drug (shiji_part))
Search formula: Treatment (imi_mokuteki_pathogen)
In this example, parentheses indicate a connection between words, and a character string written in Roman letters indicates syntax / semantic role information. These examples and search formulas are judged to have been matched by matching the part of “treatment (imi_mokuteki_pathogen)”.

  For example, it is assumed that the following sentence data is registered in the example database 28. The head number in each sentence data is given for convenience of explanation.

1. To treat YY disease patients, removal of WW pathogens ...
2. The pathogen of YY disease is WW.

3. The handling of WW pathogens must be strictly controlled.

4). UU, a therapeutic agent for YY disease ...
5). To treat VV pathogens ...
6). To disinfect VV pathogens ...
For an example database in which such sentence data is registered,
“Pathogen + <word with“ pathogen ”as a semantic object>”
When a search is performed using the search expression, it is possible to search for an example (sentence data) including a word having “pathogen” as a semantic object as shown below. Here, the sentence data 2 and 4 that are included in the search result only by specifying “pathogen” or “treatment” as a keyword are not included in the search result. In this way, it is understood that the search results are narrowed down by performing the search by specifying the syntactic and semantic role information in the sentence.

1. To treat YY disease patients, removal of WW pathogens ...
3. The handling of WW pathogens must be strictly controlled.

5). When treating VV pathogens ...
6). To disinfect VV pathogens ...
When searching for a verb having “pathogen” as an object, for example, a method of searching for “pathogen” instead of “pathogen” is conceivable. However, this method cannot detect a case where the particle to be connected is not “O” while using “pathogen” as a semantic object, and there is a limit to the example (sentence data) that can be detected. On the other hand, if we try to cover particles that may be connected to other words (such as “no”, “ga”, etc.), many examples that do not have “pathogen” as a semantic object are detected, and the accuracy of the search is reduced. . On the other hand, in this data search system 11, since a word having “pathogen” as a semantic object can be directly searched, it is possible to perform a search more suited to the purpose.

  As described above, in the data search system 11 according to the first embodiment, the search target word for character string designation and the syntactic / semantic role information given to the search target word for role information designation The example data registered in the example database 28 is searched using a search expression including As a result, a word having a specific syntactic / semantic role for a search target word in the search character string can be searched from the example database 28 without fixing the character string.

Further, the example data registered in the example database 28 is registered via, for example, a storage medium inserted in the disk drive 16. This eliminates the need to create a dictionary in which related words and case relationships are associated with certain keywords in advance, as in the conventional search method. Even if the syntactic and semantic roles of a keyword in the sentence are known, it is possible to efficiently search for a word that does not know the word itself from an example.

  Furthermore, the role information on the syntax and semantics is acquired by the language analysis of the search character string by the search character string language analysis unit 25, and a search expression is generated according to the analysis result of the language analysis. This makes it possible to search for an example desired by the user without requiring a special process.

(Second Embodiment)
In the data search system 11 according to the first embodiment, when the corresponding sentence data is not obtained by the search process of the example database 28 using the search expression, the fact is output as a search result. On the other hand, in the data search system 11 according to the second embodiment, if the corresponding sentence data is not obtained by the search processing of the example database 28 using the search formula, a new document to the example database 28 is obtained. This is different from the data search system 11 according to the first embodiment in that the user can select the registration process and search process for documents not registered in the example database 28.

  FIG. 8 is a functional block diagram of the data search system 11 according to the second embodiment of the present invention. In FIG. 8, the same elements as those in FIG. In the data search system 11 according to the second embodiment, as shown in FIG. 8, an element for registering a new document in the example database 28 and a document not registered in the example database 28 are searched. The data retrieval system 11 according to the first embodiment is different from the data retrieval system 11 according to the first embodiment.

  In FIG. 8, a new document input unit 30 receives an input of a document to be newly registered in the example database 28 (hereinafter referred to as “new document”). For example, a new document is input from the disk drive 16 via a storage medium. Note that the input method of the new document is not limited to these, and any input method can be selected.

  The database construction control unit 31 functions as a database registration unit and controls a related document language analysis unit 33 described later. By controlling the related document language analysis unit 33, for example, language analysis of a new document newly registered in the example database 28 is performed.

  The sentence division unit 32 divides the input new document into one sentence unit.

  The related document language analysis unit 33 functions as a new document language analysis unit, and performs a language analysis of the input new document, thereby performing a syntactic and semantic role of each word constituting a sentence included in the new document. Information is acquired. Specifically, by performing morphological analysis, syntax analysis, dependency analysis, and semantic analysis of sentences included in a new document, the role information on the syntax and meaning of each word constituting the sentence included in the new document is obtained. To do. Further, the related document language analysis unit 33 functions as an external document language analysis unit, and performs the language analysis of the primary information described later to thereby construct the syntax and meaning of each word constituting the sentence included in the primary information. The above role information is acquired.

  The external document search control unit 34 controls an external document search unit 35 and a related document language analysis unit 33 described later. By controlling these, a search is performed in accordance with a search target word included in the search expression generated by the search expression generation unit 26 described above, and language analysis of primary information obtained as a result of the search is performed.

  The external document retrieval unit 35 retrieves a document (hereinafter referred to as “external document”) that is not registered in the example database 28 from the Web or another existing database. Specifically, an external document is searched for the Web or another existing database according to the search target word included in the search expression generated by the search expression generation unit 26 described above.

  Next, search processing in the data search system 11 according to the second embodiment will be described with reference to FIG. FIG. 9 is a flowchart for explaining search processing in the data search system 11 according to the second embodiment. The search processing according to the second embodiment determines whether or not the search result of the example database 28 is obtained, and performs related document processing described later when the search result is not obtained. Differs from search processing. In FIG. 9, the same processes as those in FIG.

  As shown in FIG. 9, in the data search system 11 according to the second embodiment, after the database search unit 27 searches the example database 28, the search control unit 24 checks whether the search result is obtained. Determination is made (step S41). If no search result is obtained, a related document process for registering a new document or searching for an external document is performed (step S42).

  FIG. 10 is a flowchart for explaining related document processing of the data search system 11 according to the second embodiment. As shown in FIG. 10, in the related document processing, first, the user is asked whether or not to register a new document (step S50). When a new document registration instruction is received from the user, the new document is input following the registration instruction, and the new document input unit 30 receives the input of the new document (step S51). The new documents here include documents that have a reputation in the field of search character strings, documents that are sufficiently reliable in terms such as books and standards, and in-house documents. Can be considered.

  When the new document is received, the sentence dividing unit 32 divides the new document into one sentence unit (step S52). Then, the related document language analysis unit 33 performs language analysis of a sentence included in the new document divided into one sentence (step S53). As a result, the syntactic and semantic role information of each word constituting each sentence is acquired, and the syntactic and semantic role information is given to each word. When the syntactic / semantic role information is given, the database construction control unit 31 registers the new document in the example database 28 (step S54). As a result, the sentence included in the new document is registered in the example database 28 with syntax and semantic role information added to each word.

  If the new document is registered, the database search unit 27 stores the search formula generated in step S34 shown in FIG. 9 and each sentence data (including sentence data of the new document) registered in the example database 28. The search is performed by comparing the search target word and the role information in the syntax and meaning (step S55). Then, the search control unit 24 determines whether a search result has been obtained (step S56). If a search result is obtained, the process proceeds to step S36 shown in FIG. 9, and the output unit 29 outputs the search result. On the other hand, when the search result cannot be obtained, the process returns to step S50, and the processes after step S50 are performed again.

  On the other hand, in step S50, for example, when a registration instruction for a new document is not accepted for a certain period of time, the user is asked whether or not to search for an external document (step S57). Note that the external document here may include a document that exists on the Internet or the like and that is not sufficiently reliable in terms. When an external document search instruction is received from the user, the external document search unit 35 searches for an external document (step S58). In this case, the external document search unit 35 performs a primary search using only the search target words for character string designation among the search expressions generated in step S34 shown in FIG. 9, and includes those search target words. Collect sentences.

  When the primary search result (primary information) is collected, the related document language analysis unit 33 performs language analysis of the primary information (step S59). As a result, the syntactic and semantic role information of each word constituting each sentence included in the collected primary information is acquired, and the syntactic and semantic role information is assigned to each word. When the syntactic / semantic role information is given, the external document search unit 35 performs a secondary search using the search expression generated in step S34 of FIG. 9 for the sentence included in the primary information. It performs (step S60). If a secondary search is performed, the process proceeds to step 56, and the search control unit 24 determines whether a search result has been obtained. If a search result is obtained, the process proceeds to step S36 shown in FIG. 9, and the output unit 29 outputs the secondary search result. On the other hand, when the search result cannot be obtained, the process returns to step S50, and the processes after step S50 are performed again. In this way, a series of search processing in the second embodiment is completed.

  When a desired search result is obtained by searching an external document, the search result can be sent to the database construction control unit 31 and registered in the example database 28.

  As described above, in the data search system 11 according to the second embodiment, when a desired search result cannot be obtained from the example data registered in the example database 28, a new one is selected according to the user's selection. Since the document can be registered, the example database 28 including syntax and semantic role information can be constructed from any new document.

  If a desired search result cannot be obtained from the example data registered in the example database 28, data including a part of the search target word is arbitrarily searched from an arbitrary external document according to the user's selection. Since secondary search can be performed by assigning syntactic and semantic role information to the search results, not only the example data registered in the example database 28 but also external documents and the like can be searched. As a result, it is possible to perform search processing on an ad hoc basis.

  In the data search system 11 according to the second embodiment, a case where a new document registration process in the example database 28 and an external document search process are executed in combination is shown. Are not necessarily executed in combination, but only one of the processes may be executed.

(Third embodiment)
In the data search system 11 according to the first embodiment and the second embodiment, search processing is performed in accordance with a search character string input from the user. On the other hand, the data search system 11 according to the third embodiment includes a machine translation system, handles the translation result in this machine translation system as a search character string, and uses the information obtained in the translation process for the search process. It differs from the data search system 11 according to the first embodiment and the second embodiment in that it is used.

  FIG. 11 is a functional block diagram of the data search system 11 according to the third embodiment of the present invention. In FIG. 11, the same elements as those in FIG. As shown in FIG. 11, the data search system 11 according to the third embodiment is different from the data search system 11 according to the second embodiment in that it includes an element for performing machine translation.

  In FIG. 11, the first language document input unit 36 functions as an original document input unit, and inputs a document (hereinafter referred to as “first language document”) described in the first language, which is an original document to be translated. Is to accept. For example, the first language document is input from the disk drive 16 via a storage medium.

  The translation processing control unit 37 controls the sentence division unit 32 and a machine translation processing unit 38 to be described later. By controlling these, the first language document is divided into sentence units, language analysis is performed on each sentence of the first language document, and translations in the second language are generated.

  The machine translation processing unit 38 interprets the language of the first language document and generates a translation of the second language from the document of the first language divided into one sentence with reference to the dictionary 39 for machine translation. It is. Each constituent word of the generated translation in the second language is associated with a correspondence relationship between the word in the corresponding sentence in the first language and another translation word candidate. In the machine translation dictionary 39, words in the first language and words in the second language are registered in association with each other.

  Next, search processing in the data search system 11 according to the third embodiment will be described with reference to FIG. FIG. 12 is a flowchart for explaining search processing in the data search system 11 according to the third embodiment. The search process according to the third embodiment is different from the search process according to the second embodiment in that an input is received as a search character string using a translation result obtained by the machine translation system. In FIG. 12, the same processes as those in FIG.

  When a search process is performed using the data search system 11 according to the third embodiment, first, a first language document is input from the user. Therefore, the first language document input unit 36 stores the first language document. An input is accepted (step S61). When the first language document is input, the sentence dividing unit 32 divides the first language document into one sentence unit (step S62).

  The machine translation processing unit 38 performs language analysis on each sentence of the first language document thus divided. As a result, the syntactic and semantic role information of each word constituting each sentence is acquired, and the syntactic and semantic role information is given to each word. Then, the machine translation processing unit 38 generates a translation in the second language while referring to the machine translation dictionary 39 (step S63). At this time, each constituent word of the translated sentence in the second language is associated with a correspondence relationship with the word in the first language document, and a syntactic and semantic relation between each constituent word of the translated sentence is also associated. .

  When the translation of the second language is generated by the machine translation processing unit 38, the search character string input unit 22 receives this as a search character string (step S64). When the search character string is received, the search target word in the search character string is specified by the user, and the search target word specifying unit 23 receives specification of the search target word (step S65). The search target word for specifying the character string and the search target word for specifying the role information are specified as in the first embodiment.

  When a search target word is specified, the search expression generation unit 26, the search target word for character string specification, the syntactic / semantic role information given to the search target word for role information specification, and other translation candidates A search expression is generated from (step S66). These other translated word candidates are considered as a narrowing-down condition in the search.

  If the search expression is generated, the database search unit 27 determines the search target word, syntactic / semantic role information, and other information between the search expression and each sentence data registered in the example database 28. A search is performed by comparing translated word candidates (step S67). If the search by the database search unit 27 is completed, the search control unit 24 determines whether a search result has been obtained (step S41). If no search result is obtained, the related document processing described above is performed (step S42). On the other hand, when a search result is obtained, the output unit 29 outputs the search result (step S68). In this way, a series of search processing in the third embodiment is completed.

  Here, the data search operation shown in FIG. 12 will be described using a specific example. Here, it is assumed that “The agents seem to be effective in the treatment of the pathogen” is input from the user as the first language document. In addition, it is assumed that “pathogen” is designated as a search target word for character string designation and “treatment” is designated as a search target word for role information designation by the user. Here, for convenience of explanation, a case where only one sentence is input as the first language document will be described. For this reason, the process which divides | segments the 1st language document in step S62 is skipped.

  When the input of the first language document is accepted by the first language input unit 36 (step S61), the machine translation processing unit 38 performs language analysis of the first language document. Thereby, the tree structure of the first language shown in FIG. 13 is obtained. Then, after obtaining the tree structure of the first language, the contents of the tree structure of the first language are translated into the second language, and the tree structure of the second language is obtained. Here, the same tree structure as that shown in FIG. 4 is obtained. Note that the machine translation dictionary 39 is referred to during such language conversion. Further, after obtaining the second language tree structure, a translation is generated in the second language (step S63). Here, the translation “The drug seems to be effective for the treatment of the pathogen” is generated.

  It should be noted that the translation sentences and the words in the second language tree structure are associated with other translation word candidates for the corresponding words in the original sentence. For example, “treatment” is associated with other candidate words for “treatment” in the corresponding original text. Here, it is assumed that “other treatment, handling, processing, hospitality, therapeutic agent, processing,...” Is associated as another candidate translation.

  When the translation in the second language is generated, the search character string input unit 22 accepts the translation as a search character string (step S64). Thereafter, the search target word designation unit 23 accepts the designation of the search target word designated by the user as described above (step S65). When the specification of the search target word is accepted, the search expression generation unit 26 generates a search expression from the second language tree structure shown in FIG. 4 (step S66). In this case, the generated search expression includes a search target word for designating a character string, role information on the syntax and meaning of the search target word for designating role information, and other translated word candidates.

  When such a search expression is generated, the database search unit 27 searches the sentence data registered in the example database 28 using the search expression (step S67). The search process by the database search unit 27 is different from the search process in the first embodiment in that the search result is narrowed down by other translation word candidates included in the search expression.

  Here, a case where the sentence data in the example database 28 used in the first embodiment is searched using the search formula generated in the third embodiment will be described. As described above, when a search is performed using the search formula generated in the first embodiment, the following search results are obtained.

1. To treat YY disease patients, removal of WW pathogens ...
3. The handling of WW pathogens must be strictly controlled.

5). When treating VV pathogens ...
6). To disinfect VV pathogens ...
On the other hand, when a search is performed using the search expression generated in the third embodiment, as described above, “removal” and “disinfection” are not included in the other translated word candidates. For this reason, the sentence data 1 and 6 can be excluded from the search results. As a result, only the sentence data 3 and 5 are output as search results. Thereby, it becomes possible to obtain a more accurate search result in line with the user's intention.

  As described above, in the data search system 11 according to the third embodiment, the search target word for designating the character string and the role information in syntax and meaning given to the search target word for specifying the role information. And the database search unit 27 searches the example data registered in the example database 28 using a search expression including the other translation word candidates, and uses the other translation word candidates included in the search expression to obtain the search result. Since the search is narrowed down, a word having a specific syntactic / semantic role for a certain search target word in the search character string can be searched from the example database 28 without fixing the character string. At the same time, it is possible to obtain a search result narrowed down according to the user's intention. As a result, for example, when it is unclear whether the co-occurrence relationship between a specific word and other words in a translation output from a machine translation system is correct, it is possible to determine the suitability from the narrowed-down example data It becomes.

The block block diagram which shows the hardware constitutions of the data search system concerning the 1st Embodiment of this invention. 1 is a functional block diagram of a data search system according to a first embodiment of the present invention. The flowchart for demonstrating the data search operation | movement in the data search system concerning the 1st Embodiment of this invention. Explanatory drawing which shows the content of the tree structure obtained as a result of the linguistic analysis in the data search system concerning the 1st Embodiment of this invention. Explanatory drawing about the result of the morphological analysis in the data search system concerning the 1st Embodiment of this invention. Explanatory drawing which shows the content of the tree structure obtained as a result of the syntax analysis or dependency analysis in the data search system concerning the 1st Embodiment of this invention. Explanatory drawing about the content of the search formula produced | generated by the data search system concerning the 1st Embodiment of this invention. The functional block diagram of the data search system concerning the 2nd Embodiment of this invention. The flowchart for demonstrating the search process in the data search system concerning the 2nd Embodiment of this invention. The flowchart for demonstrating the related document process of the data search system concerning the 2nd Embodiment of this invention. The functional block diagram of the data search system concerning the 3rd Embodiment of this invention. The flowchart for demonstrating the search process in the data search system concerning the 3rd Embodiment of this invention. Explanatory drawing which shows the content of the tree structure obtained as a result of the linguistic analysis of the 1st language document in the data search system concerning the 3rd Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 ... Data search system, 12 ... Arithmetic control device, 13 ... Display device, 14 ... Mouse, 15 ... Keyboard, 16 ... Disk drive, 17 ... Auxiliary storage device, 18 ... Processor, 19 ... Memory, 20 ... Program, 21 ... Work area, 22 ... Search character string input section, 23 ... Search target word designation section, 24 ... Search control section, 25 ... Search string text language analysis section, 26 ... Search formula generation section, 27 ... Database search section, 28 ... Example database 29 ... Output unit 30 ... New document input unit 31 ... Database construction control unit 32 ... Sentence division unit 33 ... Related document language analysis unit 34 ... External document search control unit 35 ... External document search , 36... First language document input unit, 37... Translation processing control unit, 38... Machine translation processing unit, 39.

Claims (5)

A database in which example data including words constituting a sentence and syntax / semantic role information of each word in the sentence is registered, a search character string input unit that accepts input of a character string to be searched, and the search target A search for specifying role information for specifying the search target word for specifying the character string of the search target word as the search target word and the syntax / semantic role information. A search target word designating unit that accepts designation of the target word, a character string language analysis unit that analyzes the role information on the syntax and semantics of each word constituting the search target character string, and the character string designating A search expression generation unit for generating a search expression including a search target word of the user and a role information in syntax and meaning corresponding to the search target word for specifying the role information, and example data in the database using the search expression Search for Data retrieval system characterized by comprising a database search unit, and an output unit for outputting a search result of the example data. A new document input unit that receives input of a new document newly registered in the database, and a new document language analysis unit that analyzes syntax / semantic role information of each word constituting a sentence included in the new document And a database registration unit for registering example data including words constituting a sentence included in the new document and syntax / semantic role information of each word in the sentence in the database. The data search system according to claim 1. An external document search unit that searches for an external document that is not registered in the database, and a word included in a primary search result using the search target word for specifying the character string by the external document search unit An external document language analysis unit for analyzing syntactic / semantic role information, and the external document search unit includes a sentence included in a primary search result to which syntactic / semantic role information is assigned, The data search system according to claim 1 or 2, wherein a search is performed using a search expression. A source document input unit that accepts input of a source document written in a language different from the language in which the example data is described, and a sentence included in the source document document is translated into the language in which the example data is described and the source document A translation processing unit that analyzes the role information on the syntax and semantics of each word constituting the sentence included in the text, and the search character string input unit converts the translation result by the translation processing unit into the character to be searched accepted as a column, the search expression generating unit, the search target word of the character string searched word and the role information retrieval object word corresponding syntactic and semantic roles information for specifying said role information for the specified for the specified The database search unit searches the example data in the database using the search formula and generates other search terms included in the search formula. Data retrieval system according to 1, wherein in any one of claims 1 to 3, characterized in that refine the search results by using a. A program for searching a database in which example data including words constituting a sentence and syntax / semantic role information of each word in the sentence is registered, and accepting input of a character string to be searched to a computer And a role for designating the search target word for specifying the character string specifying the search target word itself as the search target word from the search target character string and the syntax / semantic role information A function for accepting designation of a search target word for information specification, a function for analyzing role information on the syntax and semantics of each word constituting the search target character string, and a search target word for the character string specification and a function of generating a search expression that contains role information syntactic and semantic corresponding to the search target word of the role information for specifying a function for searching for example data in the database by using the search formula Program for realizing a function of outputting the search result of the example data.

Images