JP2005063185A - Document retrieval system, question-answering system, and document retrieval method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a highly precise document retrieval system capable of providing a retrieval result desired by a user. <P>SOLUTION: An information type desired by the user is detected from a retrieval request sentence inputted by the user. The retrieval request sentence is morphologically analyzed to generate a retrieval condition. A retrieval term appropriate for the detected information type is added to the generated retrieval condition to correct the retrieval condition. A retrieval is performed by use of the corrected retrieval condition. The retrieval result is presented to the user. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a document search system that presents a document by performing a database search based on a search request input by a user, and a document search system that presents an answer by performing a database search based on a question input by a user.

When searching for information from an enormous amount of data that has been digitized based on a search request entered by a user, it is common practice to generate a search condition that uses a word extracted from the search request sentence as a search term. ing. In order to improve search accuracy, the search request is analyzed using a thesaurus, synonym dictionary, different notation dictionary, or co-occurrence word dictionary, etc., and related terms of the search term (for example, synonyms, synonyms, co-occurrence words, etc.) ) Has been proposed (for example, Patent Document 1 and Patent Document 2).
Japanese Patent No. 2970443 JP 2002-269133 A

  In the conventional search condition expansion process, a search term is added simply by being a related term without considering the type of information that the user is seeking. In this case, it is not always possible to add a search term effective for searching for information required by the user. Therefore, in the conventional method, the accuracy of the search result is deteriorated and the user's request may not be sufficiently satisfied.

  Accordingly, an object of the present invention is to provide a system capable of searching for a document with higher accuracy than before by performing search condition expansion processing in consideration of the type of information requested by a user.

In order to solve the above problems, a document search system of the present invention is a search system for searching a document database storing document data based on a search request sentence input by a user, and has the following configuration. .
(1) An input unit for inputting a user search request text.
(2) An information type determination unit that determines an information type requested by a user from the search request text. (3) A search condition generation unit that extracts a search term from the search request sentence and generates a search condition using the search term and a word or phrase corresponding to the information type.
(4) A search unit that searches the document database using the search condition.
(5) An output unit for presenting the search result to the user.

The question answering system of the present invention is a question answering system that searches a knowledge database storing knowledge data based on a question sentence input by a user and obtains an answer to the question sentence. Is provided.
(1) An input unit for inputting a user's question text.
(2) An information type determination unit that determines an information type requested by the user from the question text.
(3) A search condition generation unit that extracts a search term from the question and generates a search condition using the search term and a phrase according to the information type.
(4) A search unit that searches the knowledge database using the search condition.
(5) An answer generator that generates an answer to the question sentence based on the search result.
(6) An output unit for presenting the generated answer to the user.

  According to the present invention, the search condition expansion process is performed in consideration of the type of information requested by the user, so that a document can be searched with higher accuracy than before.

  Further, according to the present invention, the search range is limited in consideration of the type of information requested by the user, so that documents can be searched more efficiently than before.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  Each embodiment of the present invention described below is realized as a program that is operated by a computer such as a workstation (WS) or a personal computer (PC). FIG. 16 is a diagram of a computer used in each embodiment.

  The computer includes a central processing unit 1601 that executes a program, a memory 1602 that stores the program and data being processed by the program, and a magnetic disk drive 1603 that stores the program, data to be searched, and an OS (Operating System). And an optical disc drive 1604 for reading and writing programs and data on the optical disc.

  Furthermore, an image output unit 1605 that is an interface for displaying a screen on a display, an input receiving unit 1606 that receives input from a keyboard, a mouse, a touch panel, and the like, and an input / output interface (for example, USB (Universal Serial) Bus), audio output terminal, etc.).

  Further, the display device 1608 includes a display device 1608 such as an LCD, CRT, and projector, an input device 1609 such as a keyboard and a mouse, and an external device 1610 such as a memory card reader / speaker.

The central processing unit 1601 executes the program after reading the program from the magnetic disk drive 1603 and storing it in the memory 1602. During the program execution, a part or all of the search target data may be read from the magnetic disk drive 1603 and stored in the memory 1602. The basic operation is that a search request from the user is input via the input device 1609. In response to the search request, the search target data stored in the magnetic disk drive 1603 or the memory 1602 is searched. Then, the search result is displayed on the display device 1608.

  The search result may be displayed not only on the display device 1608 but also presented to the user by voice by connecting a speaker as the external device 1610, for example. Alternatively, a printer may be connected as the external device 1610 and presented as a printed matter.

  (First Embodiment) FIG. 1 is a diagram for explaining a schematic configuration of a document search system according to a first embodiment of the present invention. In this embodiment, a database is searched for a document that the user seems to be seeking, and a list of documents is presented to the user. The document search system of this embodiment is realized as a program that is operated by a computer.

The document search system according to the present embodiment includes an input unit 101 through which a user inputs a search request, an information type determination unit 102 that determines the type of information that the user is seeking from the search request, and a search condition from the search request. A search unit 103 that generates and searches, and an output unit 104 that outputs a search result are provided.

  The search request input to the input unit 101 is output to the information type determination unit 102. The information type determination unit 102 performs morpheme analysis using the morpheme analysis dictionary 113 and estimates the information type requested by the user using the information type determination knowledge 111.

  The search unit 103 generates a search condition using the search condition expansion dictionary 114 and the estimated information type, and searches information stored in the document database 112. The output unit 104 displays the search result on the screen and outputs it to the user.

  (Information type determination unit 102) Information type determination refers to what kind of information the user is requesting from the search request text (search request text) input by the user during document search. Is a technique for determining

  FIG. 2 is a diagram for explaining the flow of processing in the information type determination unit 102 of this embodiment.

  (S201) The search request text input by the user is received from the input unit 101. The search request text is, for example, “Where was President XX born?”.

  (S202) Morphological analysis is performed on the search request sentence using the morphological analysis dictionary 113.

  (S203) With reference to the information type determination knowledge 111, the information type requested by the user is determined from the search request sentence.

  The information type determination knowledge 111 of this embodiment is data in which “expression pattern” and “information type” are associated with each other. In the present embodiment, it is assumed that the “expression pattern” is described by a regular expression. Therefore, the “expression pattern” may be only a word or phrase, or may be information including a position (a sentence head, a sentence end, etc.) in a search request sentence.

  In the information type determination knowledge 111, for example, an information type “PLACE (location)” is associated with an expression pattern “ending in“ somewhere ”or“ where is? ””. An information type “DATE (date and time)” is associated with an expression pattern “ending in“ someday ”or“ someday ””. The expression pattern “PERSON (person name)” is associated with the expression pattern “ending with“ someone? ”.

  In the information type determination S203, the “expression pattern” registered in the information type determination knowledge 111 is detected from the search request text by pattern matching of a regular expression. The “information type” associated with the detected “expression pattern” is set as the determination result.

  Information type determination knowledge 111 is prepared in advance.

  (S204) The morphological analysis result and the information type determination result are output to the search unit 103.

  (Search Unit 103) FIG. 3 is a diagram for explaining the flow of processing of the search unit 103 in this embodiment.

  (S301) The search unit 103 according to the present embodiment receives the morphological analysis result of the search request statement and the information type determination result from the information type determination unit 102.

  (S302) A search condition is created from the morphological analysis result. As a result of the morphological analysis performed by the information type determination unit 102, the search request sentence can be decomposed for each word. Therefore, a search condition is generated by extracting a phrase as a search term from a search request sentence based on a morphological analysis result. For example, from the search request sentence “Where President XX was born”, the phrases “XX”, “President” and “Born” were extracted as search terms to generate search conditions.

  (S303) With reference to the search condition expansion dictionary 114, the search condition is expanded based on the information type.

  When expanding the search condition, it is conceivable to add a phrase related to the search term (for example, a related term including synonyms and synonyms). However, whether a related word is appropriate for a search should vary depending on the type of information that the user is seeking.

  For example, there is a search term “born”. Related words that can be inferred from this phrase include “birthday”, “birthplace”, “fabric”, “date of birth”, and so on. If the information that the user wants to know is “place of birth”, it would be appropriate to add the phrases “place of birth” and “fabric”. In this case, it is not appropriate to add the phrases “birthday” and “birth date”, and conversely, there is a possibility of adding noise to the search result.

  However, it is often difficult to add an appropriate word even if focusing only on the search condition.

  For example, if a search condition is generated from the search request sentence “when is President XX born” in S302, a search condition with search terms “XX”, “President” and “born” may be generated. There is. This is because the interrogative words such as “where” and “when” are often not effective in the search and are not usually included in the search term.

  Therefore, in this embodiment, an appropriate search term is added based on the information type with reference to the search condition expansion dictionary 114.

  FIG. 4 is a diagram for explaining an example of the search condition expansion dictionary 114 of the present embodiment. The search condition expansion dictionary 114 is data in which the original word 401, the information type 402, and the related word 403 are associated with each other.

  For example, for the original word “born”, “dough” and “birthplace” are associated as related words whose information type is PLACE (location), and “related information” whose information type is DATE (date and time) is “ “Birthday” and “Birth date” are associated.

Therefore, referring to the search condition expansion dictionary 114 and expanding the above-described search conditions “XX”, “President”, and “Born” based on the information type are as follows.
1. For search request texts with the information type PLACE (place), the words “fabric” and “birthplace” are added, and the search conditions are “XX”, “President”, “Born”, “Dough”, “Born” Earth ".
2. For search request sentences with the information type DATE (date and time), the words “birthday” and “birth date” are added, and the search conditions are “XX”, “President”, “Born”, “Birthday” "Birth date".

  (S304) The document database 112 is searched using the extended search condition.

  (S305) The search result is output to the output unit 104.

  (Effects of the present embodiment) As described above, in the present embodiment, the search condition is set by using a related term suitable for the “type of information requested by the user” determined from the expression pattern included in the search request sentence. The search is performed after expanding. Therefore, “information not requested by the user” can be reduced from the search result, and the search can be performed with high accuracy.

  (Modification) FIGS. 5A and 5B are examples of display screens when the information type determination result and the search condition expansion result are presented to the user in this embodiment.

  FIG. 5A shows an example of a case where the user inputs a search request sentence “Where is XX President born?”.

  The information type analysis result “PLACE (location)” and the search conditions “Ox”, “President” and “born” obtained from the search request text are displayed. Further, an expanded search term (extended term) “fabric”, “birthplace”, “birthday”, “birth date” obtained using the search term “born” and the search condition expansion dictionary 114 is also displayed. Yes.

  Now, since the information type is “PLACE (place)”, it is proposed to the user using the check boxes 501 to 504 to adopt “fabric” and “birthplace” as the expansion term.

  FIG. 5B shows an example of a case where the user inputs a search request sentence “when the president was born”. In this example, the information type determination result is “DATE (date)”. And it proposes to the user using check boxes 505 to 508 to adopt “birthday” and “birth date” as the expansion term.

  As described above, the expansion plan of the search condition according to the information type may be presented to the user, and the user may determine the search term to be finally used by operating the check box.

  (Additional note) In the first embodiment, the information type requested by the user is determined from the search request sentence, the search condition is generated from the search request sentence, and the search condition is extended using the search condition expansion dictionary. did. That is, it has been described that the search condition is once generated from the search request sentence and then the search condition is extended.

  However, the information type requested by the user may be determined from the search request text, and the search condition may be generated using the search request text, the search condition expansion dictionary, and the information type. That is, when a search condition is generated, a search condition including an appropriate search term corresponding to the information type may be generated by referring to the search condition expansion dictionary.

  For example, for each word extracted as a result of morphological analysis for the search request sentence, a related word suitable for the information type is obtained from the search condition expansion dictionary. Then, a search condition may be generated using the extracted word and the obtained related word.

  (Second Embodiment) A question answering system according to a second embodiment of the present invention will be described below. In the present embodiment, a database is searched for an answer to a user's question.

  In the document search system according to the first embodiment, a search request sentence of a user is appropriately expanded according to the information type requested by the user, a database search is performed, and a list of documents that the user thinks is necessary. Was presenting.

  On the other hand, in the question answering system according to the present embodiment, the user's question is appropriately expanded according to the information type requested by the user, the database search is performed, and the answer that the user thinks necessary is presented. To do.

  FIG. 6 is a diagram illustrating a schematic configuration of the question answering system according to the present embodiment. The question answering system according to the present embodiment includes an input unit 101 through which a user inputs a question sentence, an information type determination unit 102 that determines a type of information requested by the user from the question sentence, and a search condition from the question sentence. A search unit 103 that generates and searches, and an output unit 104 that outputs a search result are provided.

  The question text input to the input unit 101 is output to the information type determination unit 102. The information type determination unit 102 performs morphological analysis of the question sentence using the morphological analysis dictionary 113 and estimates the information type requested by the user using the information type determination knowledge 111.

  The search unit 103 generates a search condition using the search condition expansion dictionary 114 and the estimated information type, and searches information stored in the knowledge database 220. The output unit 104 displays the search result on the screen and outputs it to the user.

  Since the part for generating the search condition from the user's question sentence is the same as that for generating the search condition from the search request sentence in the first embodiment, the different part will be mainly described.

  (Preparation before System Use) In this embodiment, search data is generated from document data and stored in the knowledge database 220 prior to system use by the user.

  The information extraction unit 222 performs morphological analysis using the morphological analysis dictionary 113. Further, based on the dictionary or pattern matching rule stored in the extraction knowledge 223, tags for each information type are embedded in proper nouns and dates in the document data to generate search data. .

  FIG. 7 shows an example of search data. In the document 700, a tag “<PERSON> (person name)” is assigned to “○ × Taro”, and a tag “<DATE> (date)” is assigned to “△△△△ year ○ month × day”. Is assigned. Further, 00001 is assigned to the document 700 as DOCUMENT_ID. DOCUMENT_ID is an identifier of a document.

  The registration unit 221 registers the generated search data in the knowledge database 220.

  It should be noted that new data may be added to the knowledge database 220 using the idle time when the user is not using the system. Alternatively, it may be performed in the background while the user is using the system.

  (System Operation) The operation of the question answering system of this embodiment will be described below with reference to FIG.

  When the user inputs a question sentence, the information type determination unit 102 performs morphological analysis on the question sentence and determines the information type requested by the user in the same manner as in the first embodiment. Do. The search unit 103 generates a search condition and adds a search term according to the information type with reference to the search condition expansion dictionary 114 to expand the search condition (steps S801 to S803).

  The search unit 103 searches the knowledge database 220 using the extended search condition, and obtains an answer based on the information type requested by the user from the information extraction result of the document higher in the search result. And an answer is output from the output part 104 (step S804-step S806).

  For example, it is assumed that search conditions “XX”, “President”, “Born”, “Birthday”, and “Birth date” are generated from the question sentence “? Then, it is assumed that the document 700 with DOCUMENT_ID 00001 in FIG.

  Now, since the question sentence is “What was the day when the president was born?”, The information type determination unit 102 determines that the information type requested by the user is DATE. Therefore, the character string “ΔΔΔΔ year * month * day” that matches the information type DATE and is near the search term “◯ ×” “date of birth” in the document 700 is obtained as an answer.

  When there are a plurality of character strings whose information type is DATE, evaluation is performed using the number of search terms in the vicinity and the distance (the number of characters between the character string and the search terms). For example, evaluation may be performed using the number of search terms ÷ distance as an evaluation value.

  (Effects of the present embodiment) As described above, according to the present embodiment, an appropriate search term can be added according to the information type requested by the user in the question answering system.

  (Third Embodiment) Hereinafter, a third embodiment of the present invention will be described with reference to the drawings. FIG. 9 is a diagram for explaining the configuration of the document search system of this embodiment. In the present embodiment, the document data is divided in advance into N document databases 112-1 to 112-N. This corresponds to, for example, registering newspaper article data in a document database by age group or field.

  Except for the search unit 103, the description is omitted because it is the same as the first embodiment. The search unit 103 performs a search by limiting the search target using a search target limiting rule 900 described later.

  FIG. 10 is a diagram for explaining processing of the search unit 103 in the present embodiment.

  The search unit 103 receives the morphological analysis result and the information type determination result of the search request sentence from the information type determination unit 102 (step S1001). The search unit 103 generates a search condition from the morphological analysis result (step S1002).

  The search unit 103 refers to the search target limitation rule 900 and limits the document database 112-x (x = 1 to N or a plurality of search targets) according to the information type determination result (step S1003). ). Then, the limited document database 112-x is searched (step S1004), and the search result is passed to the output unit 104 (step S1005).

  FIG. 11 shows an example of the search target restriction rule 900. The search target limitation rule 900 includes a search target limitation trigger 1101, a limitation method 1103, and an information type 1102.

The search target limitation trigger 1101 is information that serves as a clue to limit the search target. In the present embodiment, the search target limiting trigger 1101 is described in a regular expression pattern (in FIG. 11, the description is different from the original regular expression in order to simplify the description. For example, the year of the year Regular expressions that match all issues are abbreviated as “XXXX year”.) The search unit 103 first checks whether there is an expression that matches the search target limitation trigger 1101 in the search condition.

  The information type 1102 is information for changing the way of limitation according to the determination result of the information type determination unit 102. The value of each item of the information type 1102 corresponds to one of the information types that can be determined by the information type determination unit 102, or has a value of “other”. Each item of the search target limitation trigger 1101 has an item of information type 1102 independently.

  The limiting method 1103 is information having an actual limiting method. The value of each item of the limited information 1103 represents, for example, the characteristics (field, age, etc.) of the database to be searched, or a value “not limited”. Each item of the information type 1102 is associated with each item of the limiting method 1103. Every item of the information type 1102 has only one corresponding restriction method 1103 item.

  In FIG. 11, for items whose search target limitation trigger 1101 is “XXXX year”, items “ACCIDENT (accident)”, “EVENT (event)”, and “others” are set as the information type 1102, and each item is The limiting method 1103 is associated with the items “after XXXXXX”, “after (XXXX-1)”, and “not limited”.

  The purpose of setting the limiting method 1103 in this way is as follows. “ACCIDENT” is an event that cannot be predicted in advance. Therefore, it is considered meaningless to search for information before the accident occurred. On the other hand, “EVENT (event)” is an event that becomes a topic before the event is executed and information flows. Therefore, it is considered that setting a wide search range retroactively affects the search results.

  The search unit 103 limits the search target according to the limiting method 1103 specified by the search target limiting trigger 1101 and the information type 1102.

  Hereinafter, the search target limiting process using the search target limiting rule 900 in the search unit 103 will be described with specific examples.

Regular expression pattern matching using the search target restriction trigger 1101 is performed on the user's search request text. For example, if the search request is
(Search Request 1) “What is the train accident that occurred on June 27, 1999?”
(Search Request 2) “What are sports events held jointly between Japan and Korea in 2002?”
11, the trigger “XXXX year” in FIG. 11 matches, (Search request sentence 1) becomes “XXXX = 1999”, and (Search request sentence 2) becomes “XXXX = 2002”.

  Suppose that the information type determination unit 102 determines that the information type of (search request sentence 1) is “ACCIDENT (accident)” and the information type of (search request sentence 2) is “EVENT (event)”. .

  In this case, according to each item of the information type 1102 included in the trigger “XXXX year” in FIG. 11, “after 1999” is selected as the search contrast limiting method 1103 for (search request sentence 1). For (search request sentence 2), the search target limitation “after (2002-1)”, that is, “after 2001” is selected as the search contrast limiting method 1103.

  (Effects of this embodiment) As described above, according to this embodiment, it is possible to flexibly limit the search object according to the information type determined from the search request sentence. As a result, information closer to the information requested by the user can be easily extracted as a search result.

  (Modification) FIGS. 12A and 12B and FIGS. 13A and 13B are diagrams showing an example of an interface when the search unit 103 presents the result of limiting the search target to the user. The user may change the limited range as appropriate by operating the check boxes 1201 to 1206 and the check boxes 1301 to 1304.

  (Fourth Embodiment) FIG. 14 is a diagram for explaining the configuration of a question answering system according to a fourth embodiment of the present invention. The third embodiment is a document retrieval system that outputs a list of documents, whereas the present embodiment is a question answering system that outputs an answer to a question sentence. Further, the knowledge database to be used is not limited in the second embodiment, but the present embodiment is different in that the limitation is performed according to the question text of the user. The following description will focus on differences from the second and third embodiments.

  In the present embodiment, document data is classified in advance by age or field, and information is extracted by the information extraction unit 222 in the same manner as in the second embodiment, and N knowledge databases 220 are extracted. -1, 220-2,..., 220-N are generated.

  As in the second embodiment, the information type determination unit 102 performs morphological analysis and information type determination on a question sentence input by the user. Similarly to the third embodiment, the search unit 103 limits the knowledge database 200-x (x = 1 to N) to be searched using the information type determination result and the search target limiting rule 900. The limited knowledge database 200-x (x = 1 to N) is searched to select an answer and output to the output unit 104. The knowledge database need not be limited to one, and may be limited to a plurality.

  FIG. 15 is a diagram for explaining the flow of processing of the search unit 103 in this embodiment. The search unit 103 receives the morphological analysis result and the information type determination result of the question text input by the user from the information type determination unit 102 (step S1501). Search conditions are created based on the morphological analysis results (step S1502).

  Next, as in the third embodiment, for example, a search target limiting method is determined using a search target limiting rule 900 as shown in FIG. 11, and the knowledge database to be searched is limited. (Step S1503).

  Then, the knowledge database is searched within a limited range (step S1504), and an answer that matches the information type determination result is obtained from the information extraction result of the upper document of the search result (step S1505). (Step S1506).

  For example, it is assumed that a question text “What is a train accident that occurred on June 27, 1999” is input. The information type determination unit 102 determines that the information type of this question is “ACCIDENT (accident)”. The search unit 103 generates a search condition for the question sentence and limits the search target. Here, if the search target limiting rule 900 shown in FIG. 11 is used, the search target limiting method is “after 1999”.

  The retrieval unit 103 retrieves a knowledge database storing information “after 1999”. As a result of the search, the character string with the tag <ACCIDENT> is extracted from the document obtained at the top as in the second embodiment, and presented to the user via the output unit 104 as an answer.

(Effects of the present embodiment) As described above, in the present embodiment, flexible search target limitation according to the information type detected from the question sentence can be performed, so an appropriate answer is obtained from the knowledge database. Will be able to.

  (Modification) In the fourth embodiment, there are a plurality of knowledge databases. However, it may be configured as a single database and may have attribute information such as “age” and “field”. In this case, the search target limitation may be used as a search condition for the attribute information.

  Moreover, you may implement combining the 1st-4th embodiment demonstrated above. That is, a system that performs both document search and question answering may be constructed, or a system that performs both search condition expansion according to the information type and search target limitation according to the information type may be constructed. .

  In the description of the first to fourth embodiments, the document search and the question answering in Japanese have been described. However, the present invention is not limited to the language type.

The figure of the document search system of a 1st embodiment. The figure of the flow of processing of the information type judgment part of a 1st embodiment. The figure of the flow of processing of the search part of a 1st embodiment. The figure of an example of the data structure of a search condition expansion dictionary. (A) An example of an interface for presenting information type determination results and search condition correction results to the user. (B) An example of an interface for presenting information type determination results and search condition correction results to the user. The figure of the question answering system of a 2nd embodiment. The figure of an example of the data structure of an information extraction result. The figure of the flow of processing of the search part of a 2nd embodiment. The figure of the document search system of 3rd Embodiment. The figure of the flow of processing of the search part in a 3rd embodiment. The figure of an example of a search object limitation rule. (A) An example of an interface for presenting a search target limited result to a user. (B) The figure of an example of the interface which presents a search object limitation result to a user. (A) The figure of an example of the interface which presents a search object limitation result to a user. (B) The figure of an example of the interface which presents a search object limitation result to a user. The figure of the question answering system of a 4th embodiment. The figure of the flow of processing of the search part of a 4th embodiment. The figure of the computer utilized in 1st-4th embodiment.

Explanation of symbols

101 Input unit 102 Information type determination unit 103 Search unit 104 Output unit 111 Information type determination knowledge 112, 112-1, 112-2, 112-N Document database 113 Morphological analysis dictionary 114 Search condition expansion dictionary 220, 220-1, 220 -2, 220-N Knowledge database 221 Registration unit 222 Information extraction unit 223 Extraction knowledge 900 Search target limitation rule

Claims (11)

A search system for searching a document database storing document data based on a search request sentence input by a user,
An input unit for inputting a user search request sentence;
An information type determination unit that determines an information type requested by the user from the search request sentence;
Extracting a search term from the search request sentence;
The search term;
A phrase according to the information type;
A search condition generation unit that generates a search condition using
A search unit for searching the document database using the search condition;
An output unit for presenting the search result to the user;
A document retrieval system comprising: The search condition generation unit
Means for extracting a search term from the search request sentence;
Means for generating a search condition using the search term;
Means for extending the search condition using a phrase according to the information type;
The document search system according to claim 1, comprising: The document database is a document database storing document data by classification,
The means for expanding the search condition comprises means for adding a condition for limiting the search range of the document database according to the information type in the classification to the search condition.
The document search system according to claim 2. A search system for searching a document database storing document data by classification based on a search request sentence input by a user,
An input unit for inputting a user search request sentence;
An information type determination unit that determines an information type requested by the user from the search request sentence;
A search condition generator for generating a search condition from the search request sentence;
A search target limiting unit that limits a search range of the document database to those according to the information type among the classifications;
Within this limited search range, a search unit that searches the document database using the search condition;
An output unit for presenting the search result to the user;
A document retrieval system comprising: A question answering system that searches a knowledge database storing knowledge data based on a question sentence input by a user and obtains an answer to the question sentence,
An input unit for inputting a user's question text;
An information type determination unit for determining an information type requested by the user from the question sentence;
Extract search terms from the question,
The search term;
A phrase according to the information type;
A search condition generation unit that generates a search condition using
A search unit for searching the knowledge database using the search condition;
An answer generator for generating an answer to the question sentence based on the search results;
An output unit for presenting the generated answer to the user;
A question answering system comprising: The search condition generation unit
Means for extracting a search term from the question sentence;
Means for generating a search condition from the search term;
Means for extending the search condition using a phrase according to the information type;
The question answering system according to claim 5 provided with. The knowledge database is a knowledge database storing knowledge data by classification,
The means for extending the search condition includes means for adding a condition for limiting the search range of the knowledge database to the search condition according to the information type among the classifications.
The question answering system according to claim 6. It is a question answering system that searches a knowledge database storing knowledge data by classification based on a question sentence input by a user and obtains an answer to the question,
An input unit for inputting a user's question text;
An information type determination unit for determining an information type requested by the user from the question sentence;
A search condition generator for generating a search condition from the question sentence;
A knowledge database that stores knowledge data by category;
A search target limiting unit that limits a search range of the knowledge database to a type corresponding to the information type among the classifications;
A search unit that searches the knowledge database within a limited search range using the search condition;
An output unit that presents the search result to the user as an answer to the question text;
A question answering system comprising: A method of retrieving a document from a document database storing document data,
In response to a search request from a user,
Determine the information type that the user is seeking from the search request text,
A search term is generated by extracting a search term from the search request sentence,
Extending the search criteria with words according to the information type,
Search the document database using extended search criteria;
Present search results to users,
Document search method. A method of retrieving a document from a document database storing document data,
In response to a search request from a user,
Determine the information type that the user is seeking from the search request text,
Extracting a search term from the search request sentence;
A search condition is generated using the search term and the phrase according to the information type,
Search the document database using the search condition,
Present search results to users,
Document search method. A method of retrieving a document from a document database storing document data by classification,
In response to a user search request,
Determine the information type that the user is seeking from the search request text,
Generating a search condition from the search request sentence;
The search range of the document database is limited to the classification according to the information type,
Using the search condition, search the document database within a limited search range,
Present search results to users,
Document search method.