JP2005202627A - Question answering device and question answering system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a question answering device and system for holding the priority of a database structurer. <P>SOLUTION: This question answering device is provided with a document analyzing part for acquiring a retrieval word by analyzing a question sentence, a document storage part for storing a document group including answer candidates to the question sentence, a document retrieving part for acquiring an identifier group for identifying a document corresponding to the retrieval word by referring to the document storing part, a session managing part for enabling a user to arbitrarily select the identifier group and an answer extracting part for acquiring a document corresponding to the identifier group from the session managing part by referring to the document storing part, and for extracting the answer candidates corresponding to the predetermined question classification from the document. This question answering device is also provided with an answer processing control part for performing the decision of the number of documents as the decision of answer processing based on the identifier group, and for, when the number of documents is not more than a predetermined reference value according to the decision, outputting the result that there is no answer to the question sentence as the answer result of the question sentence. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a question answering apparatus and a question answering system for obtaining an answer to a question sentence from information held in advance.

Japanese Patent Application Laid-Open No. 2004-151867 discloses a question answering system that obtains an answer to a question sentence from information held in advance and responds.
According to the invention of Patent Document 1, a morpheme of a question sentence is analyzed, a question analysis process for obtaining a search word and a question type that are the main contents of a question, and a document set (searched document) held in advance, A document search process for searching a document in which a search term is indicated, and in the document obtained by the search, a score is given to the word according to the question type based on the search word and the appearance frequency, and the answer is based on the score An answer extraction process for determining a word to be performed is performed to obtain an answer according to the question content.

For example, a description will be given of a case in which a document with a search target document is held as “About fuel cell” and a text indicating “What is a battery using fuel?” Is input as a question sentence.
In the question analysis process, morphological analysis of the question sentence is performed, and words separated by this analysis are acquired. That is, in the question analysis process, morphemes are analyzed as “What is the battery that uses fuel?” And words such as “fuel” and “battery” are acquired.

  The document search process uses each acquired word as a search word, and checks whether the search word is included in the searched document. In other words, in the document search process, the searched sentence is morphologically analyzed as “about fuel cell”. As a result, the “fuel cell” in the searched document is divided into “fuel” and “battery”, and the searched word corresponding to the question sentence is obtained by matching the divided word with the word acquired as the search word. Identified.

  When there are a plurality of documents to be searched, session management is performed so that the user can arbitrarily select. Answer extraction processing is performed on the search target document arbitrarily selected by the session management.

In the answer extraction process, a score is assigned to each word of the identified search target document based on the question type and the search word / appearance frequency. Are output together with the word extraction source document.
Japanese Patent Laid-Open No. 2002-132911

By the way, when answer candidates are extracted, if there is only one document from which the answer candidates are extracted, or if the user intentionally selects only one document to be searched in session management, The person can easily identify the document from which the answer candidates are extracted.
In other words, since the user can easily know from which document (searched document) the answer candidate is extracted, the relationship between the searched document and the answer candidate associated with the searched document, so-called database You can know the contents easily. This could reduce the superiority of database builders, which was feared.

  Accordingly, an object of the present invention is to provide a question answering apparatus and a question answering system that can maintain the superiority of a database builder.

The present invention adopts the following configuration in order to solve the above points.
<Configuration 1>
A document analysis unit that analyzes a question sentence from a user and obtains a search term, a document storage unit that accumulates a document group including answer candidates for the question sentence, and the document storage unit, and corresponds to the search term A document search unit for acquiring an identifier group for identifying a document to be performed, a session management unit for a user to arbitrarily select the identifier group from the document search unit, and the document storage unit, When a document corresponding to the identifier group is acquired from the session management unit, an answer extraction unit that extracts an answer candidate corresponding to a predetermined question type from the document, based on the identifier group An answer processing control unit that performs determination of the number of documents as an answer process judgment, and outputs an answer result of the question text that there is no answer to the question text when the number of documents is equal to or less than a predetermined reference value in this judgment. Characterized in that it obtain.

<Configuration 2>
Further, referring to the document storage unit, based on the document associated with the identifier group, the query analysis unit acquires a similar search word similar to the search word, and identifies a similar document corresponding to the similar search word A similar document search unit that causes the document search unit to acquire a similar identifier to be used, and the answer processing control unit has a number of documents equal to or less than a predetermined reference value based on the identifier group obtained by the document search unit. When the similar document search unit is activated and a similar identifier is obtained from the similar document search unit, the answer processing determination is performed based on the number of documents obtained by adding the similar identifier to the similar identifier group. When the number of documents is larger than a predetermined reference value in this determination, the answer extraction unit is activated so that the answer source is extracted and the extraction source document is output as the answer result of the question sentence.

<Configuration 3>
A document analysis unit that analyzes a question sentence from a user and obtains a search term, a document storage unit that accumulates a document group including answer candidates for the question sentence, and the document storage unit, and corresponds to the search term A document search unit for acquiring an identifier group for identifying a document to be performed, a session management unit for a user to arbitrarily select the identifier group from the document search unit, and the document storage unit, When a document corresponding to the identifier group is acquired from the session management unit, an answer extraction unit that extracts an answer candidate corresponding to a predetermined question type from the document, the question answering device includes: The number of documents is determined based on an extraction identifier group that identifies a document from which answer candidates are extracted. If the number of documents is equal to or less than a predetermined reference value in this determination, it is determined that there is no answer to the question sentence. answer Characterized in that it comprises an answer processing controller for outputting a result.

<Configuration 4>
A document analysis unit that analyzes a question sentence from a user and obtains a search term, a document storage unit that accumulates a document group including answer candidates for the question sentence, and the document storage unit, and corresponds to the search term A document search unit for acquiring an identifier group for identifying a document to be performed, a session management unit for a user to arbitrarily select the identifier group from the document search unit, and the document storage unit, When a document corresponding to the identifier group is acquired from the session management unit, an answer extraction unit that extracts an answer candidate corresponding to a predetermined question type from the document is referred to the document storage unit. Then, based on a document associated with the identifier group, the query analysis unit acquires a similar search word similar to the search word, and a similar identifier for identifying a similar document corresponding to the similar search word The number of documents is determined based on a similar document search unit to be acquired by the document search unit, and an extraction identifier group that identifies an extraction source document of answer candidates from the answer extraction unit. If it is above, the extraction source document is output as the answer result of the question sentence together with the extraction of the answer candidate, and when the number of documents is less than a predetermined reference value, the similar document search unit is activated to When the similarity identifier is acquired from the search unit, when the number of documents indicated by the similarity identifier is equal to or greater than a predetermined reference value, the answer extraction unit is activated based on a new identifier generated by adding the similarity identifier to the identifier And an answer processing control unit.

<Configuration 5>
The identifier is a file name of a document or an index that associates the document with a word that can be a search word for the document stored in the document storage unit.

<Configuration 6>
The similar document search unit removes the similar identifier when the similar identifier from the document search unit is included in the identifier group.

<Configuration 7>
In order to construct the question answering apparatus as a server / client, the client includes the session management unit, and the server operates based on the identifier selected by the session management unit, and the answer processing control. A question answering system comprising: each unit under control of a unit; and the document storage unit.

  The question answering apparatus of the present invention determines the number of documents based on an identifier group for identifying a document corresponding to a search word obtained by analyzing a question sentence, and the number of documents is equal to or less than a predetermined reference value by this determination. In some cases, by providing an answer control unit that outputs that there is no answer to the question text as the answer result of the question text, the question text only exists when there are multiple source documents and the user cannot identify the source text Therefore, the user cannot easily grasp the contents of the database, and the superiority of the database builder can be maintained.

  Hereinafter, an example of a question answering system having a client / server configuration will be described in detail with reference to the drawings.

  As shown in FIG. 1, the question answering system 10 of the present invention includes a server 20 and a client 30, and the server 20 and the client 30 are connected to a network as a transmission path (not shown). Data communication is performed between the server 20 and the client 30 via the.

  The client 30 includes an input unit 1100 that receives a question sentence from the user, a session management unit 1500 that manages communication with the server 20 based on the question sentence received by the input unit 1100, and answers to the question sentence. And an output unit 1700 to be presented to the user.

  In addition, the server 20 performs morphological analysis of the question sentence, searches the query analysis unit 1200 for acquiring a search term that is the key of the question content, searches for a document that matches the search term, and outputs the search result to the session management unit 1500. A document search unit 1400 that performs a document search, a document storage unit 1300 that holds a document for the document search unit 1400, and an answer processing control unit that controls an answer to the question sentence based on the number of documents acquired from the session management unit 1500 1600, and an answer extraction unit 1800 for extracting answer candidates for the question sentence from the document processed by the answer processing control unit 1600.

  The question analysis unit 1200 performs natural language processing such as morphological analysis on the question sentence, and acquires a word that is the key of the question sentence as a search word. For example, the question analysis unit 1200 that has acquired “XX inventor?” As a question sentence performs a morphological analysis with “XX device, developer ,?”, and from each divided word, the keyword of the question sentence A word with a possible part of speech (for example, a noun) is acquired as a search word. At this time, the question analysis unit 1200 may acquire a question type for answer extraction processing in an answer extraction unit 1800 described later based on the search word.

  The document accumulating unit 1300 holds a document set as a source for extracting question answer answers. By the way, the document accumulating unit 1300 applies the document (answer candidate) to the word (answer candidate) obtained in advance by performing natural language analysis processing such as morphological analysis for each document in order to improve the search efficiency in the document search unit 1400 described later. It is desirable to keep the identifiers associated with each other.

  The document search unit 1400 acquires a document that matches the search word from the question analysis unit 1200 with reference to the document storage unit 1300. In this embodiment, the identifier (identifier group) of the document acquired by the document search unit 1400 is output to the session management unit 1500 as a search result. At this time, the search term is also output to the session management unit 1500 together with the identifier.

  When the session management unit 1500 acquires a document identifier or search word as a search result from the document search unit 1400, the session management unit 1500 manages the identifier and the search word in association with the question sentence. Session management unit 1500 outputs an identifier and a search term to be managed to answer processing control unit 1600. The session management unit 1500 is also provided with an input display function so as to correct a search term and an identifier to be managed based on an instruction from the user as necessary.

  When the response processing control unit 1600 acquires the identifier from the session management unit 1500, the response processing control unit 1600 determines whether or not the identifier is equal to or greater than a predetermined reference value based on the number of identifiers. When the number of identifiers is equal to or less than the reference value, the output unit 1700 In response to this, information indicating “no answer” is output. On the other hand, when the number of identifiers is greater than the reference value, the identifiers and search terms are output to the answer extraction unit 1800.

  When the answer extraction unit 1800 acquires the identifier from the answer processing control unit 1600, the answer extraction unit 1800 refers to the document storage unit 1300 to extract a word having an attribute corresponding to the question type from among answer candidates associated with the identifier. The extracted answer candidates are output from the answer extraction unit 1800 to the output unit 1700. By the way, the question type may be obtained from a question sentence by a conventionally known method, or the session management unit 1500 has a GUI (Graphical User Interface) function, and the user arbitrarily designates it with the GUI or the like. Alternatively, specific question types may be limited in advance.

The answer extraction unit 1800 gives a score to the extracted answer candidate based on the number of appearances of the answer candidate, the number of appearing documents, and the like, and then outputs the score to the output unit 1700.
The output unit 1700 associates a question sentence for extracting the answer candidate with the answer candidate and presents it to the user, or associates that there is no answer with the processing history processed by the session management unit 1500. To present.

Next, operation | movement of the question answering system 10 of Example 1 is demonstrated along the processing flow of FIG.
The user inputs a question sentence from the input unit 1100 (step S10). The input question text is sent to the question analysis unit 1200 via the network. At this time, a question sentence is also sent from the input unit 1100 to the session management unit 1500.

  When the question analysis unit 1200 obtains the question sentence, the question analysis part 1200 performs morphological analysis on the question sentence, obtains a search word that is the key of the question content, and outputs the search word to the document search part 1400 (step S20).

When the search term is acquired, the document search unit 1400 refers to the document storage unit 1300 to acquire the identifier of the document (searched document) corresponding to the search term, and outputs the identifier and the search term to the session management unit 1500. (Step S30). Here, an example in which a plurality of identifiers are acquired as an identifier group by the document search unit 1400 will be described below.
The session management unit 1500 manages the query text from the input unit 1100 in association with the search word and identifier group from the document search unit 1400 (step S40). When the session management unit 1500 receives a question type from the user, the session management unit 1500 also manages the question type in association with the question sentence or the identifier group.

When the user instructs the session management unit 1500 to request an answer to a question sentence, the identifier group acquired based on the search term is output from the session management unit 1500 to the response processing control unit 1600.
When the answer processing control unit 1600 acquires the identifier group from the session management unit 1500, the answer processing control unit 1600 determines whether the number of identifiers is equal to or less than a predetermined reference value (step S50). In this embodiment, the reference value is 1. When the number of identifiers is 1 or less, the answer processing control unit 1600 outputs information indicating no answer to the output unit 1700. Upon receiving this information, the output unit 1700 presents that fact to the user.

On the other hand, when there are two or more identifiers, the answer processing control unit 1600 outputs the identifiers to the answer extracting unit 1800.
When the answer extraction unit 1800 obtains the identifier and the search word, the answer extraction unit 1800 refers to the document storage unit 1300 to obtain a word (answer candidate) associated with the identifier (step S60). At this time, the answer extraction unit 1800 acquires the question type from the session management unit 1500 and extracts answer candidates corresponding to the question type. The answer extraction unit 1800 assigns a score to the extracted answer candidates based on the number of appearances, the number of appearance documents, and the like by a conventionally known method, and then outputs the score to the output unit 1700.
The output unit 1700 associates a question sentence for extracting the answer candidate with the answer candidate and presents it to the user (step S70).

  Here, an operation when the user narrows down the identifier group managed by the session management unit 1500 using the GUI or the like will be described. In the session management unit 1500, when the user narrows down with the GUI, a new identifier group is generated. An operation when the number of identifiers in this identifier group becomes one will be described (hereinafter referred to as identifier group, but the number of identifiers is one). When the identifier group is output to the answer processing control unit 1600, the answer processing control unit 1600 determines the number of identifiers. The response processing control unit 1600 that has determined that the number of identifiers is 1 or less in this determination does not send the identifier group to the response extraction unit 1800 and outputs “no response” to the output unit 1700 via the network. To do. Thereby, when the number of identifiers is 1 or less, the user cannot obtain answer candidates.

  As described above, according to the question answering system 10 of the first embodiment, when obtaining answer candidates from each document held in the document storage unit 1300 (each document with an identifier associated with each word), the user If it is not possible to know from which document the answer candidate was extracted, that is, if it does not know from which document the answer candidate was extracted, the answer candidate is output as usual, and the document from which the answer candidate was extracted can be identified Since the answer to the question sentence is output as “no answer”, the user cannot know from which document the answer candidate is extracted.

  Therefore, the user cannot know the database configuration in which the answer candidates obtained from each document are associated with each document, and the possibility that the profit of the server operator who built the database is impaired can be reduced.

Next, a question answering system 11 in which a similar document search unit is added to the configuration of the question answering system 10 of the first embodiment will be described. In the question answering system 11, as shown in FIG. 3, the server 20 and the client 30 are connected by a network (not shown).
The client 30 includes an input unit 1100, a session management unit 1500, and an output unit 1700.

  The server 20 is newly similar to the document storage unit 1300, the answer extraction unit 1800, the question analysis unit 2200 in the second embodiment, the document search unit 2400 in the second embodiment, and the answer processing control unit 2600 in the second embodiment. A document search unit 2900.

Since the configuration content on the client 30 side is the same as in the first specific example, the description thereof will be omitted, and the server 20 will be described only for each part different from the configuration in the first specific example.
The question analysis unit 2200 acquires a search term based on the question sentence as in the first embodiment, and outputs the acquired search term to the document search unit 2400. In addition to this, the question analysis unit 2200 performs morphological analysis on a document from a similar document search unit 2900, which will be described later, and acquires a part of speech word that can be a keyword of the document as a similar search word. The data is output to the similar document search unit 2900.

  As in the first embodiment, when the document search unit 2400 acquires a search word from the question analysis unit 2200, the document search unit 2400 refers to the document storage unit 1300, and an identifier associated with an answer word (document) that matches the search word. The group is acquired, and the identifier group is output to the session management unit 1500. In addition, when the document search unit 2400 obtains a similar search word from the similar document search unit 2900, the document search unit 2400 refers to the document storage unit 1300 to identify an identifier associated with a response word (document) that matches the similar search word. Is output as a similar identifier, and the similar identifier is output to the similar document search unit 2900.

  When the answer processing control unit 2600 acquires the identifier group from the session management unit 1500, the answer processing control unit 2600 determines whether the number of identifiers is equal to or less than a predetermined reference value. In the present embodiment, the following description will be made assuming that the reference value is 1. When the number of identifiers is 1 or less, the answer processing control unit 2600 outputs the identifier group to 2900. Then, the response processing control unit 2600 that acquired the identifier of another response word (document) similar to the response word (document) associated with the output identifier group, that is, the similar identifier from the similar document search unit 2900 is included in the identifier group. A new identifier group to which similar identifiers are added is generated, and it is determined whether or not the number of identifiers in the identifier group is equal to or less than a predetermined reference value. The output based on the determination result is the same as that in the first embodiment.

  That is, when the number of identifiers is 1 or less in the newly generated identifier group by adding similar identifiers, the answer processing control unit 2600 outputs “no answer” to the output unit 1700.

  On the other hand, when the number of identifiers is 2 or more, the answer processing control unit 2600 outputs the identifier group to the answer extracting unit 1800 as in the first embodiment without performing the process in the similar document search unit 2900. To do.

  The similar document search unit 2900 refers to the document storage unit 1300, acquires an identifier of a document similar to the document associated with the identifier group from the response processing control unit 2600 as a similar identifier, and uses the similar identifier as a response processing control. Part 2600.

  That is, when the similar document search unit 2900 acquires the identifier group from the answer processing control unit 2600, the similar document search unit 2900 acquires the document associated with the identifier group with reference to the document storage unit 1300, and the query analysis unit 2200 for the document. Then, morphological analysis is performed to obtain similar search terms. Then, the similar document search unit 2900 causes the document search unit 2400 to acquire a similar identifier based on the acquired similar search word, and outputs the acquired similar identifier to the answer processing control unit 2600.

Next, the operation of the question answering system 11 will be described.
FIG. 4 is a processing flow showing the processing timing of each part in the question analysis unit 2200, document search unit 2400, similar document search unit 2900, answer processing control unit 2600, and answer extraction unit 1800 in this order. I do.

  Until the query analysis unit 2200 acquires a search term based on the question sentence acquired by the input unit 1100, acquires an identifier group by the document search unit 2400 based on the search term, and manages the identifier group by the session management unit 1500 The processing content of is the same as that of the question answering system 10 of the first embodiment. That is, the question analysis part 2200 which acquired the question sentence input by the input part 1100 performs a morphological analysis, and acquires a search word (step S201). When the search word acquired by the question analysis unit 2200 is sent to the document search unit 2400, the document search unit 2400 refers to the document storage unit 1300, and an identifier associated with an answer word (document) that matches the search word. The group is output to the session management unit 1500.

  The following description will be given with an example in which the number of identifiers in the identifier group managed by the session management unit 1500 is one. The response processing control unit 2600 that has acquired the identifier group from the session management unit 1500 determines whether or not the number of identifiers in the identifier group is equal to or less than a predetermined reference value (1) (step S203). In this determination, when the number of identifiers is 1 or less, the identifier group is output to the similar document search unit 2900.

  The similar document search unit 2900 that has acquired the identifier group acquires a document corresponding to the identifier with reference to the document storage unit 1300 (step S204). The acquired document is output from the similar document search unit 2900 to the question analysis unit 2200.

  When the question analysis unit 2200 acquires a document from the similar document search unit 2900, the question analysis unit 2200 performs a morphological analysis of the document and acquires a similar search word (step S205). The acquired similar search term is output from the question analysis unit 2200 to the similar document search unit 2900.

The similar document search unit 2900 that has acquired the similar search term from the question analysis unit 2200 outputs the similar acquisition search term to the document search unit 2400.
The document search unit 2400, which has acquired the similar search word, refers to the document storage unit 1300, acquires an identifier corresponding to the similar search word as a similar identifier, and uses the similar identifier as the similar document search unit 2900. (Step S206).

  The similar document search unit 2900 that has acquired the similar identifier checks whether or not the identifier acquired in step S202 is included in one or a plurality of similar identifiers, and if included, generates a similar identifier by removing the identifier. Then, the similarity identifier is output to the answer processing control unit 2600 (step S207).

The response processing control unit 2600 that has acquired the similar identifier generates a new identifier group obtained by adding the similar identifier to the identifier group acquired in step S202 (step S208).
The answer processing control unit 2600 that has generated a new identifier group determines whether or not the number of identifiers in the identifier group is equal to or smaller than a predetermined reference value (1) (step S209). In this determination, when the number of identifiers is 1 or less, information indicating “no answer” is output to the output unit 1700. On the other hand, when the number of identifiers is 2 or more, the identifier group is output to the answer extraction unit 1800.

  The answer extraction unit 1800 that has acquired the identifier group from the answer processing control unit 2600 extracts answer candidates for the question sentence, assigns scores to the answer candidates, and outputs them to the answer extraction unit 1800 as in the first embodiment. (Step S210).

  As described above, according to the question answering system 11 of the second embodiment, when an answer candidate is obtained from each document held in the document storage unit 1300 (document of each identifier associated with each word), the identifier group Is equal to or less than a predetermined reference value, a similar search word is acquired from a document corresponding to the identifier group, a similar identifier corresponding to the similar search word is added to the identifier group, and a new identifier group is generated. As a result, the answer extraction process is performed based on the identifier group composed of a plurality of identifiers, and the user cannot know from which document the answer candidate is extracted.

  Therefore, the user cannot know the database configuration in which the answer candidates obtained from each document are associated with each document, and the possibility that the profit of the server operator who built the database is impaired can be reduced.

Next, a question answering system 12 including an answer processing control unit 3600 that replaces the answer processing control unit 1600 of Example 1 and an answer extraction unit 3800 that replaces the answer extraction unit 1800 of Example 1 will be described.
The configuration of the question answering system 12 other than the answer processing control unit 3600 and the answer extracting unit 3800 is the same as that of the question answering system 10 of the first embodiment described above. That is, in the question answering system 12, as shown in FIG. 5, the server 20 and the client 30 are connected by a network (not shown), and the client 30 includes an input unit 1100, a session management unit 1500, and an output unit 1700. .

The server 20 includes a question analysis unit 1200, a document storage unit 1300, a document search unit 1400, a new answer processing control unit 3600, and a new answer extraction unit 3800.
Other than the answer processing control unit 3600 and the answer extraction unit 3800 are the same as those in the first embodiment, and the description thereof will be omitted.

  When the answer extraction unit 3800 obtains an identifier group from the answer processing control unit 3600 described later, the answer extraction unit 3800 refers to the document storage unit 1300 and selects an attribute word corresponding to the question type from among answer candidates associated with the identifier. Extract. The answer extraction unit 1800 assigns a score to the extracted answer candidates based on the number of appearances and the number of appearance documents by a conventionally known method, and then outputs the score to the answer processing control unit 3600. In addition to this, the answer extraction unit 3800 acquires an extracted identifier group composed of identifiers associated with each extracted answer candidate, and outputs the extracted identifier group to the answer processing control unit 3600. The output to the answer processing control unit 3600 may be the number of documents for which answer candidates are obtained.

  When the answer processing control unit 3600 outputs the identifier group from the session management unit 1500 to the answer extracting unit 3800 and acquires the answer candidate and the extracted identifier group as the processing result in the answer extracting unit 3800, the extracted identifier group It is determined whether or not the number of identifiers in is less than or equal to a predetermined reference value. In the present embodiment, the following description will be made assuming that the reference value is 1. When the number of identifiers is 1 or less, the answer processing control unit 3600 outputs “no answer” to the output unit 1700, and if the number of identifiers (number of documents) is 2 or more, the answer candidate and its score are output. Part 1700.

Next, the operation of the question answering system 12 will be described using a flowchart showing the operation of the answer processing control unit 3600 showing the features of this embodiment.
The query analysis unit 2200 acquires a search word based on the question sentence, the document search unit 2400 acquires an identifier group based on the search word, the session management unit 1500 manages the identifier group, and the identifier group is processed as an answer The processing content until output to the control unit 3600 is the same as that of the question answering system 10 of the first embodiment.

The answer processing control unit 3600 outputs the acquired identifier group to the answer extracting unit 3800, and acquires the answer candidate and the extracted identifier group from the answer extracting unit 3800 as a processing result in the answer extracting unit 3800 (step S301).
The answer processing control unit 3600 determines whether or not the number of identifiers in the extracted identifier group is 1 or less (step S302). When the number of identifiers (number of documents) is 2 or more, the answer processing control unit 3600 outputs answer candidates and their scores to the output unit 1700.
On the other hand, when the number of identifiers is 1 or less, the answer processing control unit 3600 outputs “no answer” to the output unit 1700 (step S304).

  As described above, according to the question answering system 12 of the third embodiment, when the number of identifiers (extraction identifier group) associated with the answer candidates extracted by the answer extraction unit 3800 is equal to or less than a predetermined reference value, Since the answer is not made, that is, the answer is not made except when there are a plurality of question candidate extraction target documents, the user cannot know from which document the answer candidate is extracted.

  Therefore, the user cannot know the database configuration in which the answer candidates obtained from each document are associated with each document, and the possibility that the profit of the server operator who built the database is impaired can be reduced.

Next, a question answering system 13 combining the second embodiment and the third embodiment will be described.
As shown in FIG. 7, the question answering system 13 includes a server 20 and a client 30 connected via a network (not shown). The client 30 includes an input unit 1100, a session management unit 1500, and an output unit 1700.

  The server 20 includes a question analysis unit 2200, a document search unit 2400, a similar document search unit 2900, a document storage unit 1300, an answer extraction unit 3800, and a new answer processing control unit 4600.

Except for the answer processing control unit 4600, the explanation is omitted because it is the same as the above-described embodiment.
When the answer processing control unit 4600 according to the fourth embodiment outputs the identifier group from the session management unit 1500 to the answer extracting unit 3800 and acquires the answer candidate and the extracted identifier group as the processing result in the answer extracting unit 3800, It is determined whether or not the number of identifiers in the extracted identifier group is equal to or less than a predetermined reference value. In the present embodiment, the following description will be made assuming that the reference value is 1. When the number of identifiers is 1 or less, the answer processing control unit 4600 outputs the identifier group to 2900. The response processing control unit 4600 obtained from the similar document search unit 2900 as an identifier of another response word (document) similar to the response word (document) associated with the output identifier group is similar to the identifier group. A new identifier group to which the identifier is added is generated.

  Then, the answer processing control unit 4600 outputs the newly generated identifier group to the answer extracting unit 3800, and when the answer candidate and the extracted identifier group are acquired as the processing result in the answer extracting unit 3800, the identifier in the extracted identifier group It is determined whether or not the number is less than or equal to a predetermined reference value. The answer processing control unit 4600 repeats the above process a predetermined number of times, or repeats until the number of identifiers (number of documents) in the extracted identifier group from the answer extraction unit 3800 becomes a predetermined value or more, or from the similar document search unit 2900 Repeat until no similar identifier is available.

  As a result, when the number of identifiers in the identifier group is equal to or smaller than the predetermined value, the answer processing control unit 4600 outputs information indicating “no answer” to the output unit 1700, and the number of identifiers in the identifier group is larger than the predetermined value. At this time, a score is assigned to the answer candidate, and this is output to the answer extraction unit 1800.

Next, the operation of the question answering system 13 will be described.
FIG. 8 is a processing flow showing the processing timing of each part in the question analysis unit 2200, document search unit 2400, similar document search unit 2900, answer processing control unit 4600, and answer extraction unit 3800 in this order. I do.

  Until the query analysis unit 2200 acquires a search term based on the question sentence acquired by the input unit 1100, acquires an identifier group by the document search unit 2400 based on the search term, and manages the identifier group by the session management unit 1500 The processing content of is the same as in the above-described embodiment. That is, the question analysis unit 2200 that has acquired the question text input by the input unit 1100 performs morphological analysis and acquires a search word (step S401). When the search term acquired by the question analysis unit 2200 is sent to the document search unit 2400, the document search unit 2400 refers to the document storage unit 1300, and an identifier associated with an answer word (document) that matches the search term. The group is output to the session management unit 1500 (step S402).

  The answer processing control unit 4600 that has acquired the identifier group from the session management unit 1500 outputs the identifier group to the answer extraction unit 3800. The answer extraction unit 3800 that has acquired the identifier group refers to the document storage unit 1300, acquires the answer candidate and the extracted identifier group, and outputs them to the answer processing control unit 4600 (step S403). The following explanation will be given using an example in which the number of identifiers in the extracted identifier group in the answer extraction unit 3800 is one.

  The answer processing control unit 4600 determines the number of identifiers in the acquired extracted identifier group as a predetermined reference value (step S404). When the number of identifiers in the extracted identifier group is two or more in this determination, the answer processing control unit 4600 gives a score to the answer candidate and outputs it to the output unit 1700.

  On the other hand, when the number of identifiers in the extracted identifier group is 1 or less, the response process control unit 4600 determines whether or not the similar identifier acquisition process has been performed at least once (step S404). When the similar identifier acquisition process is performed at least once, the answer process control unit 4600 outputs information indicating “no answer” to the output unit 1700.

When the similar identifier acquisition processing is not yet performed, the answer processing control unit 4600 outputs the identifier group to the similar document search unit 2900.
The similar document search unit 2900 that has acquired the identifier group acquires a document corresponding to the identifier with reference to the document storage unit 1300 (step S406). The acquired document is output from the similar document search unit 2900 to the question analysis unit 2200.

  When the question analysis unit 2200 acquires a document from the similar document search unit 2900, the question analysis unit 2200 performs a morphological analysis of the document and acquires a similar search word (step S407). The acquired similar search term is output from the question analysis unit 2200 to the similar document search unit 2900.

The similar document search unit 2900 that has acquired the similar search term from the question analysis unit 2200 outputs the similar acquisition search term to the document search unit 2400.
The document search unit 2400, which has acquired the similar search word, refers to the document storage unit 1300, acquires an identifier corresponding to the similar search word as a similar identifier, and uses the similar identifier as the similar document search unit 2900. (Step S408).

  The similar document search unit 2900 that has acquired the similar identifier checks whether or not the identifier acquired in step S202 is included in one or a plurality of similar identifiers, and if included, generates a similar identifier by removing the identifier. Then, the similarity identifier is output to the answer processing control unit 4600 (step S409).

  When the response processing control unit 4600 acquires a similar identifier from the similar document search unit 2900, it determines whether the number of identifiers in the similar identifier is equal to or greater than a predetermined reference value (1) (step S410). In this determination, when the number of identifiers is not 1 or more, that is, when the number of identifiers is 0, the answer processing control unit 4600 outputs information indicating “no answer” to the output unit 1700.

  On the other hand, when the number of identifiers in the similar identifier is one or more, the answer processing control unit 4600 generates a new identifier group in which the similar identifier is added to the identifier group, and the generated identifier group to perform the processing from step S403. Is output to the answer extraction unit 3800 (step S411).

The answer extraction unit 3800 in step S403 adds a new similar identifier in step S411, and based on the added similar identifier, acquires an answer candidate and an extracted identifier group including the new identifier, and the following description is given. To do.
The answer processing control unit 4600 in step S404 determines the number of identifiers in the extracted identifier group. In this determination, in step S403, the number of identifiers in the extracted identifier group including a new identifier is determined. The extracted identifier group includes a new identifier obtained based on the similar identifier added in the process in step S403. That is, the extracted identifier group includes the previously acquired identifier and another identifier acquired this time. ing. Therefore, since at least two or more identifiers are included, the number of identifiers in the extracted identifier group is at least two, and the answer processing control unit 4600 gives a score to the answer candidate and outputs it to the output unit 1700.

  As described above, according to the question answering system 13 of the present invention, when the number of identifiers (extracted identifier group) associated with the answer candidates is equal to or smaller than a predetermined reference value, similar search terms are retrieved from the document corresponding to the identifier group. A new identifier group obtained by adding the similar identifier corresponding to the similar search word to the identifier group, and re-extracting the answer candidate based on the generated identifier group, thereby being associated with the answer candidate. Other identifiers can be obtained. Therefore, since a plurality of documents to which a plurality of identifiers, that is, similar documents are added, can be obtained, the user cannot know from which document the answer candidate is extracted.

  As a result, the user cannot know the database configuration in which the answer candidates obtained from each document are associated with each document, thereby reducing the possibility that the profit of the server operator who built the database is impaired. it can.

  Although the embodiment has been described with the example of the question answering system, the present invention may be implemented as a question answering device without being limited to the client / server configuration.

It is a block diagram of the question answering system of Example 1. It is a figure which shows the processing flow of the question answering system of Example 1. It is a block diagram of the question answering system of Example 2. It is a figure which shows the processing flow of the question answering system of Example 2. It is a block diagram of the question answering system of Example 3. 14 is a flowchart illustrating an operation of an answer processing control unit according to the third embodiment. It is a block diagram of the question answering system of Example 4. It is a figure which shows the processing flow of the question answering system of Example 4.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Question answering system of Example 1 11 Question answering system of Example 2 12 Question answering system of Example 3 13 Question answering system of Example 4 20 Server 30 Client 1100 Input part 1200 2200 Question analysis part 1300 Document storage part 1400 2400 Document search unit 1500 Session management unit 1600 2600 3600 4600 Answer processing control unit 1700 Output unit 1800 3800 Answer extraction unit 2900 Similar document search unit

Claims (7)

A document analysis unit that analyzes a question sentence from a user and obtains a search term, a document storage unit that accumulates a document group including answer candidates for the question sentence, and the document storage unit, and corresponds to the search term A document search unit for acquiring an identifier group for identifying a document to be performed, a session management unit for a user to arbitrarily select the identifier group from the document search unit, and the document storage unit, In a question answering apparatus comprising: an answer extracting unit that extracts an answer candidate corresponding to a predetermined question type from the document when acquiring a document corresponding to the identifier group from the session management unit;
An answer that determines the number of documents based on the identifier group as an answer processing judgment, and outputs an answer to the question text that there is no answer to the question text when the number of documents is equal to or less than a predetermined reference value A question answering apparatus comprising a processing control unit. Further, referring to the document storage unit, based on the document associated with the identifier group, the query analysis unit acquires a similar search word similar to the search word, and identifies a similar document corresponding to the similar search word A similar document search unit for causing the document search unit to acquire a similar identifier for
The answer processing control unit activates the similar document search unit based on the identifier group obtained by the document search unit and activates the similar document search unit when the number of documents is equal to or less than a predetermined reference value. When the identifier is acquired, the answer processing determination is performed based on the number of documents obtained by adding the similar identifier to the similar identifier group. When the number of documents exceeds the predetermined reference value in this determination, the extraction source is extracted together with the extraction of the answer candidates. The question answering apparatus according to claim 1, wherein the answer extraction unit is activated to output the document as an answer result of the question sentence. A document analysis unit that analyzes a question sentence from a user and obtains a search term, a document storage unit that accumulates a document group including answer candidates for the question sentence, and the document storage unit, and corresponds to the search term A document search unit for acquiring an identifier group for identifying a document to be performed, a session management unit for a user to arbitrarily select the identifier group from the document search unit, and the document storage unit, In a question answering apparatus comprising: an answer extracting unit that extracts an answer candidate corresponding to a predetermined question type from the document when acquiring a document corresponding to the identifier group from the session management unit;
The number of documents is determined based on an extraction identifier group that identifies the document from which the answer candidate is extracted from the answer extraction unit. A question answering apparatus comprising: an answer processing control unit that outputs the answer as a result of the question sentence. A document analysis unit that analyzes a question sentence from a user and obtains a search term, a document storage unit that accumulates a document group including answer candidates for the question sentence, and the document storage unit, and corresponds to the search term A document search unit for acquiring an identifier group for identifying a document to be performed, a session management unit for a user to arbitrarily select the identifier group from the document search unit, and the document storage unit, In a question answering apparatus comprising: an answer extracting unit that extracts an answer candidate corresponding to a predetermined question type from the document when acquiring a document corresponding to the identifier group from the session management unit;
Referring to the document storage unit, the query analysis unit is caused to acquire a similar search word similar to the search word based on a document associated with the identifier group, and a similar document corresponding to the similar search word is identified. A similar document search unit for causing the document search unit to acquire a similar identifier for
The number of documents is determined based on an extraction identifier group that identifies an extraction source document of answer candidates from the answer extraction unit. When the number of documents is equal to or greater than a predetermined reference value in this determination, the extraction source is extracted together with the extraction of answer candidates. When the number of documents is less than a predetermined reference value and the similar document search unit is activated to obtain a similar identifier from the similar document search unit, the similar document is obtained. An answer processing control unit that activates the answer extracting unit based on a new identifier generated by adding the similar identifier to the identifier when the number of documents indicated by the identifier is equal to or greater than a predetermined reference value. Question answering device.   4. The identifier according to claim 1, wherein the identifier is a file name of the document or an index that associates the document with a word that can be a search word for the document stored in the document storage unit. 4. The question answering device according to 4.   5. The question answering apparatus according to claim 2, wherein when the similar identifier from the document search unit is included in the identifier group, the similar document search unit removes the similar identifier. In order to construct the server / client of the question answering apparatus according to claim 1, claim 3 and claim 4,
The client includes the session management unit,
The server includes the answer processing control unit that operates based on the identifier selected by the session management unit, the units that are controlled by the answer processing control unit, and the document storage unit. Response system.

Images