JP2006323670A - Answering method, answering device and program for questioning method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an answering method, an answering device and a program for questioning method for intelligibly presenting more explanatory documents to a method questioned which asks for the explanation of a method in the answering method of presenting a document of a retrieval result as a document which explains the method to a question sentence of the method when the question sentence asking for the method is given. <P>SOLUTION: The answering method for questioning method has a discriminating step for an explanatory document of a method to discriminate whether each retrieved document is the explanatory document of the method based on a explanatory document characterizing reference which is the reference for characterizing the explanatory document, and storing the discriminated result in a storage device, and a presenting step of a retrieval step changing the display method of a retrieval result list when the retrieved document is discriminated as the explanatory document for the method, and storing it in the storage device and presenting it. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

The present invention generally relates to a question answering method for outputting an answer to a question sentence, and in particular, given a question sentence asking for a method, a document as a method explanation document, which is a document including a method explanation for the method question sentence. The present invention relates to a method question answering method, a method question answering apparatus and a program for presenting search results.

  The conventional question answering technology obtains a search key from a given question sentence, uses the search engine to extract related documents from a large number of documents based on the obtained search key, and further answers the question sentence. This is a technique for extracting from the extracted related document.

  For example, a technique is known in which a question type is determined from a question sentence, a specific expression or a numerical expression is extracted from a related document, an appropriate expression that matches the question type is used as an answer, and extracted (for example, , See Patent Document 1).

In addition, with respect to a question sentence expressed in a natural sentence, a document in which the word n-gram included in the question sentence frequently appears is re-ranked to the top, and the unique / A technique for extracting a numerical expression from a higher-level document is known (for example, see Patent Document 2).
Japanese Patent Laid-Open No. 2002-132911 JP 2004-355550 A

  However, the above-mentioned conventional question-and-answer techniques are intended for responding to a named entity or numerical expression that can be expressed in short words, such as a person name, place name, quantity, etc. The reason, method, procedure, etc. Cannot respond to questions that require descriptive answers.

  For example, “Who is the President of XX?”, “Where is the Olympic Games in 20xx?”, “How many meters is the tower height?” Questions that can be answered are the subject of the prior art.

  That is, the conventional question answering technique does not target a question sentence for which explanation is requested. For example, in the conventional example, questions such as "Why is the sea blue?" And "How can I make delicious curry?" That are not subject to.

  On the other hand, there is known a method for searching and solving an existing collection of Q & A for a question that requires explanation. That is, this solution is a method of searching a document for an existing Q & A collection, searching for a QA pair that matches the question sentence, and obtaining a desired answer.

  However, the publicly available collection of Q & A is specialized in a specific field such as computer, medical care, law, etc. Also, because there is a small amount of Q & A collection available, find the desired explanation answer to the question. There is a problem that is often difficult.

The present invention seeks explanation of the method in a method question response method for presenting a document search result as a document explaining a method for the method question sentence when a method question sentence that is a question asking a method is given. It is an object of the present invention to provide a method question answering method, a method question answering apparatus, and a program capable of presenting more method explanation documents in an easy-to-understand manner for method questions.

The present invention provides a method question answering method that presents a document search result as a method explanation document that is a document explaining a method for the method question sentence when a method question sentence that is a question asking a method is given. A method explanation document discrimination step for discriminating whether or not each search document is a method explanation document based on a method explanation document characterization criterion that is a criterion for characterizing an explanation document, and storing the discrimination result in a storage device; When it is determined that the search document is a method explanation document, the search result list includes a search result presentation step of changing the display method of the search result list, storing it in a storage device, and presenting it.

According to the present invention, in a method question answering method for presenting a document search result as a document explaining a method for the method question sentence when a method question sentence which is a question asking a method is given, the description of the method As a result, it is possible to present more method explanation documents in an easy-to-understand manner.

  The best mode for carrying out the invention is the following examples.

  FIG. 1 is a block diagram showing a basic configuration of a method question answering apparatus 10 that is Embodiment 1 of the present invention.

  The method question answering device 10 is a document including a method description for a question by combining a search engine that searches all types of documents with respect to a method question for which a method description is requested, and a method description document discrimination process. This is an embodiment in which a document search result is presented as a method explanation document.

[Constitution]
The method question answering apparatus 10 includes a question analysis unit 11, a document search unit 12, a method explanation document determination unit 13, and a search result presentation unit 14.

  The question analysis unit 11 extracts a search word set from the inputted question sentence and stores it in a storage device (not shown).

  The document search unit 12 performs a document search using the extracted set of search terms as a search key, obtains a search result list obtained as a result, and individual search documents indicated in the search result list, and stores them in a storage device. To remember.

  The method explanation document discriminating unit 13 discriminates whether or not the search document acquired by the document search unit 12 is a document explaining the method based on the method explanation document criteria for characterizing the method explanation document. A method explanation document discrimination for storing the result in the storage device is performed.

  If the search document explains the method, the search result presentation unit 14 changes the display method, stores the search result list acquired by the document search unit 12 in the storage device, and presents it.

[Operation]
Next, the operation of the method question answering apparatus 10 will be described.

  FIG. 2 is a flowchart showing a schematic operation of the method question answering apparatus 10.

  In S1, the question analysis part 11 inputs the question sentence from a user, and memorize | stores it in a memory | storage device. In S2, the question analysis unit 11 extracts a search term set from the question sentence acquired in S1. In the present specification and claims, the “search word set” is composed of at least one search word.

  In S3, the document search unit 12 inputs the search term set extracted in S2 to the document search engine, performs a document search, obtains a search result list as a search result, and stores it in the storage device.

  In S4, all the individual search documents shown in the search result list obtained by the document search unit 12 are acquired by tracing the link URL, file path, etc. shown in the search result list.

  In S5, the method explanation document determination unit 13 determines whether or not the search document is a document explaining the method based on the method explanation document characterization criteria for all the search documents acquired in S4. And the determination result is stored in the storage device. The “method explanation document characterization standard” is a standard for characterizing a method explanation document that is a document explaining a method.

  In S6, the search result presentation unit 14 changes the display method of the portion corresponding to the search document determined to have explained the method in S5 with respect to the search result list obtained in S3, and the changed The search result list is stored in the storage device, presented to the user, and the process ends.

  In S5, when it is determined whether or not one search document is a method description document, the method description document determination unit 13 performs the following operation for searching for a bulleted portion.

  FIG. 3 is a flowchart showing an operation in which the method explanation document discriminating unit 13 detects the itemized portion.

  In S11, the location where the search term extracted in S2 appears in the search document is specified.

  In S12, an itemized list in which a plurality of identical symbols or ascending numerical strings are arranged in a predetermined range behind the appearance of the search term specified in S11 with a character string of a predetermined length or more in between. Check if the part appears.

  In S13, if a bulleted item is found in S12, it is determined that the currently searched search document is a method explanation document. On the other hand, if none is found, it is determined that the document is not a method explanation document.

  In S12, a plurality of the same symbols or ascending numerical strings are arranged in a predetermined range ahead of the appearance of the search term specified in S11, with a character string having a predetermined length or more in between. You may make it investigate whether the bulleted part appears.

  That is, when a method explanation document is determined, the search document is determined to be a method description document by determining that a bulleted portion has appeared in a predetermined range centering on the search term that appears in the search document. It is determined that The predetermined range centering on the search word that appears is a predetermined range before the search word that appears or a predetermined range after the search word that appears. .

  FIG. 4 is a flowchart illustrating an operation in which the method explanation document determination unit 13 detects a portion that matches the method explanation document feature pattern in the first embodiment.

  In the description of the first embodiment, the “method explanation document feature pattern” is a word pattern that characterizes a predetermined method explanation document.

  In S21, the location where the search word extracted in S2 appears in the search document is specified and stored in the storage device.

  In S22, it is checked whether or not a portion that matches the method description document feature pattern appears in a predetermined range behind the appearance of the search term specified in S21.

  If a part (same part) matching the method explanation document feature pattern is found (S22), it is determined in S23 that the search document is a document explaining the method (method explanation document). If no part that matches the method description document feature pattern is found, it is determined that the method description document is not a method description document.

[Concrete example]
Next, a more specific operation regarding the first embodiment will be described.

  In the following, the method question answering apparatus 10 responds to the question sentence “How to change the name of a car” according to the flowcharts of FIGS. 2, 3, and 4. The operation of extracting and presenting will be described.

  First, the question analysis unit 11 inputs a question sentence “How to change the name of a car?” To the method question answering apparatus 10 and stores it in the storage device (S1).

  Next, the question analysis unit 11 extracts one or more search terms from the inputted question sentence (S2). In the first embodiment, when extracting a search word, a noun obtained by morphological analysis of the question sentence is regarded as a search word. In other words, morphological analysis of "How to change the name of a car?" And the nouns obtained are "car" and "name change", and the nouns "car" and "name change" are extracted as search terms. To do. When extracting a search term, use a method of searching for a verb stem as a search term in addition to a noun, or a method of searching for a word that matches a dictionary prepared in advance without using morphological analysis, etc. You may do it. Here, a method for extracting a search term is not defined.

  Next, the document search unit 12 inputs the search term string “car name change” obtained by the question analysis unit 11 into the document search engine, and searches the document.

  Further, a search result list as the search result is acquired and stored in the storage device (S3).

  FIG. 5 is a diagram illustrating an example of a search result list that is a result screen searched by the document search engine that has input the search term set “car name change” in the first embodiment.

  Each item in the search result list shown in FIG. 5 includes a search order (for example, “6.”), a search document title (for example, “car name change (transfer registration)”), a search document, a URL (for example, “ http://hhh.iii.com/abcdef/ ”).

  The document search engine used by the document search unit 12 may be a commonly used Web search engine such as Google or Google, or a search engine specialized in Q & A collections such as OKWeb or Hatena, and is closed to an intranet, a PC local, or the like. It may be a document search engine. In other words, any search engine may be used as long as it is a search engine that inputs a search keyword and returns a list of documents related to the input search keyword.

  Further, the document search unit 12 acquires all the individual search documents shown in the search result list by tracing the search document URLs in the list, and stores them in the storage device (S4). In FIG. 5, for the sixth item, for example, the document indicated by the URL “http://hhh.iii.com/abcdef/” is actually acquired and stored in the storage device.

  FIG. 6 is a diagram showing an example of the contents of the third search document included in the search result list shown in FIG.

  In actual processing, document correction processing is performed such as conversion to plain text or normalization of character codes, excluding HTML tags, but this is not the essence of the above embodiment. The detailed explanation is omitted.

  Next, the method explanation document discriminating unit 13 explains the method for all the search documents acquired by the document search unit 12 based on the method explanation document characterization criteria for characterizing the method explanation document. It is determined whether or not the document is a document (S5).

  In the first embodiment, if the condition of either the itemized list or the method explanation document feature pattern appears in a predetermined range after the search word, it is determined as the method explanation document and stored in the storage device. Only one of the two conditions may be used to determine that the document is a method description document.

  Next, a description will be given of a specific operation in which the method explanation document determination unit 13 detects itemized items for the search document shown in FIG. 6 among the stored search documents.

  First, in the search document shown in FIG. 6, the part where the search terms “car” and “name change” appear is specified and stored in the storage device (S11).

  In FIG. 6, the search terms “car” and “name change” appearing are shown in bold.

  Next, with respect to a search word appearing in the search document, a plurality of identical symbols or ascending numerical strings are not less than b characters having a predetermined length within a character that is a predetermined range behind the appearance location. It is checked whether or not the lined items appearing across the character string appear (S12).

  Here, the predetermined threshold values such as a and b are experimentally determined and set, and no specific numerical values are defined here. In the first embodiment, for example, a = 300 and b = 30.

  As a result, in the search document shown in FIG. 6, the underline “1”, “2”, “3”, “4” within the last 300 characters from the search term “name change” appearing in the first line. An ascending number string (item number part of the bulleted list) in which “” is drawn can be used to detect the itemized part where the character string (item content part of the bulleted item) of 30 characters or more is sandwiched.

  Therefore, since it is found that the itemized portion exists in the search document shown in FIG. 6 (S12), it is determined that the search document shown in FIG. 6 is a document explaining the method, and the storage device (S13).

  On the other hand, FIG. 7 is a diagram showing an example of the content of the sixth search document in the search result list shown in FIG.

  Among the saved search documents, the search document shown in FIG. 7 cannot be detected as a bulleted list by the same process, and therefore, at this stage, it is determined that the document is not a document explaining the method. . After that, if a predetermined condition is satisfied, it is determined that the document explains the method.

  Next, a specific operation when the method explanation document determination unit 13 detects a portion matching the method explanation document feature pattern for the search document shown in FIG. 7 will be described.

  First, in the search document shown in FIG. 7, the part where the search terms “car” and “name change” appear is specified and stored in the storage device (S21).

  In FIG. 7, the search terms “car” and “name change” appearing are shown in bold.

  Next, with respect to the search word appearing in the search document, it is checked whether or not a portion matching the method explanation document feature pattern has appeared in the c character that is a predetermined range behind the appearance location (S22).

  Here, the above-mentioned “method explanation document feature pattern” is a pattern of words characterizing the method explanation document, and is an experimentally predetermined pattern, for example, “how”, “how”, “procedure”, “procedure”. And so on. In the first embodiment, it is assumed that the four words exemplified above are method description document feature patterns. In addition, as the method description document feature pattern, <verb union form> + “method” may be set with a part of speech or a condition based on utilization, or a word having a meaning attribute of “procedure”, etc. You may make it set with the conditions by the meaning of a word. Method Description Document feature pattern setting method is not the essence of the above embodiment, and is not defined here.

  Further, the threshold value determined in advance for c is determined experimentally and is not defined here.

  In the first embodiment, for example, c = 10. Accordingly, in the search document shown in FIG. 7, there is a character underlined “how” within 10 characters behind the search word “name change” appearing in the first line. This means that a part that matches one of the feature patterns could be detected. The “procedure” can also be detected in the “car procedure” part of the middle part of the search document.

  Therefore, since a portion that matches the method description document feature pattern was found in S22 for the search document shown in FIG. 7, it is determined that the search document shown in FIG. 7 is a document explaining the method, and the storage device (S23).

  In the same manner as described above, the method explanation document determination unit 13 determines whether or not other search documents are documents explaining the method.

  In the first embodiment, among the search documents described in the search result list shown in FIG. 5, the third and sixth documents are determined to be method explanation documents as described above, and the others are Otherwise it is determined.

  Finally, the search result presentation unit 14 changes the display method of the search result list shown in FIG. 5 to a corresponding part of the search document that the method explanation document determination unit 13 determines to explain the method, The search result list in which the display method is changed is stored in the storage device, presented to the user of the method question answering apparatus 10, and the process ends (S6).

  In the first embodiment, the search character string in the method explanation document is displayed in bold and the display method in the search result list is changed.

  FIG. 8 is a diagram illustrating an example of a screen presented by the search result presentation unit 14 after the display method in the search result list is changed by the above operation.

  In the screen shown in FIG. 8, since the method explanation document determination unit 13 determines that the third search document and the sixth search document are method description documents, the display of those items is bold. has been edited.

  As a method of changing the display, a method of re-ranking and displaying so that the item of the method explanation document is in a higher rank may be adopted, and only the search document determined as the method explanation document is displayed. You may make it employ | adopt a method.

  According to the first embodiment, a method explanation document that matches an input question sentence is obtained by combining a search engine that searches all types of documents with a method explanation document determination process for a method question for which a method explanation is requested. Can be distinguished from other documents and provided to the user.

  In addition, according to the above-described embodiment, since a wording and document style peculiar to a method description appearing in the vicinity of a search keyword in a search result document are detected, a description of a predetermined method can be provided without preparing a special Q & A collection. A method explanation document that matches the question sentence can be provided to the user separately from the other documents for the method question to be sought.

In addition, the said Example can be grasped | ascertained as invention of a program, that is, the said Example is a document explaining the method with respect to the said method question sentence, when the method question sentence which is a question which asks a method is given. In the program for presenting the document retrieval result, the method explanation document discrimination for discriminating whether each search document is a method explanation document based on the method explanation document characterization criteria for characterizing the method explanation document and storing it in the storage device A program for causing a computer to execute a procedure and a search result presenting procedure for storing and presenting in a storage device by changing a display method of the search result list when it is determined that the search document is a method explanation document It is an example.

It is a block diagram which shows the basic composition of the method question answering apparatus 10 which is Example 1 of this invention. 4 is a flowchart showing a schematic operation of the method question answering apparatus 10. It is a flowchart which shows the operation | movement in which the method description document discrimination | determination part 13 detects an itemized part. 6 is a flowchart illustrating an operation in which the method explanation document determination unit 13 detects a portion that matches a method explanation document feature pattern in the first embodiment. In Example 1, it is a figure which shows an example of the search result list | wrist which is a result screen searched by the document search engine which input search word set "car name change". It is a figure which shows the example of the content of the 3rd search document contained in the search result list | wrist shown in FIG. It is a figure which shows the example of the content of the 6th search document of the search result list | wrist shown in FIG. It is a figure which shows an example of the screen which the search result presentation part 14 has shown, after changing the display method in a search result list by the said operation.

Explanation of symbols

10 ... Method question answering device,
11 ... Question analysis unit,
12 ... Document search section,
13 ... Method explanation document discrimination section,
14: Search result presentation unit.

Claims (10)

In a method question answering method for presenting a document search result as a method explanation document that is a document explaining a method for the method question sentence when a method question sentence that is a question asking a method is given,
A method explanation document discriminating step for discriminating whether or not each search document is a method explanation document based on a method explanation document characterization criterion which is a criterion for characterizing the method explanation document; and storing the discrimination result in a storage device;
A search result presentation step of changing the display method of the search result list, storing it in a storage device, and presenting it when it is determined that the search document is a method explanation document;
A method for answering questions, comprising: In a method question answering method for presenting a document search result as a method explanation document that is a document explaining a method for the method question sentence when a method question sentence that is a question asking a method is given,
A question analysis step of analyzing a question by extracting a set of search terms composed of at least one search term from the inputted method question sentence, and storing the analysis result in a storage device;
Document search using the extracted search term set as a search key, and storing the search result list obtained as a result of the document search and the individual search documents included in the search result list in a storage device A search step;
A method explanation document discrimination step for discriminating whether each search document is a method explanation document based on a method explanation document characterization criterion, which is a criterion for characterizing the method explanation document, and storing the discrimination result in a storage device; ;
A search result presentation step of changing the display method of the search result list, storing it in a storage device, and presenting it when it is determined that the search document is a method explanation document;
A method for answering questions, comprising: In claim 1 or claim 2,
In the method description document determining step, it is determined that a bulleted portion has appeared in a predetermined range centering on the search term that appears in the search document, whereby the search document explains the method. A method for answering questions, which is a step of discriminating that the document is a document. In claim 3,
In the method description document determining step, a plurality of identical symbols or ascending numerical strings are character strings having a predetermined length or more in a predetermined range centering on the search word appearing in the search document. A method question answering method characterized in that it is a step of detecting a part lined up and sandwiched as being the above-mentioned itemized part. In claim 1 or claim 2,
The method explanation document discriminating step includes a part of the search document that matches a method explanation document feature pattern that is a pattern of a word characterizing a predetermined method explanation document in a predetermined range centered on the search word that appears. A method question answering method characterized in that it is a step of discriminating that the document is a method explanation document when it is detected that the character appears. In any one of Claim 1, Claim 2, and Claim 5,
The method explanation document determining step is a method explanation document that is a pattern of words characterizing the itemized list or the predetermined method explanation document in a predetermined range centering on the search word that appears in the search document. A method question answering method comprising a step of discriminating that a part that matches the feature pattern is a method explanation document. In any one of Claims 3-6,
The predetermined range centering on the search word that appears is a predetermined range before the search word that appears or a predetermined range after the search word that appears. A method characterized by question answering. In any one of Claims 1-7,
The search result presenting step is a step of displaying only the search document explaining the method as the search result list. In a method question answering apparatus that presents a document search result as a method explanation document that is a document explaining a method for the method question sentence when a method question sentence that is a question asking a method is given.
A method explanation document discriminating means for discriminating whether or not each search document is a method explanation document based on a method explanation document characterization criterion which is a criterion for characterizing the method explanation document; and storing the discrimination result in a storage device;
Search result presentation means for changing the display method of the search result list, storing it in a storage device and presenting it when it is determined that the search document is a method explanation document;
A method question answering apparatus comprising: When a method question sentence that is a question asking a method is given, in a program that presents a document search result as a method explanation document that is a document explaining a method for the method question sentence,
A method explanation document discriminating procedure for discriminating whether or not each search document is a method explanation document based on a method explanation document characterization criterion which is a criterion for characterizing the method explanation document; and storing the discrimination result in a storage device;
A search result presentation procedure for changing the display method of the search result list, storing it in the storage device, and presenting it when it is determined that the search document is a method explanation document;
A program that causes a computer to execute.

Images