JP2012043115A - Document search device, document search method, and document search program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a document search device capable of rapidly searching for documents about things currently in conversation according to situations.SOLUTION: A document search device comprises: a situation analysis part 1 which analyses situations such as a place or a date in input documents; a search index creation part 2 which creates a search index including the situations; a feature word extraction part 3 which extracts feature words from the documents in the situations according to distances from the situations; a feature word index creation part 4 which creates a feature word index for searching for the extracted feature words from a situation input in searching for documents; a situation input part 5 which inputs a situation specified by an user; a feature word presentation part 6 and a feature word input part 7 which select feature words about the input situation referring to the feature word index; and a search result acquisition part 8 which searches for documents about the input situation and the selected feature words referring to the search index.

Description

  The present invention relates to a technique for searching a document according to a situation including a designated place or date and time from a document set.

  Currently, there are many systems that search for documents related to a keyword when the keyword is input. In particular, in recent years, a method for searching for a document related to a keyword according to a specified location, date, and other situations has been proposed.

  In Patent Document 1, a document related to a keyword corresponding to a place is searched. In Patent Document 2, a document related to a keyword corresponding to a date and time is searched.

JP 2000-163436 A JP-A-2005-85109

  However, the methods disclosed in Patent Document 1 and Patent Document 2 have a problem in that when a document corresponding to a situation such as a place or a date / time is to be searched, an appropriate keyword must be designated as an input in addition to the situation.

  When you want to know what is being talked about in a certain situation, it is difficult for the user to enter the appropriate keywords to get the desired search results because the user does not know what is being talked about. It is thought that.

  In addition, by searching for documents related to the situation, extracting and selecting feature words from the retrieved documents, and presenting the documents related to the feature words and the situation, it is related to things that are topical according to the situation Although it is possible to search for a document, there is a problem that it takes time to perform a search because it takes time to extract a feature word from the searched document.

  The present invention has been made in view of the above-described problems, and creates a feature word index for acquiring a feature word from various situations in advance, and refers to the feature word index to obtain a feature word related to the situation. Document retrieval apparatus and method capable of retrieving documents related to featured items according to the situation at high speed by retrieving and selecting documents related to the feature word and situation The purpose is to provide a program.

  The above-described problem is a document search device that searches a document according to a situation that is considered as a classification axis for a specified search from a document set, and the situation analysis that analyzes the situation included in the document And a search index creating means for creating a search index including the situation analyzed by the situation analyzing means, and extracting a feature word from the document in each situation analyzed by the situation analyzing means based on the distance to the situation A feature word extraction means for creating a feature word index for searching for a feature word extracted by the feature word extraction means from a situation input at the time of document search, a specified search A situation input means for inputting the situation considered as a classification axis for the situation, and the situation input by the situation input means with reference to the feature word index A search result for searching for a document related to a feature word selected by the feature word selection unit and a feature word selection unit that selects a feature word related to the situation and the feature word selected by the feature word selection unit with reference to the search index This can be solved by providing the acquisition means.

  According to the first to seventh aspects of the present invention, a feature word included in a document related to an input situation is selected, and a situation-related document including the selected feature word is presented. It is possible to search documents related to things at high speed.

1 is a configuration diagram of a document search apparatus 100 according to an embodiment of the present invention. 6 is a flowchart showing an operation of preprocessing of the document search apparatus 100 according to the embodiment of the present invention. 6 is a flowchart illustrating an operation of document search processing of the document search apparatus 100 according to an embodiment of the present invention. Explanatory drawing which shows the example of the document in the Example of this invention. Explanatory drawing which shows the example of the analysis result of the place in the Example of this invention. Explanatory drawing which shows the example of the analysis result of the date in the Example of this invention. Explanatory drawing which shows the example of the transposition index in the Example of this invention. Explanatory drawing which shows the example of the status index in the Example of this invention. Explanatory drawing which shows the example which extracted the feature word candidate from the document containing "May 5, 2010" in the Example of this invention, and those documents. Explanatory drawing which shows the example which provided the score to the feature word candidate extracted from the document regarding "May 5, 2010" in the Example of this invention. Explanatory drawing which shows the example of the characteristic word index regarding the date in the Example of this invention. Explanatory drawing which shows the example of the characteristic word index regarding the place in the Example of this invention. Explanatory drawing which shows the example of the acquired characteristic word in the Example of this invention. Explanatory drawing which shows the example of the feature word presentation result in the Example of this invention. Explanatory drawing which shows the example of the search result in the Example of this invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings, but the present invention is not limited to the following embodiments.

  FIG. 1 is a diagram showing a configuration of a document search apparatus 100 that is an embodiment of the present invention.

  A document search apparatus 100 shown in FIG. 1 includes a situation analysis unit 1 as a situation analysis unit, a search index creation unit 2 as a search index creation unit, a feature word extraction unit 3 as a feature word extraction unit, and a feature word index. A feature word index creation unit 4 as a creation unit, a situation input unit 5 as a situation input unit, a feature word presentation unit 6 as a feature word presentation unit, a feature word input unit 7 as a feature word input unit, and a search A search result acquisition unit 8 as a result acquisition unit, a search index database 9, and a feature word index database 10 are provided.

  The document retrieval apparatus 100 is configured by, for example, a computer, and is a normal computer hardware resource, for example, an input device such as a ROM, a RAM, a CPU, and a keyboard, an output device such as a display display and a printer, a communication interface, a hard disk, and a recording medium A driving device or the like (not shown) is provided.

  As a result of the cooperation between the hardware resource and the software resource (OS, application, etc.), the document search apparatus 100 has a situation analysis unit 1, a search index creation unit 2, a feature word extraction unit 3 as shown in FIG. , A feature word index creation unit 4, a situation input unit 5, a feature word presentation unit 6, a feature word input unit 7, a search result acquisition unit 8, a search index database 9, and a feature word index database 10 are implemented.

  The search index database 9 and the feature word index database 10 are constructed in storage means / storage means such as a hard disk or RAM.

  The situation analysis unit 1 analyzes a situation such as a place or date / time considered as a classification axis for retrieval, which is included in a document input to the document retrieval apparatus 100 in a preprocessing stage described later.

  The search index creation unit 2 creates a search index including the situation analyzed by the situation analysis unit 1 and stores it in the search index database 9.

  The feature word extraction unit 3 extracts a feature word from the document in each situation analyzed by the situation analysis unit 1 based on the distance from the situation.

  The feature word index creation unit 4 creates a feature word index for searching for the extracted feature word from the situation input in the document search stage described later, and stores it in the feature word index database 10.

  The situation input unit 5 inputs the situation from the user at the document search stage.

  The feature word presentation unit 6 refers to the feature word index database 10 to acquire and present the feature words related to the input situation.

  The feature word input unit 7 inputs, for example, one or more feature words selected by the user from the feature words acquired by the feature word presentation unit 6.

  The search result acquisition unit 8 refers to the search index database 9 to search for a document related to the situation input from the situation input unit 5 and the feature word input from the feature word input unit 7.

  Next, the operation of the document search apparatus 100 will be described.

  The present invention is divided into two stages: a pre-processing stage for creating a search index for searching a document and a feature word index, and a stage for actually searching for a document.

  FIG. 2 is a flowchart showing the preprocessing operation of the document search apparatus 100.

  First, the situation analysis unit 1 analyzes the situation included in the input document (step S101).

  The search index creation unit 2 creates a search index including the situation and stores it in the search index database 9 (step S102).

  The feature word extraction unit 3 extracts a feature word from the document in each situation based on the distance from the situation (step S103).

  The feature word index creation unit 4 creates a feature word index for searching for a feature word from the situation, and stores it in the feature word index database 10 (step S104).

  FIG. 3 is a flowchart showing the document search processing operation of the document search apparatus 100.

  First, a user who performs a document search inputs a situation using the situation input unit 5 (step S201).

  The feature word presentation unit 6 acquires the feature words related to the situation input with reference to the feature word index database 10, and displays the feature words on the display, for example, to present to the user (step S202).

  The feature word input unit 7 inputs one or more feature words selected by the user from the acquired and presented feature words (step S203).

  The search result acquisition unit 8 refers to the search index database 9 and searches for a document related to the feature word selected by the user and the feature word input unit 7 selected by the situation input unit 5 (step). S204).

  Next, the operation of the document search apparatus 100 will be described more specifically. In this example, two places, a place and a date / time, are considered as situations. The situation to be considered is not limited to this, and various situations such as the weather may be considered as long as they represent the situation.

  The pre-processing operation will be specifically described with reference to the flowchart shown in FIG.

  In step S101, the situation analysis unit 1 analyzes the situation included in the input document. An example of a document is shown in FIG. Here, as a method of analyzing a place included in a document, a position information dictionary in which the document is morphologically analyzed and divided into words, a unique name is extracted, a place name is extracted, and position information is associated with the place name. The position information is given with reference to FIG. The place analysis method is not limited to this, and a character string representing a place name may be searched from a document, or the place name may be represented as a place without giving position information to the place name. An example of the location analysis result is shown in FIG.

  In addition, here, as a method of analyzing the date and time contained in the document, the document is morphologically analyzed and divided into words, the specific expression is extracted to extract the date and time expression, and the date and time is obtained from the pattern of the date and time expression. And The date and time analysis method is not limited to this, and if the document matches a string pattern such as “… Month… Day”, the relevant part is expressed as a date, or a word such as “Today” is added to the document. It may be converted to the actual date based on the creation date. An example of the date and time analysis result is shown in FIG.

  In step S <b> 102, the search index creation unit 2 creates a search index including a situation and stores it in the search index database 9. Here, as a search index including a situation, a transposed index for performing a keyword search and a situation index for performing a search regarding the situation are created.

  In creating a transposed index, a transposed index creating method generally used in document retrieval is used. An example of the transposed index is shown in FIG. In this figure, (1, 1) for the word “ABC” means that the first word of the document text of document number 1 was “ABC”.

  In creating the status index, the status analyzed by the status analysis unit 1 is assigned to the document. An example of the status index is shown in FIG. The search index creation method is not limited to this, and an N-gram based index may be created instead of the transposed index.

  In step S103, the feature word extraction unit 3 extracts feature words from the document in each situation based on the distance from the situation. Here, for each situation of location and date and time, the situation is divided into a plurality of units, a document relating to each situation is acquired, and feature words are extracted from the document.

  As for the location, the target area is divided into 1 km × 1 km meshes, and feature words are extracted from documents relating to the locations included in each mesh. As for the date and time, the target date and time is divided every day, and feature words are extracted from documents related to the date and time included in each day. The method of determining each situation is not limited to this, and the size of the mesh may be changed, or the date and time may be further divided into morning and afternoon.

  Also, here, the names of names, places, and organizations are extracted as feature word candidates from the entire document in each situation, and a score is assigned to the feature word candidate based on the distance between the situation included in the document in each situation and the feature word candidate. The feature words are extracted based on the score of the feature word candidate.

  The method of extracting feature words is not limited to this, as long as it is based on the distance from the situation included in the document, and a specific expression other than a person name, place name, or organization name can be used. Can be extracted as feature word candidates, general nouns other than proper expressions can be extracted as feature word candidates, those with parts of speech other than nouns can be extracted as feature word candidates, and frequently appearing character strings in each situation Extract as feature word candidates, consider the appearance frequency of feature word candidates included in documents in each situation, consider the appearance frequency of feature word candidates included in documents in each situation in other situations, Character strings and word strings common to a plurality of document titles may be extracted as feature words, or all feature word candidates may be used as feature words regardless of the score.

  As an example of the date and time, FIG. 9 shows an example in which feature word candidates are extracted from documents including “May 5, 2010” and those documents. Here, the distance between the situation included in the document in each situation and the feature word candidate is a value obtained by adding 1 to the number of characters in the character string sandwiched between the character string representing the situation and the character string of the feature word candidate. The reciprocal is used, and the sum of distances in the document in which the feature word appears is used as the score of the feature word candidate. The method for obtaining the distance is not limited to this, and the maximum value of the distance may be used as a score.

  FIG. 10 shows the result of obtaining the score for each feature word candidate in FIG. A feature word is extracted based on the score of the feature word candidate. Here, the top three feature word candidates with high feature word scores are extracted as feature words. The method of extracting feature words is not limited to this, and the number of feature words to be extracted may be changed, or feature words of the top X% may be extracted. From FIG. 10, “ABC”, “XYZ”, and “DEF” are extracted as feature words.

  In step S <b> 104, the feature word index creation unit 4 creates a feature word index for searching for a feature word from the situation, and stores it in the feature word index database 10. FIG. 11 shows the result of creating the feature word index related to the date and time. The feature word index is created in the same manner for the place. FIG. 12 shows the result of creating the feature word index related to the place.

  Next, the operation of the document search process will be specifically described according to the flowchart shown in FIG. In this example, the user who performs the document search exists at a position represented by “35 ° 37′N 139 ° 53′E”, and the date and time when the document search is performed is “2010/5/5 (Wednesday)”. It shall be.

  In step S <b> 201, the situation input unit 5 inputs a situation. Here, it is assumed that the location “35 ° 37′N 139 ° 53′E” and the date and time “2010/5/5 (Wednesday)”, which are the situations where the user is currently placed, are input. The method of inputting the situation is not limited to this, and it is possible to enter other than the current location, enter the date other than the current location, enter the location name instead of location information,・ You may specify detailed date and time up to minutes and seconds.

  In step S <b> 202, the feature word presenting unit 6 acquires and presents a feature word related to the situation input with reference to the feature word index database 10. Here, a feature word corresponding to a date and time that matches the input date and time is searched, and a feature word corresponding to a mesh including the input place is searched.

  The method of searching for feature words is not limited to this, and it searches for feature words corresponding to the date / time range including the input date / time, or corresponds to the mesh range centered on the mesh including the input location. You can search for feature words. Since the input date / time coincides with the date / time “May 5, 2010” in FIG. 9, the corresponding feature word is searched. Similarly, since the input place is included in the mesh of “35 ° 30′N 139 ° 45′E to 35 ° 60′N139 ° 90′E” in FIG. 12, the feature word corresponding to this mesh is searched. Is done. Based on the searched feature word, the feature word related to the input situation is acquired.

  Here, the product of the scores of the feature words searched for the place and date / time is calculated, and those having a value of 0.1 or more are acquired as feature words. The method of acquiring feature words is not limited to this. The sum of scores is calculated, all the feature words searched regardless of the score, the threshold value of the score is changed, the higher score X% feature words may be acquired. An example of the result of acquiring the feature word is shown in FIG. The acquired feature word is displayed on a display display (not shown), for example. Here, the top two feature words are presented in descending order of the feature word score.

  The feature word presentation method is not limited to this. The top N feature words, the top X% feature words, the feature word score value of Y or more is presented. Or present one feature word for each type of person name, etc., present a feature word score together with the feature word, or present it by changing the font size according to the value of the feature word score. It doesn't matter. An example of the result of presenting the feature word is shown in FIG.

  In step S203, the feature word input unit 7 inputs one or more feature words selected by the user from the acquired and presented feature words. Here, “XYZ” is selected as the feature word. The method of selecting feature words is not limited to this, and a plurality of feature words may be selected.

  In step S204, the search result acquisition unit 8 refers to the search index database 9 to search for a document related to the input situation and the selected feature word. Here, for example, referring to the transposed index of FIG. 7, a document including a feature word is searched using a method generally used in the conventional document search, and, for example, a diagram including, for example, FIG. Referring to the status index of FIG. 8, the selected feature word “XYZ” has a date and time that matches the input date and time and within a certain distance from the input location, and the selected feature word “XYZ” is the title or text of the document. Let's search for the documents contained within.

  The way to search for documents is not limited to this. Search for documents that have a date and time that matches part of the input date and time, or if the input location is a character string that represents a place name, You can also search for documents included in the text. Further, the score for a document is assumed to be a fitness obtained by searching a document including a feature word. The method for calculating the score for a document is not limited to this. Consider the proximity of the feature word in the document to the date and time, or the proximity of the feature word to the location in the document. It doesn't matter. An example of the retrieved document is shown in FIG.

  In this embodiment, the feature word presenting unit 6 presents the feature word to the user, and the user selects one or more of the presented feature words and inputs the feature word through the feature word input unit 7. The feature word may be automatically selected according to the value of the feature word score calculated by the feature word presentation unit 6 without presenting the feature word to the user.

  As described above, according to the situation, the feature word included in the document related to the input situation is selected by the process described in the present embodiment, and the situation related document including the selected feature word is presented. You can search for documents related to the topic.

  The document search apparatus 100 according to the present embodiment can be realized by installing a program for executing the above-described series of operations in a computer. It is also possible to store the program in a portable storage medium such as a hard disk device connected to a computer used as a document retrieval device, a flexible disk, or a CD-ROM, and install and execute the program on a computer. is there. It is also possible to store the program in a network server and download and install it from there.

  The present invention is not limited to the above-described embodiments and examples, and various modifications and applications are possible.

  The present invention can be used for development of a search system that performs a search in consideration of a situation such as a place and a date.

DESCRIPTION OF SYMBOLS 1 ... Situation analysis part 2 ... Search index creation part 3 ... Feature word extraction part 4 ... Feature word index creation part 5 ... Situation input part 6 ... Feature word presentation part 7 ... Feature word input part 8 ... Search result acquisition part 9 ... Search Index database 10 ... feature word index database 100 ... document retrieval device

Claims (7)

A document search device for searching a document according to a situation considered as a classification axis for a specified search from a document set,
Situation analysis means for analyzing the situation included in the document;
Search index creating means for creating a search index including the situation analyzed by the situation analyzing means;
Feature word extraction means for extracting a feature word from a document in each situation analyzed by the situation analysis means based on a distance from the situation;
A feature word index creating means for creating a feature word index for retrieving a feature word extracted by the feature word extracting means from a situation input at the time of document search;
A status input means for inputting the status considered as a classification axis for the specified search;
Referring to the feature word index, feature word selection means for selecting a feature word related to the situation input by the situation input means;
Search result acquisition means for searching for a document related to the situation input by the situation input means and the feature word selected by the feature word selection means with reference to the search index;
A document search apparatus characterized by comprising: The feature word extraction means includes
A feature word candidate is extracted from the document, a score is given to the feature word candidate based on a distance between the situation included in the document and the feature word candidate, and the feature word is extracted based on the score of the feature word candidate The document search apparatus according to claim 1. The feature word selection means includes
Referring to the feature word index, a feature word presenting means for acquiring and presenting a feature word related to the input situation;
Feature word input means for inputting one or more feature words from the feature words acquired by the feature word presenting means;
The document search device according to claim 1, comprising: A document search method for searching a document according to a situation considered as a classification axis for a specified search from a document set,
A situation analysis step in which situation analysis means analyzes the situation included in the document;
A search index creating step for creating a search index including a situation analyzed by the situation analyzing means;
A feature word extracting unit that extracts a feature word based on a distance from the situation from a document in each situation analyzed by the situation analyzing unit;
A feature word index creating step for creating a feature word index for searching for a feature word extracted by the feature word extracting unit from a situation input at the time of document search;
A situation input step in which the situation input means inputs the situation considered as a classification axis for the designated search;
A feature word selection unit that selects a feature word related to the situation input by the situation input unit with reference to the feature word index; and
A search result acquisition step for searching for a document related to the feature word selected by the situation input by the situation input unit and the feature word selection unit by referring to the search index;
A document search method characterized by comprising: The feature word extraction step includes:
A feature word candidate is extracted from the document, a score is given to the feature word candidate based on a distance between the situation included in the document and the feature word candidate, and the feature word is extracted based on the score of the feature word candidate The document retrieval method according to claim 4, wherein: The feature word selection step includes:
A feature word presenting means refers to the feature word index, acquires and presents a feature word related to the input situation, and a feature word presenting step,
A feature word input means for inputting one or more feature words from the feature words acquired by the feature word presentation means; and
The document search method according to claim 4, wherein:   A document search program for causing a computer to execute each step in the document search method according to claim 4.

Images