JP2005010880A - Information search method, information search program, and information search apparatus - Google Patents

Abstract

[PROBLEMS] To correct a search condition using a semantic system, a classification system, geographical knowledge, etc., present an appropriate search result, and facilitate search of information related to the search result.
A step of retrieving a search condition including a relationship between keywords from an inputted question in a natural language format, a document on the Internet corresponding to each keyword included in the search condition is acquired, and based on the relationship between keywords The search result of the tree structure is acquired by linking the documents (201). It is determined whether the number of search results corresponding to each keyword is zero or does not exceed a set threshold (202, 203). A step of widening a limited range of search conditions (204), and if the number of search results exceeds a threshold, presenting a more specific keyword according to the semantic system to which the keyword belongs, and asking the user to select from among them (205).
[Selection] Figure 4

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an information search method, an information search program, and an information search device for searching for documents on the Internet. Specifically, when finding the search results for the search conditions entered by the user, the search conditions are corrected using background knowledge such as semantic system, classification system, geographical knowledge, etc. according to the number of results, and more appropriate search It presents the results to the user and facilitates a derivative search to information related to the requested search results.
[0002]
[Prior art]
Currently, a large amount of information is transmitted on the Internet, and information about real-world things such as street corner shops and tourist spots is now easily available via the network.
Such information can be obtained by designating a URL (Uniform Resource Locator) indicating the location on the Internet, but in reality, Google (http://www.google.com) or goo (http: //Www.goo.ne.jp) and the like, it is considered that the user is accessing the desired information using a search service by keyword specification.
[0003]
In these services, an input keyword is mechanically evaluated, and all documents including the keyword are returned as search results.
The search results are sorted from various viewpoints, and the documents required by the user are ranked higher.
In this method, there is a problem that an effective search result cannot be obtained unless an appropriate keyword is input during the search.
[0004]
To solve this problem, Digital City Kyoto and Georing Kyoto have been studied as follows.
1. As a method of collecting and providing coordinated web pages, Patent Document 1 discloses a method of collecting web pages from the Internet, a method of creating a database for searching from collected web pages, and a network connection. It consists of search processing methods from multiple clients.
First, Web pages are recursively collected from the Internet using a Web robot, and are selected according to the described information.
[0005]
Then, the HTML tag is deleted from the selected Web page, a proper noun such as an address or name is extracted by morphological analysis, and the geographical coordinates corresponding to the proper noun are obtained from the geographical information of the target area.
The geographic information of this Web page is registered in the search database together with the URL, and search processing from a plurality of terminals connected to the computer network is realized.
[0006]
2. As a method for receiving map information and information related to the map from a computer network and displaying them superimposed on a terminal screen, the one disclosed in Patent Document 2 realizes the method for collecting and providing a web page with coordinates. It becomes the interface of the device to do.
The method described above has improved the method of searching for documents on the Internet by extracting meta information of the page from information described in the Web page and organizing it into a database.
[0007]
However, active search assistance that modifies search conditions according to the number of search results and support for related information search based on search results have not been started yet.
As a method related to this unstarted problem, the method disclosed in Patent Document 3 is to refine a warm question by interactive correction and provide it together with position related information.
However, the interactive modification in this method is limited to location, and its scope is also limited to finding a goal.
[0008]
Also, the World Wide Web Consortium (http://www.w3c.org), which establishes and discloses technical documents related to the Internet, plays a central role, and activities for the realization of the Semantic Web are being promoted.
The Semantic Web aims to enable machines to understand meaning, search for information by inference, and solve problems.
RDF (http://www.w3c.org/RDF/),
RDF schema (http: //www.w3.0rg/TR/2000/CR-rdf-schema-20000327/),
OWL (http://www.w3.org/TR/owl-features/)
There is a discussion about.
[0009]
By using these technologies, it is possible to unify the description format of meta-information about information described in a document and the description of its constraints, and to explain the vocabulary used in the description of meta-information and the ontology that describes the system. Standardization is expected.
[0010]
[Patent Document 1]
JP 2000-339330 A [Patent Document 2]
JP 2000-339328 A [Patent Document 3]
JP-A-11-265141 [Non-patent Document 1]
“Extended Web Space for Regional Information Services”, Atsushi Hiramatsu, Atsushi Ishida, Transactions of Information Processing Society of Japan: Database, Vol. 41, no. SIG6 (TOD7), pp. 81-90 (2000)
[Non-Patent Document 2]
“SAIQA: Question Answering System Based on Mass Documents”, Hiroshi Sasaki et al., Information Processing Society of Japan 2001-FI-64, pp. 77-82 (2001)
[0011]
[Problems to be solved by the invention]
The present invention searches for appropriate related information from a problem relating to correction of search conditions necessary for obtaining an appropriate number of search results, and a search result obtained once, among problems when searching for documents on the Internet. The problem related to the generation of search conditions necessary for this purpose is solved by using ontology describing general relationships and systems and meta-information about documents.
The meta information description format uses the technology standardized by W3C, and implements its mechanical interpretation by referring to the ontology describing its various relationships.
[0012]
[Means for Solving the Problems]
An information search method of the present invention that achieves such an object includes a step of retrieving a search condition comprising a keyword and a relationship between keywords from an inputted natural language question in a method of searching a document on the Internet based on a search condition. , When retrieving documents on the Internet corresponding to each keyword included in the search conditions and linking the documents based on the relationship between the keywords to obtain a tree structure search result, the search result corresponding to each keyword is obtained. If the number of search results does not exceed zero or exceeds a preset threshold value, and if the number of search results is zero, the search condition is expanded according to the semantics to which the keyword belongs and the search condition limit range is expanded. If the number of steps and the number of search results exceeds the threshold value, a more specific process is performed according to the semantic system to which the keyword belongs. When the search results from the above steps are integrated based on the relationship between the keywords, the number of connections between the search results is set to zero or preset as required. Check if the threshold is exceeded, and if the number of connections between the search results is zero, the step of relaxing the connection condition semantically or numerically, and if the number of connections between the search results exceeds the threshold Steps to strengthen connection conditions semantically or numerically, and if necessary, display the tree structure search results obtained through the above steps in multiple formats: graph format, table format, and overlay display on a map A search condition for searching a document related to a semantic and geographical relationship with each attribute of a document included in a search result, if necessary, and a step presented to the user together. Was prepared in accordance with semantics, characterized in that it comprises the step of providing the search result to the user.
[0013]
The information retrieval program of the present invention that achieves the above object is a program for retrieving documents on the Internet based on a retrieval condition, and a process of retrieving a retrieval condition comprising a keyword and a relationship between keywords from an inputted natural language question When acquiring a document on the Internet corresponding to each keyword included in the search condition, and obtaining a tree structure search result by linking the documents based on the relationship between the keywords, the search result corresponding to each keyword is displayed. And check whether the number of search results is zero or not exceeding a preset threshold value.If the number of search results is zero, abstraction is performed according to the semantics to which the keyword belongs, and a limited range of search conditions is set. If the number of search results exceeds the threshold and the processing to expand, it will be more specific according to the semantics to which the keyword belongs. When a keyword is presented and the user is asked to select a keyword, and if necessary, the search results from the above process are integrated based on the relationship between keywords, the number of joints between the search results is zero or preset. If the number of connections between search results is zero, and if the number of connections between search results exceeds the threshold Is a process that strengthens the connection conditions semantically or numerically, and if necessary, the tree structure search results obtained through the above processes in multiple display formats: graph format, table format, and overlay display on a map In addition to creating a search condition for searching for documents that are semantically and geographically related to each attribute of the document included in the search results, according to the semantic system, Characterized in that it comprises a process of providing the results of the search to the user.
[0014]
The information retrieval device of the present invention that achieves the above object is a device that retrieves documents on the Internet based on a retrieval condition, a means for retrieving a retrieval condition comprising a keyword and a relationship between keywords from an inputted natural language question, When retrieving documents on the Internet corresponding to each keyword included in the search conditions and linking the documents based on the relationship between the keywords to obtain a tree structure search result, the search result corresponding to each keyword is obtained. If the number of search results is zero or does not exceed a preset threshold, and if the number of search results is zero, means for abstracting according to the semantic system to which the keyword belongs and expanding the limited range of search conditions If the number of search results exceeds the threshold, more specific keywords according to the semantics to which the keywords belong Presenting and requesting the user to select from among them, and if necessary, when integrating the search results by the above means based on the relationship between keywords, the number of connections between the search results is zero or a preset threshold value If the number of connections between search results is zero and the number of connections between search results is zero, the connection condition is semantically or numerically relaxed. Combine the means for strengthening the conditions semantically or numerically and, if necessary, the tree structure search results obtained by the above means in combination with multiple display formats: graph format, table format, and overlay display on the map A search condition for searching for a document that is related to the semantic and geographical features of the means to be presented to the user and, if necessary, for each attribute of the document included in the search result is created according to the semantic system. Characterized in that it comprises means for providing to the user.
[0015]
DETAILED DESCRIPTION OF THE INVENTION
The present invention enables efficient retrieval of documents on the Internet by using an ontology that describes general relationships and systems between concepts.
Among the above issues, the correction of search conditions to obtain an appropriate number of search results is based on an ontology that describes general relationships in order to avoid cases where the search results are zero or very large. The search condition is corrected based on the semantic relationship between the words described based on the numerical value included in the search condition.
[0016]
When the system actively performs this correction according to the number of search results, an appropriate number of search results can be acquired, and the distribution of the search results and its statistical information are more meaningful.
In addition, regarding the generation of search conditions necessary to search for relevant information from the search results once acquired, conventionally, the user himself examined the search results once and found keywords that lead to related information. It was necessary to search again.
[0017]
For this reason, the range of related information that a user can consider using his / her knowledge within a limited time is limited by himself.
In order to solve this problem, in the present invention, a search condition that leads to related information is automatically created from the attribute of the document included in the search result using the ontology, and the related information is searched based on the search condition.
This makes it possible to use a search condition that leads to a wider range of related information than when the user himself / herself creates a search condition for searching related information.
[0018]
【Example】
A search condition input window is shown in FIG.
The upper frame is for inputting search conditions in the form format, and the condition input in each line is paired with the search expression described in the search language for the extended Web space (see Non-Patent Document 1). One correspondence.
The lower frame is for inputting a question sentence described in a natural language.
The analysis of the question sentence is performed using the SAIQA question analysis module (see Non-Patent Document 2) to create a search expression in a search language format.
[0019]
The search result corresponding to this search expression has a graph structure including documents on the Internet and the link relationship therebetween, and is displayed in the search result output window of FIG.
In the upper right and lower right frames, graphs of link relationships between documents are displayed in two dimensions and three dimensions, respectively.
Each graph node corresponds to a search result document, and its connection relation is indicated by an edge.
[0020]
In the lower center frame, attributes such as the title and category of the document as a search result are displayed in a table format.
If the search result document has a geographic attribute, its coordinates are also displayed in the table, and an icon representing the document is displayed at the corresponding position on the map at the center.
In the left frame, a button for displaying a search result of related information search described below is displayed.
[0021]
FIG. 3 shows a configuration diagram of an information search system according to an embodiment of the present invention.
As shown in FIG. 3, this information search system includes a search condition input module 100, a SAIQA question analysis module 101, a search module 102, a structured database (DB) 103, a related information search condition generation module 104, and a search result output module 105. It may be realized by hardware that performs each processing, or may be realized by software by causing a computer to read a predetermined program.
[0022]
The structured database 103 acquires a document existing on the Internet using a web robot such as wget (http://www.wget.org/), and extracts a URL indicating the position of the document and each document. Meta information based on ontology is structured and stored in advance.
With this structured database 103, the following processing is speeded up.
The search condition input module 100 converts the received search condition into a search expression described in a search language for the extended Web space.
[0023]
At this time, in the case of the input in the form format, the search formula is converted as it is in one-to-one correspondence. In the case of the input in the natural sentence format, the keywords included in the input and the relationship between them are extracted by the SAIQA question analysis module 101, and the search formula Create
Next, FIG. 4 shows a procedure for correcting search conditions for obtaining an appropriate number of search results.
This search condition correction is realized by cooperation between the search module 102 and the structured database 103.
[0024]
First, keywords included in the search expression are extracted (step 200), and search results corresponding to the keywords are acquired (step 201).
At this time, it is checked whether the number of search results is zero or not exceeding a preset threshold value (steps 202 and 203). If the number of search results is zero, abstraction is attempted according to the semantic system to which the keyword belongs ( Step 204), expanding the limited range of search conditions.
For example, to abstract according to the semantic system means to apply keywords of one or more nodes having a higher concept than the node to which the current keyword belongs in the semantic category tree structure.
If the number of search results exceeds the threshold, a more specific keyword is presented according to the semantic system to which the keyword belongs, and the user is requested to select from the keyword (step 205).
[0025]
For example, it is assumed that “bus stop” and “restaurant” are extracted as keywords from the question sentence “What is the restaurant near the bus stop?”.
At this time, it is assumed that the search result corresponding to “bus stop” is zero and the number of search results corresponding to “restaurant” exceeds the threshold.
In this case, for the keyword “bus stop”, the “stop” category to which the keyword “bus stop” belongs, the “traffic” category above it, and the search conditions are abstracted until a sufficient number of search results are obtained.
[0026]
On the other hand, for the keyword “restaurant”, a more detailed concept included in the category of “restaurant” such as “Chinese restaurant”, “French restaurant”, “Kappo”, etc. is presented to prompt the user to select the keyword To materialize.
When the acquired search results are integrated based on the relationship between keywords (step 206), it is checked whether the number of connections between the search results is zero or does not exceed a preset threshold value (steps 207 and 208).
[0027]
If the number of connections between search results is zero, the connection conditions are relaxed semantically or numerically (step 209).
If the number of connections between search results exceeds the threshold, the connection condition is strengthened semantically or numerically (step 210).
For example, in the case of the same question sentence, “What is the restaurant near the bus stop?”, It is evaluated in two ways: the geographical relationship condition that “the soba is in the same block” and the distance condition that is “within 250 m”. The
[0028]
Of these, if the search result is zero under the condition of “in the same block”, the search result is relaxed to the condition of “in the same district” indicating a wider geographical range, and the search is performed under the condition of “in the same block” If the result exceeds the threshold, it is enhanced to the condition of “adjacent” indicating a narrower geographic range.
When the search result is zero under the distance condition “within 250 m”, the target range is expanded, and when the search result exceeds the threshold, the target range is narrowed.
Note that when the search expression includes only one keyword, the above-described procedure for relaxing and strengthening the connection condition is not applied.
[0029]
The search result (step 211) obtained as a result of the above-described search condition correction may be provided to the user as it is. Further, based on this search result, a search condition generation procedure for searching related information is shown in FIG. Show.
The search condition generation for the related information search is realized by the related information search condition generation module 104.
First, the process is started based on the search result obtained as a result of the above-described search condition correction (step 300), and attributes such as the URL, title, keyword, and category of the document included in the search result are acquired from the structured database 302 ( Step 301).
[0030]
The structured database 302 shown in FIG. 5 is the same database as the structured database 103 shown in FIG. 3, and information that can be acquired as document attributes is based on information that is acquired in advance and stored in the database.
Then, on the condition of the acquired attribute value or the upper and lower concept derived from the value, related information corresponding to each attribute is searched by each related information search module 303 to 307, and the search result is shown in FIG. Display in the search result output window.
[0031]
The related information search modules 303 to 307 are a link destination link source search module 303, a peripheral information search module 304, a category search module 305, a time condition search module 306, and a fee search module 307.
At this time, it is also possible to display a search result of related information corresponding to all the attributes acquired in step 301 by setting the user interface, or to search related information for only the attributes preset by the user among the acquired attributes. It can also be made.
[0032]
In the search result output, when you select one of the documents included in the first search result, the related information is displayed for each attribute, the search method displays all the related information, the search result of related information search For example, there may be a method of displaying only the related information of the document or its attribute that the system judges to be useful from the number and the amount of information.
Which method is adopted depends on the design of a search result output window which is a user interface, and each procedure of the present invention is not affected by the design.
In FIG. 5, five types of related information search modules 303 to 307 are shown, but this corresponds to the attribute of the document in the search result, and is not limited to the five types in the drawing.
[0033]
As the related information search module, a module corresponding to an attribute that can be acquired from the structured database 302 is prepared.
As a specific example, FIG. 6 shows a search result for the question “What is the bus stop near Kinkakuji?” And its related information search result.
First, the search result for the question sentence has a tree structure in which the Web page of “Kinkakuji” and the Web page of the bus stop nearby are connected by a link based on the distance relationship.
Then, the value of each attribute shown in FIG. 6 is acquired by the procedure 301. The description format and vocabulary of the attribute acquired at this time is meta information described based on the W3C standard, and mechanically. Can be processed.
[0034]
Then, the related information search is performed for each attribute of each Web page acquired from the database, the link destination based on the URL, the Web page of the link source, the peripheral information based on the address or geographic coordinates, the temporal relationship based on the visit time and the timetable Each type of related information such as information is additionally displayed in each corresponding frame in FIG.
[0035]
As described above, the main feature of the present invention is to apply an ontology that systematically represents the relationship between concepts when searching for documents on the Internet, and by using the ontology, The effect that search efficiency improves is produced.
The present invention also compares the number of URLs obtained by keyword search with a predetermined threshold number, applies an ontology depending on the comparison result, generates a new keyword, and performs a search again. It is also a feature that when the documents obtained as a result of the search are integrated, the connection condition is relaxed and strengthened depending on the comparison result between the number of connections suitable for the connection condition and the threshold number of connections. It is also characterized by having a feedback loop.
[0036]
【The invention's effect】
As described above, according to the present invention, based on the meta information described based on the ontology, (1) the search condition input by the user is appropriately modified according to the number of search results, and (2) acquired. This makes it easy to obtain information related to the search results.
[Brief description of the drawings]
FIG. 1 is a user interface screen (screen shot) for search condition input of a search program according to the present invention.
FIG. 2 is a user interface screen (screen shot) for search result output of a search program according to the present invention.
FIG. 3 is a block diagram showing an internal configuration of an information search system according to the present invention.
FIG. 4 is a flowchart showing a procedure for correcting a search condition according to the present invention.
FIG. 5 is a flowchart showing a search condition generation procedure for searching related information according to the present invention.
FIG. 6 is an explanatory diagram showing an example of attributes included in a search result based on the present invention.
[Explanation of symbols]
100 Search condition input module 101 SAIQA question analysis module 102 Search module 103, 302 Structured database (DB)
104 Related Information Search Condition Generation Module 105 Search Result Output Module 303 Link Destination Link Source Search Module 304 Peripheral Information Search Module 305 Category Search Module 306 Time Condition Search Module 307 Price Search Module

Claims (12)

In a method for searching documents on the Internet based on a search condition, a step of retrieving a search condition including a relationship between a keyword and a keyword from an inputted natural language format question, and an Internet corresponding to each keyword included in the search condition When obtaining a tree structure search result by linking documents based on the relationship between keywords, the search result corresponding to each keyword is obtained, and the search result number is zero or a preset threshold value If the number of search results is zero and if the number of search results is zero, the step of abstracting according to the semantic system to which the keyword belongs and expanding the limited range of search conditions, and if the number of search results exceeds the threshold, A more specific keyword is presented according to the semantic system to which the keyword belongs, and the user can select from the keywords. Information retrieval method characterized by comprising the steps of determining. When integrating the search results by the above steps based on the relationship between keywords, check whether the number of connections between search results is zero or exceeds a preset threshold, and if the number of connections between search results is zero, The method includes the steps of relaxing the connection condition semantically or numerically, and strengthening the connection condition semantically or numerically when the number of connections between search results exceeds a threshold value. Item 2. The information search method according to Item 1. 3. The step of presenting a tree structure search result acquired through the above steps to a user in combination with a plurality of display formats of a graph format, a table format, and a superimposed display on a map. Information retrieval method described. It has a step of creating a search condition for searching a semantically and geographically related document for each attribute of the document included in the search result by the above step according to a semantic system and providing the search result to the user. The information search method according to claim 1, 2 or 3. A program for searching documents on the Internet based on a search condition, which extracts a search condition consisting of a relationship between a keyword and a keyword from an inputted natural language format question, and corresponds to each keyword included in the search condition When retrieving documents on the Internet and linking documents based on the relationship between keywords to retrieve tree structure retrieval results, retrieve the retrieval results corresponding to each keyword, and the number of retrieval results is zero or preset If the number of search results is zero, if the number of search results is zero, the processing is performed in accordance with the semantic system to which the keyword belongs and the search condition is expanded to expand the limited range of search conditions. Presents more specific keywords according to the semantic system to which the keywords belong, and allows the user to select from them. Computer readable information retrieval program, which comprises a Mel process. When integrating the search results by the above process based on the relationship between keywords, check whether the number of connections between search results is zero or exceeds a preset threshold, and if the number of connections between search results is zero, Claims characterized by including processing for relaxing the connection condition semantically or numerically, and processing for strengthening the connection condition semantically or numerically when the number of connections between search results exceeds a threshold value Item 6. The information search program according to Item 5. 7. A process for presenting a tree structure search result obtained through the above process to a user in combination with a plurality of display formats of a graph format, a table format, and a superimposed display on a map. The information retrieval program described. Including a process for creating a search condition for searching for a semantically and geographically related document for each attribute of a document included in a search result by the above process according to a semantic system, and providing the search result to a user. The information search program according to claim 5, 6 or 7. In an apparatus for searching documents on the Internet based on a search condition, means for extracting a search condition comprising a relationship between a keyword and a keyword from an inputted natural language question, and an Internet corresponding to each keyword included in the search condition When obtaining a tree structure search result by linking documents based on the relationship between keywords, the search result corresponding to each keyword is obtained, and the search result number is zero or a preset threshold value If the number of search results is zero, and if the number of search results is zero, the means of abstracting according to the semantics to which the keyword belongs and expanding the limited range of search conditions, and if the number of search results exceeds the threshold, Present more specific keywords according to the semantic system to which the keywords belong, and ask the user to select from them Information retrieval apparatus characterized by comprising a stage. When integrating the search results by the above means based on the relationship between keywords, check whether the number of connections between search results is zero or exceeds a preset threshold, and if the number of connections between search results is zero, Claims characterized by comprising means for relaxing the connection condition semantically or numerically and means for strengthening the connection condition semantically or numerically when the number of connections between search results exceeds a threshold. Item 10. The information search device according to Item 9. 11. The apparatus according to claim 9, further comprising means for presenting the search result of the tree structure acquired by the means to the user in combination with a plurality of display formats of graph format, table format, and overlay display on a map. Information retrieval device. It has means for creating a search condition for searching a semantically and geographically related document for each attribute of a document included in a search result by the above means according to a semantic system and providing the search result to a user. 12. The information search apparatus according to claim 9, 10 or 11.

Images