JP2001229178A - Document retrieval method and apparatus and recording medium recording this method - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To provide a document search method and apparatus capable of supporting a user conscious of a group and re-searching by directly and efficiently eliminating unnecessary search results in a group unit. A search gateway that presents search results presents pages that belong to the same group and that are found by a search, and do not limit the link relationship between the groups, rather than the conventional list of grouped page data. Present to assist the user in determining the suitability. Further, the search server performing the search (re-search) uses the deletion instruction means provided at the search gateway for excluding the specified group from the search target, and when the user instructs to delete the specified group, At the time of the search, pages belonging to the specified group are collectively excluded from subsequent search results (or search targets).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search procedure and presentation of search results in a WWW search engine (hereinafter referred to as a search engine) and the like. Related to interactive information retrieval technology.

[0002]

2. Description of the Related Art A conventional search engine returns a ranking list of documents to be searched based on a unique scoring method. FIG. 11 shows a flow of a conventional search process. The user makes relevance judgments only based on the information of the presented ranking list title and summary, and if the results of the first search are unsatisfactory, add or delete keywords to improve this, and Search for.

At present, a general search result display has a drawback that a plurality of pages expressing the same subject are scattered throughout the search result display.

FIG. 12 is a Venn diagram showing the distribution of pages in a conventional WWW search. In the following, a Web page is interpreted as a file corresponding one-to-one with a URI, and a Web document is interpreted as a single subject represented by one or more Web pages or a set of constituent pages thereof. Here, | A |
The set of the top 10 pages is | B |. When the search result is large (for example, | A | = 100, | B | = 10), the conventional search engine returns pages belonging to | A | by dividing them into ten lists belonging to | B |. Web pages are often treated as independent documents by general search engines, but reference [1] "Masaki Harada, Shinya Sato, Kazuhiro Kazama" Estimation of hierarchical structure of WWW pages and application to search system " As described in the report of the 118th Database Research Group, pp. 105-112, May, 1999, etc., a plurality of Web pages often constitute one document. A group consisting of a plurality of pages expressing such a single subject is referred to as a group A.

It is clear from FIG. 12 that pages belonging to group A are scattered throughout the search results. The user looks at the ten lists given as search results from the beginning and determines the suitability of each page. For example, in the list of elements of B, the page belonging to group A (that is, the subject represented by group A) Even if it is determined that the content is not suitable, since the page belonging to the group A exists in AB, the user sees the page having the content out of the search purpose again as the page hit in the search.

However, recently, the above-mentioned references [1] and [2]
"Michel Chen, Martin A. Hea
rst, "Cha-Cha: Contextulizi
ngIntranet Search Result
s ", http: //cha-cha.berkele
y. edu / ", and reference [3]" Gen Takano and Shinya Kubo, "WWW search ranking system using information structure analysis", Information Processing Society of Japan 59th Annual Convention, pp. 91-9.
2, 1999, etc., a system has been developed that groups a plurality of pages from a certain viewpoint and displays search results for each group. These show the top several dozens of pages that hit the keyword search by grouping them together, so that the pages that hit the search belonging to a page set that constitutes a group of pages from a certain viewpoint are listed in the search result list The purpose of the present invention is to solve the problem of the conventional search result display that the display is scattered.

[0007] On the other hand, re-search methods are generally (1) a method of adding or deleting a keyword, and (2) a similar document search. The former is, for example, a method of further narrowing down the search results obtained earlier by adding search keywords, or obtaining a wider range of search results by deleting some keywords that are too many. The latter is a method of searching for a page similar to the page determined to be correct by the user (determining the similarity by comparing the feature values of the quantified pages) using a technique called relevance feedback. Reference [4] "http: //www.excit"
e. co. jp "realizes this by a function called" similarity search ".

[0008]

What is common to the conventional search result display on a group basis is that the result of dividing a set of pages to be searched into a plurality of groups is simply rearranged as a result of keyword search, that is, one sort of result. This means that it is used only as a sort function. Specifically, when presenting the pages belonging to B, the results of the keyword search are not arranged in the order of the page scores (the score of the pages is not reflected at all), but are arranged in groups as shown in FIG. indicate.

However, in FIG. 13, although the result of the group A is displayed in the search result list 1 (page belonging to B), the result (page belonging to AB) which cannot be displayed on the same page is browsed. In this case, search results belonging to group A are also displayed in search result list 2. In each of the studies [1], [2], and [3] described above, since this situation always occurs, pages that hit a search belonging to the same group are scattered on a plurality of search result display pages. Resulting in.

Further, regarding the conventional re-search method,
There is a problem that even if a user adds or deletes a keyword to a document determined to be non-conforming, or specifies a conforming document, the document cannot be removed properly. The problem is that there is no way to directly remove a document and its related documents. Narrowing down by keywords is indirect, and erroneous selection of keywords may result in missing required documents, as well as better results. The same applies to similar document search. In the search on a group basis, a match is determined for one subject represented by a page set belonging to the group.
If the content expressed by a certain group does not meet the user's search purpose, all pages belonging to the group should be excluded from the search target by narrowing down.

An object of the present invention is to perform user navigation (user support) with more awareness of a group, and to perform a search again by directly and efficiently eliminating unnecessary search results for each group. An object of the present invention is to provide a search method and apparatus and a recording medium on which the method is recorded.

[0012]

In order to solve the above-mentioned problems, a document search method according to the present invention comprises a first step of dividing a document set including text into subsets based on topics, and A second procedure for searching for a document related to a search term from among them, and a presentation instruction for dividing the search result into subsets divided by the first procedure, and instructing presentation of each document of the subset. Means, link instruction means for instructing presentation of another subset linked to the subset, and third procedure for presentation together with deletion instruction means for instructing deletion of the subset, and division into each of the presented subsets When the user determines that the retrieved search result is not suitable, the user has a fourth procedure of repeating the second procedure and the third procedure again according to an instruction.

Alternatively, in the fourth step, the third step
When the deletion of the subset is instructed by the deletion instructing means presented in the procedure of the above, the subset obtained in the first procedure from the search target or the search result in the repeated second procedure is used. , And excludes those included in the designated subset.

The document search apparatus according to the present invention also includes a first means for dividing a document set including text into subsets based on topics, and searches for documents related to a search term from the entire set. A second unit, a presentation instructing unit that divides a search result obtained by the second unit into subsets divided by the first unit, and instructs presentation of each document of the subset; Link instruction means for instructing the presentation of another subset linked to the sub-set, and third means for presenting together with deletion instruction means for instructing the deletion of the subset, wherein the second means comprises: When the user determines that the search result divided for each is inappropriate, the document is searched again according to the instruction, and the search result is sent to the third means.

[0015] Alternatively, the third means, when a deletion of a subset is instructed by the presented deletion instructing means, from the re-search result of the second means or of the second means. It is characterized in that those included in the specified subset are excluded from the search targets.

Further, a program for causing a computer to execute the procedure in the text document search method described above is recorded on a computer-readable recording medium.

The third means / procedure for presenting the search result is that, for example, dozens of pages having a high score among pages hit by a search by a search term (keyword or the like) are grouped for each document set (group). The user is presented with the following instruction means for each group.・ Documents belonging to the group (for example, search result page)
Document presenting means for presenting (displaying) only in another area (an area different from the area where the search result is displayed), for example, a link instruction for presenting (displaying) the link destination group of the group in another area. Means, a deletion instruction means for deleting the group from the search target.

The second means / procedure for retrieving (re-retrieving) the above-mentioned group as a basic unit is a method of retrieving documents (pages) belonging to the group designated to be deleted by the above-mentioned deletion instructing means (or a retrieval object). ) Are excluded together.

In the present invention, a third method for presenting a search result is provided.
Means / procedure is not limited to the conventional list of grouped page data, but presents pages that belong to the same group and is found by a search, presents a link relationship between groups, and matches the user. Support judgment.

[0020] The second means / procedure for searching (re-searching) includes a deletion instruction means provided in the third means / procedure for presenting the search result for excluding the designated group from the search target. When the user instructs to delete a group that is determined to be unnecessary, pages belonging to the specified group are excluded from subsequent search results (or search targets) at the time of re-search.

FIG. 2 shows a state in which pages belonging to group A are excluded from the state of FIG. It is clear from the figure that pages belonging to group A are not displayed in the subsequent results.

[0022]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a diagram showing the configuration of an embodiment of a search device according to the present invention and the flow of processing thereof.

[0024] The present apparatus comprises: first means for dividing a set of WWW pages into subsets based on topics; and a means for retrieving each of the divided sets from the entire page set for pages related to a search term. And a third means for dividing the searched result into subsets divided by the first means and presenting the divided results together with instruction means described later.

The “grouping based on topics of a set of pages to be searched” in FIG. 1 corresponds to the above-described first means. Several grouping methods have been proposed. For example, in Document [1], pages belonging to the same directory are basically regarded as one group, and a set of pages to be searched is grouped.

In FIG. 1, "index word extraction", "index", "search server", "search gateway (G
W) "corresponds to the second means described above." Grouping based on topics of a set of pages to be searched ",
“Index word extraction” and “index” creation are performed before the search processing. An index is a file (database) from which all document IDs of a search including the word can be subtracted from a certain word, and a certain group ID
From which all document IDs belonging to the group ID can be subtracted. Index words can be extracted by ch
asen [“Japanese morphological analysis system [cha sen]”, h
ttp: // cl. aist-nara. ac. jp /
lab / nlt / chasen /]. In addition, the index creation is performed in conjunction with the search server (in FIG. 1) using namaz [“Namazu the
full text retrieval sea
chsystem ", http: // openlab.
ling. gr. jp / namazu /] can be used. Namaz is package software that collects tools necessary for conventional information search, such as index creation and search server.

The functions required of the search server in the present invention include "Boolean search" and "field search". The former receives, for example, a query “search for information and” (a search request to the search server) and “information”.
This is a function of searching for documents that include the word “search” and the word “search”. The latter receives, for example, a query of “information and group ID = 00230142”,
Contains the word "information" and belongs to group ID 00
230142 is a function for searching for a document. Namaz described above provides a search server that satisfies these requirements.

Each part of "search GW" and "user interface" in FIG. 1 corresponds to the above-mentioned third means. The search GW is, for example, CGI (Common Gateway)
Interface). The user interface is a part that receives an input from a user among those displayed on the hypertext browser.
What is input from here is sent to the search GW in a specific format.

The "metafile with index creation tag" portion in FIG. 1 is data stored in an external storage device used when creating the "index" in the above figure.

Next, as an example, a description will be given of a search for a document including "information search". A user sends a search request to a search gateway (GW) program via a user interface. This request includes a keyword or phrase to be included in the document of the search result, and a “group ID list that does not display the search result” (group IDs in this list).
Pages with are not displayed as search results. Hereafter, D
-List), a command (see Table 1), and a specific group ID if necessary. In the first search (not a re-search), D-list is empty (｛｝), and the search request is as shown in 1 of Table 2. The command “NewSe
"arch" means a new search request, and D-
If the list has been created, the list is cleared.

[0031]

[Table 1]

[0032]

[Table 2]

The search GW converts a user's search request into a format (query) understandable by the search server and sends it to the search server. Query generation refers to, for example, when a search request is described in a natural language, performing a process such as morphological analysis to generate a query Q including a logical combination of keywords. In this example, Q = “information and search”. The search server interprets Q and returns a list of pages corresponding to the search request. Table 3 is an example of the list. Table 4 is a list of pages belonging to each group. The second row is the identification number of the group, each group having a unique identification number.
Pages with "*" at the beginning (pages A4, etc.) indicate pages that did not hit the search.

[0034]

[Table 3]

[0035]

[Table 4]

The search GW sorts the top several items in the list of Table 3 for each group. Since the information of the group to which each page belongs is indexed as page information when creating an index, it is described in a list returned by the search server as a result as shown in Table 3. FIG. 3 shows this as a screen for the user to confirm.

The search results are such that the pages that hit the keyword search are sorted by groups (in this example) every five pages and displayed on the left page portion of FIG. The processing up to this point can be performed by using a conventional technique such as literature [1]. In the present invention, areas (hyperlinks, button widgets, etc.) indicated by 1 to 3 in the figure are further displayed for each group. By designating each part with a pointing device, the following operations can be provided to the user.

1. List of Applicable Documents in Group A description will be given of an example of the flow of processing when “List of applicable documents in group” is selected in the “Group A” display section of FIG. When this part is specified, the user interface sends 2 of Table 2 as a search request to the search GW.
The search GW sends a query “information and search and group ID: 00230142” to the search server, and the search server returns a list of pages that include the keyword and belong to the specified group. FIG. 4 shows a display example of the search result list on the browser. In this example, “page A1” which did not appear in the list of the top five pages is listed as a page belonging to group A.

As an option, pages belonging to the group browsed by this function can be regarded as read information so as not to appear in subsequent search results. This is the D-lis of the user interface
This is achieved by adding the group A ID “00230142” to t. In FIG. 3, a description will be given of an example of a processing flow in a case where “to the next page” is selected after only “List of applicable documents in group” of “Group A” is clicked. When this part is designated, the user interface adds “00230142” to the D-list and sends 3 in Table 2 as a search request to the GW. The search GW gives Q to the search server to obtain Table 3, and then, from the sixth item in Table 3, the group in the D-list (here, 00
230142), excluding the pages belonging to the five cases. According to Table 3, these are "Page D1, Page D
4, page C4, page C2, page D3 ". The search GW edits these for each group and passes them to the user interface. FIG. 5 is a display example of this search result in the browser. If "List of applicable documents in group" in the result display portion for group A in the figure was clicked before "To page" was clicked, the right side of the figure is displayed.

In this example, "Page A1" appears in the second list according to the conventional search result display method. However, in the present invention, the page is browsed all at once, and subsequent results are displayed in the same group. It is possible to hide the page to which it belongs.

2. Link destination group list As shown in FIG. 6, from page A2 to page C2, page A
If there is a link from page 3 to page D1, these are links from group A to groups C and D, respectively.
Groups C and D are called link destination groups with respect to group A.

An example of the flow of processing when "list of linked groups" in the "group A" display section of FIG. 3 is selected will be described. When this part is designated, the user interface sends 4 in Table 2 as a search request to the search GW. The search GW gives Q to the search server and obtains Table 3. Next, a group data file as shown in FIG. 7 corresponding to the designated group is read to create a list of link destination group IDs (a character string surrounded by <link_to> tags in FIG. There are as many link destination groups as possible), and scan the third column of Table 3 to create a group ID list from which duplicate group IDs have been removed. Group ID that exists in both of these two lists
Is passed to the user interface as a search result.
In this way, among the link destination groups, the groups including the page hit in the search are listed as shown in FIG. For each group, a link is provided to display the corresponding document in the same group.

3. This group is deleted. An example of the processing flow when “Delete this group” is selected in the “Group B” display section of FIG. 3 will be described.
When this part is specified, the user interface
“00711829” is added to the D-list, and 5 in Table 2 is sent to the GW as a search request. Search GW is Q
Is given to the search server, and Table 3 is obtained.
The pages excluding all pages belonging to the group in the st (here, 00711829) are edited for each group and sent to the user interface. FIG. 9 shows the result of browsing with a browser by the user who has selected “Delete this group”.

Thereafter, for example, when the keyword "Japanese" is added and the search is performed again, the user interface sends 6 in Table 2 as a search request to the search GW.
Since the search GW returns only pages that do not belong to a group in the D-list to the user interface as search results, all pages (pages B1 to B1) belonging to group B are returned.
B4 (page B4 is excluded from the search even though it has not been hit in the search) does not appear in the search result list in this case. This state continues until a search request including the command “NewSearch” as shown in 7 of Table 2 is issued.

Next, FIG. 10 shows an embodiment of a search procedure using the present invention.

The user first inputs a search request such as a keyword, and obtains a search result for each group (FIG. 3, left page). Next, in addition to the title and the summary, the matching determination of the group is performed based on the corresponding page in the group (FIG. 4) and the information of the link destination group including the corresponding page (FIG. 8).
If a group (contents represented by pages belonging to)
If does not meet the search purpose, use the "Delete this group" function to narrow down the search results. Otherwise, narrow down by adding or deleting keywords as usual.

As shown in the above-described embodiment, according to the present invention, it is possible to establish a method of performing group-conscious user navigation by presenting a list of applicable documents in a group and a link destination group. . -It is possible to establish a method for directly and efficiently excluding unnecessary search results from search targets in groups.

In the present apparatus and method, since the result is browsed in group units and the read information can be easily deleted from the list, when a keyword is added or deleted, it is determined that it is unnecessary before the keyword operation. Since the page does not appear in the search results, browsing of the results can be performed more efficiently than the conventional divided list.

It should be noted that some or all of the functions of each unit shown in FIG. 1 can be realized using a computer, or that the computer can execute the processing procedure shown in FIG. Needless to say, a program for realizing the function of each unit in the computer or a program for causing the computer to execute the processing procedure is stored in a recording medium readable by the computer, for example, an FD (floppy disk) or an MO. , ROM, memory card, CD, DVD,
It can be recorded on a removable disk, provided, and distributed.

[0050]

As described above, according to the present invention, in a search engine, etc., by presenting a list of applicable documents in a subset and a link destination subset, user navigation with a greater awareness of groups can be realized. It is possible to do. Unnecessary search results can be directly and efficiently excluded from a search result or a search target in a subset unit (the more documents that constitute a subset, the more documents can be eliminated at once). .

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an embodiment of the present invention.

FIG. 2 is a Venn diagram showing that when a re-search is performed using the present invention, a state in which pages belonging to a group, which is a problem of the conventional technology, is not scattered in search results does not occur. FIG.

FIG. 3 is a diagram illustrating a search result display example according to the embodiment of the present invention.

FIG. 4 is a diagram illustrating an example of an action that occurs when a user selects “List of applicable documents in a group” in a search result display example according to the embodiment of the present invention.

FIG. 5 is a diagram illustrating an example of an action that occurs when a user selects “next page” in a search result display example according to the embodiment of the present invention.

FIG. 6 is a conceptual diagram of a “link destination group” in the embodiment of the present invention.

FIG. 7 is a diagram illustrating a group data file according to the embodiment of the present invention.

FIG. 8 is a diagram illustrating an example of an action that occurs when a user selects “link destination group list” in a search result display example according to the embodiment of the present invention.

FIG. 9 is a diagram illustrating an example of an action that occurs when a user selects “delete this group” in a search result display example according to the embodiment of the present invention.

FIG. 10 is a diagram showing an example of a search result obtained by a user using the present invention to determine whether or not a matching is performed using information on a corresponding document in a group or a link destination group, and unnecessary results are excluded from a search target in a group unit. FIG. 9 is a diagram illustrating an example of a flow of a search process of “narrowing down”.

FIG. 11 is a diagram illustrating a flow of a conventional search process in which a user performs an operation such as addition or deletion of a keyword and performs a search again based on only title and summary information.

FIG. 12 is a diagram illustrating, using a Venn diagram, a state in which pages belonging to a certain group are scattered throughout a search result in a conventional search engine and a search engine capable of performing a search in a group unit.

FIG. 13 is a diagram specifically illustrating a state in which pages belonging to a certain group are scattered throughout a search result in a conventional search engine capable of performing a search in a group unit.

Claims (5)

[Claims] A first step of dividing a document set including text into subsets based on topics; a second step of searching for a document related to a search term from the entire set; and the search result. Is divided for each subset divided by the first procedure, and presentation instruction means for instructing presentation of each document of the subset, and link instruction means for instructing presentation of another subset linked to the subset And a third procedure for presenting together with the deletion instruction means for instructing the deletion of the subset, and when the user determines that the search result divided for each of the presented subsets is inconsistent, A text document search method, comprising: a fourth procedure of repeating the second procedure and the third procedure again. 2. In the fourth procedure, when a deletion of a subset is instructed by the deletion instructing means presented in the third procedure, a search target or a search result in the repeated second procedure is determined. 2. The document search method according to claim 1, wherein a subset included in the designated subset is excluded from the subset obtained in the first procedure. 3. A first means for dividing a document set including text into subsets based on topics, a second means for searching a document related to a search term from the entire set, and the second means. Means for dividing the search result by the first means into each of the subsets divided by the first means, and instructing presentation of each document of the subset, presentation of another subset linked to the subset And third means for presenting together with deletion instruction means for instructing deletion of the subset, wherein the second means is incompatible with a search result divided for each presented subset. A text document search apparatus characterized in that, when the user determines that the search is performed, the document is searched again according to the instruction, and the search result is sent to the third means. 4. The method according to claim 1, wherein when a deletion of the subset is instructed by the presented deletion instructing means, the third means selects one of the re-search results of the second means or the one of the second means. 4. The document search apparatus according to claim 3, wherein the search target excludes those included in the designated subset. 5. A text document search method characterized by recording a program for causing a computer to execute the procedure in the text document search method according to claim 1 or 2 on a recording medium readable by the computer. Recording medium.