JPH1185764A - Method and apparatus for statistically estimating the number of search results, and storage medium storing program for statistically estimating the number of search results - Google Patents

Abstract

(57) [Summary] [Problem] In addition to the search results obtained with a certain search word,
A method and apparatus for statistically estimating the number of search results that can be calculated without retaining the approximate change in the number of search results when another search word is added to the entire search result set, and a statistical method for the number of search results A storage medium storing an estimation program is provided. According to the present invention, a part of a search result set that matches a search word is used to estimate a ratio of an arbitrary search word included in the entire search result set that matches the search word.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for statistically estimating the number of search results and a storage medium storing a program for statistically estimating the number of search results, and more particularly to further narrowing down search results in a data search system. Therefore, the present invention relates to a method and apparatus for statistically estimating the number of search results to know how the number of search results changes when a search word is added, and a storage medium storing a program for statistically estimating the number of search results.

[0002]

2. Description of the Related Art In a data search system, it is important to narrow down search results to a number suitable for browsing. For this reason, conventionally, in a document database or the like,
There is a method in which a set of data obtained by a previous search is held in a search process, and a search result set is reduced by adding search words.

[0003]

However, in the above-mentioned conventional method, a storage area for holding a set of search results is required, and it is necessary to hold a search history performed so far. However, it is difficult to satisfy these conditions when it is necessary to accept tens of thousands of search requests at once, or when a search is performed using a communication protocol that cannot maintain a history. Therefore, in such a case, there is a problem that it is necessary to prevent the search from being performed or to add a search word, to generate a new search question and perform the search from the beginning each time.

The present invention has been made in view of the above points, and it has been found that a change in the approximate number of search results when another search word is further added to a search result obtained by a certain search word,
It is an object of the present invention to provide a method and apparatus for statistically estimating the number of search results that can be calculated without holding the entire search result set, and a storage medium storing a program for statistically estimating the number of search results.

[0005]

According to a first aspect of the present invention, there is provided a data search system for determining how the number of search results changes when a search word is added to further narrow down search results. In a method for statistically estimating the number of search results, using a part of a set of search results that match a search word, estimate a ratio of an arbitrary search word included in the entire search result set that matches the search word. .

According to a second aspect of the present invention, in a data search system, when a search term is added in order to further narrow down search results, the number of search results in order to know how the number of search results changes varies statistically. In the estimation method, a list of words that can be additional search words is generated and presented by using a part of the document set that matches the search words, so that the user can select the search words to be added, Is estimated in the entire document set that satisfies.

According to a third aspect, in a data search system, when a search term is added to further narrow down search results, the number of search results is statistically determined to know how the number of search results changes. In the estimation method, the number of search results when another search word is added to the search word is estimated using a part of the search result set that matches the search word.

According to a fourth aspect of the present invention, in a data search system, when a search term is added in order to further narrow down search results, the number of search results in order to know how the number of search results changes varies statistically. In the estimation method, a list of possible additional search terms is generated and presented using a part of the document set that matches the search terms, so that the user can select the search terms to be added, Estimate the number of search results when added.

FIG. 1 is a diagram showing the principle of the present invention. Fifth
The invention of the present invention is a statistical estimation device for the number of search results to know how the number of search results changes when a search term is added to further narrow down the search results in a data search system, Search result sample extracting means 11 for extracting a part of a search result set matching a search word
0, a search result set and an additional search word, and an additional search word ratio calculating means 120 for obtaining and outputting a ratio of the search result set including the additional search word by using a part of the search result set; and Having.

According to a sixth aspect of the present invention, in a data search system, when a search word is added in order to further narrow down a search result, a statistical search of the number of search results to know how the number of search results changes is performed. A search result sample extracting means for extracting a part of a search result set that matches a search term, and a list of potential search terms that can be an additional search term from a part of the search result set extracted by the search result sample extracting means A search word candidate generating means that sets
Selects the search term corresponding to the selection signal received from the user from the search term candidates, and selects a selection means as an additional search term, and estimates a proportion of a part of the search result set that matches the additional search term. And an additional search term ratio calculating means for outputting.

According to a seventh aspect, in the data search system, when a search term is added to further narrow down the search results, the number of search results in order to know how the number of search results changes varies statistically. An estimating apparatus, a search word candidate generating means for extracting a total number of search results and a part of a search result set matching the search word, and estimating a part of the search result set including an additional search word Means for calculating the number of additional search words to estimate and output the number of cases.

According to an eighth invention, in a data search system, when a search word is added in order to further narrow down search results, a statistical search of the number of search results to know how the number of search results changes is performed. An estimating device, a search term candidate generating means for extracting a part of a document set that matches the search result and a part of the document set that matches the search term, and an additional search term in the total number of search results and the document set A search word candidate generating unit that uses a list of search words as a search word candidate; a selection unit that uses a search word corresponding to a selection signal obtained from a user from the search word candidates as an additional search word; Means for calculating the number of additional search words and outputting the number of additional search words.

According to a ninth invention, in a data search system, when a search term is added to further narrow down a search result, a statistical search of the number of search results to know how the number of search results changes is performed. A storage medium storing an estimation program, wherein a search result sample extraction process for extracting a part of a search result set that matches a search word, a search result set and an additional search word are received, and a part of the search result set is used. An additional search term ratio calculating process for calculating and outputting a ratio of the search result set including the additional search term.

According to a tenth aspect of the present invention, in the data search system, when a search term is added to further narrow down the search results, the number of search results is statistically determined to know how the number of search results changes. A search result sample extraction process for extracting a part of a search result set that matches a search word on a storage medium storing an estimation program, and additional search from a part of the search result set extracted by the search result sample extraction process A search word candidate generation process for selecting a list of possible words as search word candidates, a selection process for selecting a search word corresponding to a selection signal received from a user from the search word candidates and setting it as an additional search word, and a search result set And a process of estimating and outputting a ratio of those that match the additional search word among a part of the additional search words.

According to an eleventh aspect, in the data search system, when a search term is added in order to further narrow down the search results, the number of search results in order to know how the number of search results changes varies statistically. A storage medium storing an estimation program, a search word candidate generation process for extracting a part of a search result set matching a total number of search results and a search word, and an additional search in a part of the search result set And a process of calculating the number of additional search terms to estimate and output the estimated number of terms containing words.

According to a twelfth aspect, in the data search system, when a search term is added to further narrow down the search results, the number of search results is statistically determined to know how the number of search results changes. A storage medium storing an estimation program, wherein a total number of search results, a search word candidate generation process for extracting a part of a set of documents matching the search word, and a total number of search results and a set of documents, A search word candidate generation process that sets a list of search words that can be additional search words as search word candidates, a selection process that sets a search word corresponding to a selection signal obtained from a user from the search word candidates as an additional search word, A process of estimating the number of additional search words in the document set and outputting the estimated number of additional search words.

In the above, the first, fifth, and ninth aspects of the present invention use any part of the search result set that matches the search term, and select an arbitrary search term in the entire search result set that matches the search term. By estimating the percentage of search terms included, the change in the approximate number of search results when one search term is added to another search term is estimated without holding the entire search result set and without performing a re-search. It is possible to

Further, the second, sixth and tenth inventions generate and present a list of words that can be additional search words by using a part of a document set that matches the search words, and present the list to the user. Search terms to be added, and at the same time, add another search term to one search term by estimating the proportion of the added search term in the entire set of documents that match the search term The approximate change in the number of search results in the case can be estimated without holding the entire search result set and without performing a re-search, and at the same time, candidates for additional search words can be presented to the user.

Further, the third, seventh and eleventh inventions estimate a search result when another search word is added to a search word by using a part of a search result set that matches the search word. Accordingly, the approximate number of search results when another search word is added to a certain search word can be calculated without holding the entire search result set and without performing a re-search. In addition, the fourth and eighth
And a twelfth invention generates and presents a list of words that can be additional search words by using a part of a document set that matches the search word, and allows the user to select a search word to be added, By estimating the number of search results when a search term is added, the approximate number of search results when adding another search term to one search term is estimated without re-searching, and at the same time Word candidates can be presented to the user.

As described above, according to the present invention, only a part of the search result is held, and when an additional search word is given, instead of performing a re-search, the search result of the held part of the search result is
The ratio of those containing the additional search word is obtained, and the ratio of the additional search word in the entire search result set is calculated based on the value. Therefore, although the calculated value is not completely accurate, it is not necessary to hold the entire search result set instead.

[0021]

DESCRIPTION OF THE PREFERRED EMBODIMENTS First, as a first embodiment,
A description will be given of a case in which a part of a search result set that matches a search word is used to estimate the ratio of an arbitrary search word included in the entire search result set that matches the search word. FIG.
Shows the configuration of the first estimating device of the present invention.

The apparatus 100 shown in FIG.
1 and a sample search unit 102. The whole search unit 101 receives the search term 103, searches for data matching the search term 103 from all the data, and uses a part of the search term as a search result sample set 104 to make the sample search unit 1
02 is output.

The sample search unit 102 receives a search result sample set 104 and an additional search word 105 from the outside, and in the search result sample set 104,
The number of elements that match 5
A value obtained by dividing by the number of elements of 04 is output as an appearance rate 106. The appearance rate 106 calculated here is based on the search result sample set 104. However, if the accuracy of the appearance rate 106 is to be improved, the search result sample set 104 may be selected to be sufficiently large.

Next, as a second embodiment, a list of words that can be additional search words is generated and presented using a part of the document set that matches the search words, and the search words added by the user are added. And a case where the proportion of the added search term in the entire document set that matches the search term is estimated. FIG. 3 shows a second embodiment of the present invention.
1 shows the configuration of the estimation device.

The apparatus 200 shown in FIG.
1. It comprises a search word candidate generation unit 202, a search word selection unit 203, and a sample search unit 204. Overall search unit 201
Receives the search term 205, searches data matching the search term 205 from all data, and outputs a part thereof as a search result sample set 206.

The search term candidate generation unit 202 receives the search result sample set 206 and outputs a list of potential additional search terms 209 as search term candidates 207. The method of generating the search term candidate 207 differs depending on the type of data. For example, in the case of data to which a keyword has been assigned in advance, a set of keywords assigned to each of the search result sample sets 206 is used. be able to. If the data is text, the text can be divided into words by morphological analysis, and search term candidates 207 can be generated by an existing keyword estimation method (for example, TF * IDF value).

The search word selection section 203 searches for a search word candidate 207.
And a selection signal 208 from the user, and outputs a search word 205 corresponding to the selection signal 208 as an additional search word 209. The sample search unit 204 receives the search result sample set 206 and the additional search term 209, and divides the number of elements of the search result sample set 206 that match the additional search term 209 by the number of elements of the search result sample set 206. Is output as the appearance rate 210. That is,
Appearance rate = MT / T, where MT = the number of elements of the search result sample set that matches the additional search word, T = the number of elements of the search result sample set Next, as a third mode, a search result set that matches the search word A case where the number of search results when another search word is added to the search word is estimated using a part of the search word will be described.

FIG. 4 is a block diagram of a third estimating device according to the present invention. The estimating device 300 shown in FIG. 7 has the same configuration as that shown in FIG. 2 except that the entire search unit 301 outputs not only a set of search result samples but also the number of search results based on search words as the total number of search results 304. is there. The sample search unit 302 acquires the search result sample set 305 and the additional search term 306, and divides the number of elements of the search result sample set 305 that match the additional search term 306 by the number of elements of the search result sample set 305. Is output as the estimated number, and the appearance rate multiplied by the total number of search results 304 is output as the estimated number.

Next, as a fourth mode, a list of words that can be additional search words is generated and presented using a part of the document set that matches the search words, and the user selects the search words to be added. A description will be given of a case where the number of search results when a search term is added is estimated while the number of search results is estimated. FIG.
It is a lineblock diagram of the 4th presumption device of the present invention.

The estimating apparatus 400 shown in FIG. 3 has the same configuration as that shown in FIG. 3, except that the entire search unit 401 outputs not only a set of search result samples but also the number of search results by a search word as the total number of search results 406. It is to be. The sample search unit 404 obtains the search result sample set 407 and the additional search word 410 from the search word selection unit 403, and outputs the number of cases that match the additional search word 410 in the search result sample set 407 as the estimated number 411. I do.

Note that the search word selection unit 403 acquires the search word candidate 408 obtained from the search word candidate generation unit 402 and the selection signal 409 from the user, as in the configuration shown in FIG. A corresponding search term is selected from the search term candidates 408.

[0032]

Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 6 shows the configuration of the estimating device according to one embodiment of the present invention. The estimating device 500 shown in FIG. 5 is based on the fourth estimating device 400 shown in FIG.

The estimating apparatus 500 shown in FIG. 6 comprises a text database 501, a document word matrix generating section 502, a number estimating section 503, a search word candidate presenting section 504, and a search word selecting section 505. Text database 501
Searches a database when a search term 506 is given. If the number of results is small enough, terminate the search operation,
The result may be output, but if the number of cases is too large and narrowing is necessary, the text database 501 randomly extracts 50 cases from the search results and outputs the result as a document sample set 507. Also, the total number of search results 508 is output. The number of samples is not particularly limited to 50, but an optimum value may be selected according to the capacity of the computer. It is also assumed that the document sample set 507 has already been divided into words by morphological analysis. Usually, in a full-text search, a morphological analysis is often performed to create an index file. Therefore, it is considered that a morphological analysis of a document is performed before a search is performed.

The document word matrix generation unit 502 receives the document sample set 507 and counts the number of appearances of each word for each document. This operation is performed on the entire document sample set 507 to generate and output a document word matrix 509 as shown in FIG. Note that the document word matrix 509 is information that is required in common by the number-of-cases estimation unit 503 and the search word candidate presentation unit 504. Therefore, the document word matrix 509 is generated by providing the document word matrix generation unit 502. The unit 502 is a lower module of the sample search unit 404 and the search word candidate generation unit 402 in FIG. In the document word matrix 509 shown in FIG. 7, the rows and columns represent the document identifier and the search word list, and the value of the cell in the table indicates how many times the search word in the corresponding column is included in the document in the corresponding row. Or represents

The search word candidate presentation unit 504 receives the document word matrix 509, removes words that are inappropriate as search words 506, such as particles and pronouns, from the list, and how important any word is in any document. Is calculated as follows. Word importance = (number of occurrences of word in document) × log ｛50
/ (The number of documents in which the word appears)｝ The above equation is an index known as TF * IDF,
Is the number of elements in the document sample set 507. As can be easily understood, the word importance is high for words appearing intensively in a specific document, and conversely, the importance of words appearing in all documents is zero.

As a result of the above-described calculation, the search word candidate presentation unit 504 determines a word having a high weight as a search word candidate 510.
Output as The search word selection unit 505 selects the search word candidate 5
10 and output it to the screen to provide the user with additional search words 5
12 is selected. The user's selection is input as a selection signal 511. The search word selection unit 505 selects the selection signal 511
Is output as the additional search word 512.

The number-of-cases estimation unit 503 calculates the total number of search results 508
And the document word matrix 509 are received and held. Upon receiving the additional search word 512, the number estimating unit 503 examines the document word matrix 509 and counts the number of rows where the column of the additional search word 512 is not 0. Dividing this by the total number of rows in the matrix gives the appearance rate of additional search terms 512. If the appearance rate is multiplied by the total number of search results 508, the total number of search results 508 when the search is performed with the additional search words 512 in all documents can be output as the estimated number 513.

FIG. 8 shows an example of a screen of a search system using the embodiment of FIG. In the figure, the user performs a document search using the search term “movie”, and further searches for additional search terms 5.
As 12, an “introduction” (additional search word) is to be added. Hereinafter, how the information displayed on this screen is generated will be described with reference to FIG. 8 in FIG.
The lower part is a part for inputting a search word 506. At first, the user inputs "movie" here to perform a search. Search word 5
06 was sent to the text database 501, and as a result, as shown in the second row, 1234 documents were found. In the text database 501, 50 out of 1234 items are randomly extracted and sent to the document word matrix generation unit 502. The 50 documents are subjected to morphological analysis and decomposed into words, and output as a document word matrix 509.

The search word candidate presentation section 504 weights the words, extracts 15 words as shown in the third row of FIG. 8, and outputs them as search word candidates 510. Search term selection part 5
In step 05, the screen shown in FIG. 8 is presented to accept input from the user. Here, the user has specified "Introduction",
"Introduction" is added to the search term input field in the first row,
The additional search word 512 “introduction” is sent to the number estimation unit 503.

The number estimating unit 503 includes a document word matrix 509.
To find out the number of documents that include "Introduction". In this case, eight documents included "Introduction". this is,
This is 16% of the 50 reference sample sets. This is multiplied by the total number of search results 1234 to output 197 cases as the estimated number 513. This output corresponds to the second stage in FIG.

In order to output the estimated number 513 more accurately, the value of the estimated number may have a certain range (for example, 197 ± 20). In this case, the probability that the true value is in this range can be calculated statistically. It should be noted that the present invention is not limited to the above-described embodiment, but can be variously modified and applied within the scope of the claims.

[0042]

As described above, according to the first, fifth, and ninth aspects, the change in the approximate number of search results when another search word is added to a certain search word is determined by the entire search result set. , And can be estimated without re-searching. According to the second, sixth, and tenth aspects, the change in the approximate number of search results when another search word is added to a certain search word is determined without holding the entire search result set and performing a re-search. , And at the same time, a candidate for a search word to be added can be presented to the user.

According to the third, seventh and eleventh aspects of the present invention, the approximate number of search results when another search word is added to a certain search word can be re-searched without holding the entire search result set. Can be calculated without the need. According to the fourth, eighth, and twelfth inventions, the approximate number of search results when another search word is added to a certain search word is estimated without performing a re-search, and at the same time, candidates for additional search words are estimated. Can be presented to the user.

[Brief description of the drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a configuration diagram of a first estimation device of the present invention.

FIG. 3 is a configuration diagram of a second estimation device of the present invention.

FIG. 4 is a configuration diagram of a third estimation device of the present invention.

FIG. 5 is a configuration diagram of a fourth estimation device of the present invention.

FIG. 6 is an example of an estimation device according to an embodiment of the present invention.

FIG. 7 is an example of a document word matrix according to one embodiment of the present invention.

FIG. 8 is a screen example of a search system according to an embodiment of the present invention.

[Explanation of symbols]

100 Statistical estimation device for the number of search results, estimation device 101, 201, 301, 401 Overall search unit 102, 204, 302, 404 Sample search unit 103, 205, 303, 405, 506 Search terms 104, 206, 305, 407 Search result sample set 105, 209, 306, 410, 512 Additional search word 106 Appearance rate 110 Search result sample extraction means 120 Additional search word ratio calculation means 200, 300 Estimation device 202, 402 Search word candidate generation unit 203, 403, 505 Search word selection unit 207, 408, 510 Search word candidate 208, 409, 511 Selection signal 210 Appearance rate 304, 406, 508 Total number of search results 307, 411, 513 Estimated number 400 Estimating device 500 Estimating device 501 Text database 502 Literature word matrix Living Part 503 number estimation section 504 search word candidate presenting unit 507 literature sample set 509 literature word matrix

Continued on the front page (72) Inventor Kazuo Tanaka Nippon Telegraph and Telephone Corporation, 3-9-1-2 Nishishinjuku, Shinjuku-ku, Tokyo

Claims (12)

[Claims] In a data search system, a method for statistically estimating the number of search results to know how the number of search results changes when a search word is added to further narrow the search results, A statistic of the number of search results characterized by estimating, by using a part of a search result set matching a search term, a ratio of an arbitrary search term included in the entire search result set matching the search term. Estimation method. 2. A method for statistically estimating the number of search results in order to know how the number of search results changes when a search word is added to further narrow down search results in a data search system. Using a part of the document set that matches the search term, a list of terms that can be additional search terms is generated and presented, so that the user can select the additional search terms,
A statistical estimation method of the number of search results, characterized by estimating a ratio of an added search term included in a whole document set that matches the search term. 3. A method for statistically estimating the number of search results for knowing how the number of search results changes when a search word is added to further narrow the search results in a data search system, A statistical estimation method for search result training, comprising estimating the number of search results when another search word is added to the search word using a part of a search result set that matches the search word. 4. A method for statistically estimating the number of search results in order to know how the number of search results changes when a search term is added to further narrow down the search results in the data search system. Using a part of the document set that matches the search term, a list of terms that can be additional search terms is generated and presented, so that the user can select the additional search terms,
A statistical method for estimating the number of search results, comprising estimating the number of search results when a search word is added. 5. A statistical estimation device for the number of search results to know how the number of search results changes when a search word is added to further narrow down the search results in the data search system. A search result sample extracting means for extracting a part of the search result set that matches the search word; receiving the search result set and the additional search word; And an additional search word ratio calculating means for calculating and outputting the ratio of words containing the additional search word. 6. A data estimating apparatus for statistically estimating the number of search results to know how the number of search results changes when a search term is added to further narrow down the search results. A search result sample extracting means for extracting a part of a search result set matching the search word; and a list of possible additional search words from a part of the search result set extracted by the search result sample extracting means. A search word candidate generating means to be a candidate, a search word corresponding to a selection signal received from a user to select from the search word candidates, and a selection means to be an additional search word, and a part of the search result set A statistical unit for estimating the number of search results, comprising: an additional search word ratio calculating means for estimating and outputting a ratio of the search result matching the additional search word. 7. A statistical estimation device for the number of search results to know how the number of search results changes when a search word is added to further narrow down the search results in the data search system. A search word candidate generating means for extracting a total number of search results and a part of a set of search results that match the search word; and estimating an estimated number of cases including an additional search word in a part of the set of search results. And a means for calculating the number of additional search terms to be output. 8. A statistical estimation device for the number of search results to know how the number of search results changes when a search word is added to further narrow down the search results in a data search system. Search word candidate generation means for extracting the total number of search results and a part of the document set that matches the search term, and a list of search terms that can be additional search terms in the total number of search results and the document set A search word candidate generating unit that sets a search word as a search word candidate, a selection unit that sets a search word corresponding to a selection signal selected by a user from the search word candidates as an additional search word, and the addition in the document set A statistical device for estimating the number of search results, comprising: an additional search word number calculator for estimating and outputting the number of search words. 9. A program for statistically estimating the number of search results to know how the number of search results changes when a search term is added to further narrow down search results in a data search system. A retrieval result sample extraction process for extracting a part of a search result set that matches a search word; receiving the search result set and an additional search word, and using the part of the search result set to A storage medium storing a program for statistically estimating the number of search results, characterized by having a process of calculating a ratio of additional search words that calculates and outputs a ratio of the search result set that includes the additional search word. 10. In a data search system, a program for statistically estimating the number of search results to know how the number of search results changes when a search word is added to further narrow the search results is stored. A search result sample extraction process for extracting a part of a search result set that matches the search word, and a search result sample extracted from the search result sample extraction process may be an additional search word. A search word candidate generation process of using a list of things as a search word candidate; a selection process of selecting a search word corresponding to a selection signal received from a user from the search word candidates and setting it as an additional search word; And estimating a ratio of those that match the additional search word, and outputting the additional search word ratio calculation process. Results count storage medium storing a statistical estimation program. 11. In a data search system, a program for statistically estimating the number of search results to know how the number of search results changes when a search word is added to further narrow the search results is stored. A search medium candidate generating process for extracting a total number of search results and a part of a search result set that matches the search word; and including an additional search word in a part of the search result set. A storage medium storing a statistical estimation program for the number of search results, characterized by having an additional search word number calculation process for estimating and outputting the estimated number of items. 12. In a data search system, a program for statistically estimating the number of search results to know how the number of search results changes when a search word is added to further narrow the search results is stored. A search term candidate generation process that extracts the total number of search results and a part of the document set that matches the search term, and an additional search term in the total number of search results and the document set. A search word candidate generation process that uses a list of obtained search words as search word candidates; a selection process that uses a search word corresponding to a selection signal acquired from a user from the search word candidates as an additional search word; And a program for estimating the number of the additional search words in the program and outputting the number of additional search words. Media.