JP3564999B2 - Information retrieval device - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an information search apparatus capable of obtaining a bibliographic list at a high speed even when a plurality of search engines are used to search for information from digitized document data.
[0002]
[Prior art]
In recent years, with the spread of word processors and personal computers, a large amount of document information has been accumulated, and interest in a document database for searching document information as necessary has been increasing. Also, a full-text search method for searching the document information from the contents of the document without adding a keyword has attracted attention, and is used in a homepage search on the Internet. A search system using this full-text search method can be used by a user in a form connected to a server client or a WWW server. In such a search system, a method of processing user requests simultaneously by displaying a list of search results in units of dozens when displaying a list of search results instead of letting one user occupy the search system is adopted. ing. Further, the search result list is not simply output in the order of registration, but is digitized (score) by setting a reference between the document and the search condition, and the bibliographic list is ranked according to the obtained score. By doing so, it is possible to output a search result close to the user's request. Here, "quantification" means not only uniform numerical conversion but also numerical conversion by weighting words to be searched (for example, short sentences are weighted more than long sentences, etc.) and words. The meaning also includes. Also, the “bibliographic list” is difficult for the user to understand only by the document number, and thus means, for example, the title of a homepage or a URL (Universal Resource Locator).
[0003]
Hereinafter, a conventional information retrieval apparatus will be described.
FIG. 8 shows a configuration of a conventional information retrieval apparatus. In FIG. 8, 81-1, 81-2,..., 81-n are client units, 82 is a communication unit, 83 is a search engine unit, 84 is an index storage unit, and 85 is a bibliographic storage unit. It is.
[0004]
The operation of the information retrieval device configured as described above will be described below. First, each client unit 81-1, 81-2,..., 81-n sends a search request from the user to the communication unit 82. The communication unit 82 internally stores search requests from a plurality of client units and sends them to the search engine unit 83. The search engine unit 83 reads out the index information from the index storage unit 84, performs a high-speed search, and returns it to the communication unit 82. The communication unit 82 returns the number of search results to the client units 81-1, 81-2,..., 81-n. Also, the client unit 81-1, 81-2,..., 81-n sends to the communication unit 82 a request for a list of bibliographies hit in the search. The communication unit 82 sends a bibliography list request to the search engine unit 83. The search engine unit 83 reads the bibliographic list from the bibliographic storage unit 85, creates a bibliographic list, and returns it to the communication unit 82. The communication unit 82 returns the bibliography list to the client units 81-1, 81-2,..., 81-n.
[0005]
[Problems to be solved by the invention]
However, in the above-described conventional configuration, when the number of documents to be stored is increased by tens of millions, a single computer has a limit due to a decrease in search performance and a limitation of physical computer resources such as a hard disk and a memory. Yes, it was necessary to deal with multiple computers. However, when multiple computers are used, multiple search engines are used, and when an ordered bibliographic list is obtained, the communication load between the search engines increases, and the overall performance decreases. Had the problem of doing so.
[0006]
An object of the present invention is to solve the above-mentioned problems of the prior art, and to minimize the communication load even when an ordered bibliographic list is obtained with a configuration of a plurality of search engines.
[0007]
[Means for Solving the Problems]
In order to achieve this object, an information retrieval apparatus according to the present invention comprises, firstly, a plurality of search engine units each independently performing at least retrieval of document data, creation of a bibliographic list, and ordering of retrieval results according to a certain reference value. An index storage unit for storing search information for performing a search, a bibliographic storage unit for storing information for creating a bibliographic list, and an entire sort unit for ordering the entire search results of a plurality of search engine units. In addition, a bibliographic list of search results is obtained by obtaining, from each search engine unit, up to a predetermined ordered reference value from the beginning of the search results in the overall sort unit. With the above configuration, a bibliographic list can be created at high speed.
[0008]
Secondly, when the bibliographic list positioned after half of the entire search result is obtained, the entire sort unit obtains a reference value for ordering from the end of the search result of the search engine unit. With this, when a bibliography list is acquired by a plurality of search engine units, the entire sort unit does not extract all the scores of the search histories of the plurality of search engine units, but obtains the acquisition number from the beginning or end of the search result. By extracting a score partially according to the number of acquisitions, a bibliographic list can be created at high speed.
[0009]
Third, the entire sort unit divides the search result into n parts by the reference value for ordering (n ≧ 2), first obtains the number of records that are equal to or more than the lower limit of the divided reference values from each search engine, and By accumulating the number of cases within the range, it was possible to determine the position of the target bibliographic list. As a result, a bibliographic list can be created at high speed by obtaining the number distribution of scores from each search engine unit, calculating the position of the required search result from the number distribution again, and partially obtaining the score. Can be.
[0010]
BEST MODE FOR CARRYING OUT THE INVENTION
(Embodiment 1)
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings.
[0011]
FIG. 1 is a configuration diagram of an information search device according to an embodiment of the present invention. In FIG. 1, 11-1, 11-2,..., 11-n are client units, 12 is a communication unit, 13-1, 13-2,. 14 is an overall sort unit, 15 is a high score storage unit, 16 is an index storage unit, and 17 is a bibliographic storage unit.
[0012]
The operation of the information retrieval device configured as described above will be described. First, each of the client units 11-1, 11-2,..., 11-n transfers a search request from a user to the communication unit 12, and the communication unit 12 receives the search request from each of the search engine units. , 13-n read index information for search from the index storage unit 16 and perform search. The number of results is passed to the communication unit 12. The communication unit 12 sums up the number of search results of each of the search engine units 13-1, 13-2,..., 13-n and returns the total to the client unit. When a request for a bibliography list is received from the client units 11-1, 11-2,..., 11-n, the communication unit 12 obtains the number from the top of the bibliography list and obtains the number from the top of the bibliography list. Send the number to do. .., 13-n for the search engine units 13-1, 13-2,..., 13-n. The requested values (scores) are requested for ({acquisition start number} + {acquisition number} −1). The search engine units 13-1, 13-2,..., 13-n sort according to the scores, and return the scores to the overall sorting unit 14 by the requested number. In the overall sort unit 14, the obtained scores of the search engine units 13-1, 13-2,..., 13-n are rearranged in the order of scores, and the search engine units 13-1, 13-2,. ··· Find the start number of 13-n and the number of acquisitions. The whole sort unit 14 sends the obtained start number and the number of obtained records to each of the search engine units 13-1, 13-2,..., 13-n, and the search engine units 13-1, 13-2,. .., 13-n reads the document number of the search history from the start number, creates bibliographic contents from the bibliographic storage unit 17 from the document number, and sends it to the overall sorting unit 14. The whole sort unit 14 sorts the bibliography list by sorting the bibliographies from the contents of the bibliographies obtained from the search engine units 133-1, 13-2,..., 13-n and the information sorted in the score order. Create and return to communication unit 12. The communication unit 12 transfers the bibliography list to the client units 11-1, 11-2,..., 11-n, and the process ends.
[0013]
FIG. 2 shows an example of a search history that is information of search results stored in the search engine units 13-1, 13-2,..., 13-n. Is shown. 21 is a search history of the first search engine unit, 22 is a search history of the second search engine unit, and 23 is a search history of the third search engine unit. Each history indicates a state in which the history is sorted in descending order by score. When a bibliography list having an acquisition start number of 1 and the number of acquisitions of 10 is acquired for the search history, the entire sort unit 14 retrieves {1 + 10-1 = 10} score acquisition requests in each search. The search engine units 13-1, 13-2,..., 13-n extract the top ten scores from the search engine units 13, 24, 25, and 26, respectively. Reference numeral 24 denotes the score of the first search engine unit for ten searches, reference numeral 25 denotes the score of the second search engine unit for ten searches, and reference numeral 26 denotes the score of the third search engine unit for ten searches. As shown in the above example, a higher score can be obtained by each search engine.
[0014]
FIG. 3 is a diagram showing an example in which the scores of FIG.
In this figure, ten search histories of each of three search engines are obtained, and a total of 30 search histories are rearranged in the order of score. In this example, since the acquisition start number is 1 and the number of acquisitions is 10, the first to tenth search histories are bibliographic lists to be returned to the client units 11-1, 11-2,..., 11-n. FIG. 3 is a diagram in which the start numbers and the number of acquisitions for each of the search engine units 13-1, 13-2,..., 13-n are obtained from the search history. In this example, the first search engine unit has the start number 1 and the number of acquisitions is 2, the second search engine unit has the start number 1 and the acquisition number is 4, and the third search engine unit has the start number 1 and the acquisition number is 4. As shown in the above example, the scores are sorted by the overall sorting unit 14 and the start numbers and the number of acquired records for requesting the search engine units 13-1, 13-2,. You can ask.
[0015]
FIG. 4 is a diagram showing a process of acquiring a bibliography from the start number and the number of acquisitions of each search engine unit in FIG. 3 and creating a bibliography list.
[0016]
.., 13-n finds the document number from the search history, reads the bibliographic contents corresponding to the document number from the bibliographic storage unit 17, and sends the bibliographic contents to the overall sort unit 14. The bibliographies obtained by the respective search engine units 13-1, 13-2, ..., 13-n are transferred. The overall sorting unit 14 sorts the bibliographies in the order of score, creates a bibliographic list, and returns the bibliographic list to the communication unit 12.
[0017]
As shown in the above example, a bibliographic list can be created by creating bibliographic contents from the start number of each search engine and the number of acquisitions, and rearranging them again in the overall sort unit.
[0018]
As described above, according to the present embodiment, in an information search device including a plurality of search engine units, when acquiring a target bibliographic list from search results ordered by scores or the like, an overall sort unit and a high score storage By providing the section, a bibliographic list can be created at high speed only by acquiring a necessary search history partially.
[0019]
In the first embodiment, the client unit, the communication unit, the search engine unit, and the overall sorting unit may be performed by one computer, or may be performed by separate computers. In addition, it may be partially performed by one computer.
[0020]
Also, in the first embodiment, the communication unit holds the number of search results of each search engine unit and passes it to the overall sort unit, so that the overall sort unit performs bibliography on the search engine unit with zero search results. By not making an acquisition request, the communication time with zero search engine units can be reduced.
[0021]
(Embodiment 2)
Hereinafter, a second embodiment of the present invention will be described with reference to the drawings.
[0022]
FIG. 5 is a diagram showing an information search device according to one embodiment of the present invention.
In FIG. 5, 51-1, 51-2,..., 51-n are client units, 52 is a communication unit, 53-1, 53-2,. Reference numeral 56 denotes an index storage unit, and reference numeral 57 denotes a bibliographic storage unit, which has the same configuration as that of FIG. The difference from the configuration of FIG. 1 is that when the score information is acquired from the search history, the overall sorting unit 54 and the score storage unit 55 are not acquired and stored in descending order of the score, but by the position of the acquisition start number. The point is that it is possible to automatically select whether to acquire in the order of higher score or lower score.
[0023]
For example, when newspaper articles are not arranged in chronological order and a new article is to be acquired, retrieval may be more efficiently performed by acquiring from the end than acquiring from the beginning.
[0024]
The operation of the information retrieval device configured as described above will be described below.
First, the client unit 51-1, 51-2,..., 51-n transfers a search request from the user to the communication unit 52. , 53-2,..., 53-n, and the search engine units 53-1, 53-2,. The index information for the search is read out and searched, and the number of search results is passed to the communication unit 52. The communication unit 52 sums up the number of search results of each of the search engine units 53-1, 53-2,..., 53-n and returns the result to the client unit 51. When a request for a bibliography list is received, the communication unit 52 sends the bibliography list acquisition start number and the number of acquisitions to the overall sorting unit 54. If the acquisition start number is larger than the value obtained by dividing the total number of search results by 2 with respect to each of the search engine units 53-1, 53-2,. From the end of the search history, request that (｛the total number of search results｝-開始 acquisition start number｝-｛acquisition number｝ +2) (番 目 the total number of search results｝-｛acquisition start number｝ +1) be acquired I do. The search engine units 53-1, 53-2,..., 53-n sort according to the scores, and return the scores to the overall sorting unit 54 for the number of requests requested from the beginning or end of the sorted result. In the overall sorting unit 54, when the obtained scores of the search engine units 53-1, 53-2,..., 53-n are obtained from the top, the scores are rearranged in descending order and obtained from the end. In this case, the scores are rearranged in ascending order, and the start numbers and the number of obtained search engines 53-1, 53-2,..., 53-n are obtained. The whole sort unit 54 sends the obtained start number and the number of obtained records to each of the search engine units 53-1, 53-2,..., 53-n, and the search engine units 53-1, 53-2,. .., 53-n reads the document number of the search history from the start number, creates bibliographic contents from the bibliographic storage unit 57 from the document number, and sends it to the overall sort unit 54. The whole sort unit 54 sorts the bibliography list from the contents of the bibliographies obtained from the respective search engine units 533-1, 53-2,... Created and returned to communication unit 52. The communication unit transfers the bibliography list to the client units 51-1, 51-2,..., 51-n, and the process ends.
[0025]
As described above, by acquiring the score of the search history from the beginning or end according to the position where the bibliography list is acquired, the overall sorting unit reduces the number of scores to be transferred to the overall sorting unit. The number of cases to be retrieved is reduced, and a faster bibliographic list can be obtained.
[0026]
Although the search history is obtained from the end of the search history in the second embodiment, the search in the search engine unit may be performed from the descending order to the ascending order to obtain the search history from the start.
[0027]
(Embodiment 3)
Hereinafter, a third embodiment of the present invention will be described with reference to the drawings.
[0028]
FIG. 6 is a diagram showing an information retrieval device according to one embodiment of the present invention.
6, 61-n are client units, 62 is a communication unit, 63-1, 63-2, ..., 63-n is a search engine unit, Reference numeral 67 denotes an index storage unit, and reference numeral 67 denotes a bibliographic storage unit. The above is the same as the configuration shown in FIG. 1 is different from the configuration of FIG. 1 in an overall sort section 64, a score storage section 65, and a score distribution storage section 66. When a bibliographic list is acquired, the overall sort section 64 obtains scores and distribution information of the number of records by the search engine section 63-. , 63-n, and by summing (accumulating) the number of cases within each score range, the search engine units 63-1, 63-2,. It is possible to reduce the number of scores obtained from 63-n.
[0029]
The operation of the information retrieval device configured as described above will be described below.
First, the client unit 61-1, 61-2,..., 61-n transfers a search request from the user to the communication unit 62. When the search request comes, the communication unit 62 , 63-2,..., 63-n, and the search engine units 63-1, 63-2,. Then, the index information for the search is read out and searched, and the number of search results is passed to the communication unit 62. The communication unit 62 sums up the number of search results of each of the search engine units 63-1, 63-2,..., 63-n and returns the result to the client unit 61.
[0030]
When a request for a bibliographic list is received, the communication unit 62 sends the number from the head of the bibliographic list and the number of records to be obtained to the overall sorting unit 64. The whole sort unit 64 determines the number of cases equal to or larger than the lower limit of the score within each score range obtained by dividing the score from 0 to m into n, with the maximum value of the score being m, by the search engine units 63-1, 63-2,. , 63-n. The search engine units 63-1, 63-2,..., 63-n obtain the number of cases equal to or more than the lower limit value in each score range from the scores of the search history, and send them to the overall sort unit 64. The score distribution obtained from the entire sort unit or each search engine is stored in the score distribution storage unit 66, and the entire score distribution is created. With this, it is possible to know in which score range the acquisition start number of the bibliographic list is located, so that the score is again smaller than or equal to s for the search engine units 63-1, 63-2, ..., 63-n. And ({acquisition start number} − {number of items having a value equal to or more than the lower limit score of the score range higher by one} + {number of acquired items} −1) and the serial number are acquired and sent to the overall sorting section 64. The overall sorter 64 rearranges the obtained scores of the search engines 63-1, 63-2,..., 63-n in the order of the scores, and searches the search engines 63-1, 63-2,. .., 63-n start number and number of acquisitions. The overall sort unit 64 sends the obtained start number and the number of obtained records to each of the search engine units 63-1, 63-2,..., 63-n, and the search engine units 63-1, 63-2,. , 63-n reads the document number of the search history from the start number, creates bibliographic contents from the bibliographic storage unit 68 from the document number, and sends the bibliographic contents to the overall sort unit 64. The overall sort unit 64 sorts the bibliography list by rearranging the bibliographies from the bibliographic contents obtained from the search engine units 63-1, 63-2, ..., 63-n and the information rearranged in the score order. Created and returned to communication unit 62. The communication unit 62 transfers the bibliography list to the client units 61-1, 61-2,..., 61-n, and the process ends.
[0031]
FIG. 7 is a diagram showing an example in which a score distribution is created by the search engine unit. The numbers 71, 72, and 73 are obtained by the three search engines obtaining the number of cases equal to or more than the lower limit of each score range from the scores of the search history. Each score distribution is sent to the overall sort section 64, and the scores of the three search engine sections 63 are summed up by the overall sort section 64 within each score range as shown at 74. In the example of this figure, when the acquisition start number is 501 and the number of acquisitions is 20, since the score is 800 or more and 476, the score of the 501st is 801 or less. Therefore, the overall sorting unit 64 acquires 46 scores (501-476 + 20 + 1 = 46) and serial numbers for each search engine with a score of 801 or less. Further, the overall sorting unit 64 sorts the scores obtained from the respective search engines in descending order, and obtains 20 engine numbers, serial numbers, and numbers from the 34th (501st-476), thereby obtaining the target bibliography. The list can be obtained from each search engine unit 63.
[0032]
In the third embodiment, the number of divisions of the search history is set to n in the overall sort unit. However, n may be changed according to the number of search results. For example, when the number of search results is large, n is increased, and when the number is small, n is decreased. The number n of divisions may be changed by making the average number of records within the score range the same.
[0033]
Further, in the third embodiment, the overall sorting unit may combine the processes of the first and second embodiments according to the number of search results and the position where the bibliographic list is obtained. For example, when the number of search results is as small as about 100, the processing may be performed by the processing method of the first embodiment without acquiring the score distribution. Even when the number of search results is large, if the number is about 20 from the top, the processing may be performed by the processing method of the first embodiment without acquiring a score distribution.
[0034]
In the third embodiment, the maximum value of the score is m in the overall sorting unit. However, M may obtain the maximum value of the score at the same time as obtaining the number of cases from the search engine, and use the maximum value.
[0035]
In the first embodiment, a score obtained by quantifying the relationship between a search request and a document is used as a criterion for rearranging the search history. However, the score may be rearranged using numerical information such as date. It is needless to say that the use of this numerical information can be processed in the same manner in the second and third embodiments.
[0036]
【The invention's effect】
As described above, the present invention stores a plurality of search engine units that independently perform at least search of document data, creation of a bibliographic list, and ordering of search results according to a certain reference value, and search information for performing the search. An index storage unit, a bibliographic storage unit for storing information for creating a bibliographic list, and an overall sort unit for ordering the entire search results of the plurality of search engine units, wherein the overall sort unit starts the search results To a predetermined ordered reference value from each search engine unit, reducing the communication volume of the entire sort unit and each search engine unit to obtain a bibliography list, and bibliography at high speed even in the environment of multiple search engines This has the effect that a list can be obtained.
[0037]
Also, when obtaining the bibliographic list positioned after half of the entire search result, the overall sort unit obtains the ordering reference value from the end of the search result of the search engine unit. Creates a bibliographic list at high speed by extracting partial scores according to the acquisition number and the number of acquisitions from the beginning or end of the search results instead of extracting all the scores of the search history of multiple search engine parts It has the effect that it can be done.
[0038]
In addition, the search result is divided into n by the reference value for ordering, the number of cases that are equal to or more than the lower limit of the divided reference values is obtained from each search engine, and the number of cases is accumulated, thereby distributing the number distribution of scores to each. The bibliographic list can be created at high speed by acquiring from the search engine unit, recalculating the position of the necessary search result from the number distribution, and partially acquiring the score.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of an information search device according to a first embodiment of the present invention; FIG. 2 is a diagram illustrating an operation example of a search engine unit according to the first embodiment; FIG. FIG. 4 is a diagram showing an example. FIG. 4 is a diagram showing an example of creating a bibliographic list in the first embodiment. FIG. 5 is a configuration diagram of an information search device in the second embodiment of the present invention. FIG. 7 is a diagram showing an operation example of an overall sorting unit according to the third embodiment. FIG. 8 is a diagram showing the configuration of a conventional information search device.
11-1 11-2 11-n Client unit 12 Communication unit 13-1 13-2 13-n Search engine unit 14 Overall sort unit 15 High score storage unit 16 Index storage unit 17 Bibliographic storage unit 51-1 51-2 51 -N client unit 52 communication unit 53-1 53-2 53-n search engine unit 54 overall sort unit 55 score storage unit 56 index storage unit 57 bibliographic storage unit 61-1 61-2 61-n client unit 62 communication unit 63 -1 63-2 63-n Search engine unit 64 Overall sort unit 65 Score distribution storage unit 66 Score storage unit 67 Index storage unit 68 Bibliographic storage unit

Claims (4)

At least a plurality of search engine units that independently search document data, create a bibliography list, and order search results according to scores, create an index storage unit that stores search information for performing a search, and create a bibliography list A bibliographic storage unit for storing information for, and an overall sorting unit for ordering the entire search results of the plurality of search engine units,
In the overall sorting unit, the scores from the beginning of the search result to the number obtained from the acquisition start number and the number of acquisitions are obtained from each of the search engine units, and the obtained scores are merged to obtain the scores from each of the search engine units. By determining the position and number of bibliographic lists to be obtained, the overall sort unit requests bibliographic contents from the search engine unit according to the determined position and number of bibliographic lists, and creates a bibliographic list. Search device. The system according to claim 1, further comprising: a client unit that processes the search request; and a communication unit that transfers the search request from the client unit to the search engine unit and the overall sort unit and returns a search result to the client unit. 1. The information retrieval device according to 1. The overall sort unit, when acquiring a bibliographic list positioned after half of the entire search result, the number of search results from the end of the search results of each search engine unit and the number of search results obtained from the number of search results and the acquisition start number to obtain a score from each of the search engine unit, by acquiring the bibliography list indexing the position and number of bibliographic list acquired from each of the search engine by using the score, the position and number of bibliographic list of indexing 2. The information retrieval apparatus according to claim 1, wherein the entire sort unit requests the search engine unit for bibliographic contents in accordance with the following, and creates a bibliographic list . The overall sorting unit divides the score of the search result into n (n ≧ 2), acquires the number of cases for each divided score from each search engine, stores it in the score distribution storage unit, and accumulates the number of stored cases. By obtaining the score of the corresponding portion from each search engine and storing it in the score storage unit, merging the stored scores and calculating the position and number of bibliographic lists obtained from each search engine unit , 2. The information search apparatus according to claim 1, wherein the entire sort unit requests the search engine unit for bibliographic contents in accordance with the position and the number of cases, and creates a bibliographic list .

Images