JP4922240B2 - Retrieval processing apparatus, method, and program for selectively applying pseudo feedback processing in web retrieval - Google Patents

Description

  The present invention relates to a search processing apparatus, method, and program that apply pseudo feedback processing.

  Conventionally, in the field of information retrieval, the technology of pseudo feedback processing has been applied, and its effectiveness has been confirmed particularly in evaluation using a large-scale test collection (see Non-Patent Documents 1 and 2).

  However, in the Web search, the technique of pseudo feedback processing is rarely used. The reason for this is that the pseudo-feedback processing increases the search processing load due to the term addition. However, with the recent enhancement of hardware performance and the like, it has become possible to execute high-load search processing at a relatively low cost.

Further, as a problem of pseudo feedback processing, in web search, what is improved by pseudo feedback processing is mainly the reproduction rate of search results, and the idea that it is not effective in web search where the highest search accuracy is important. There is. Particularly in web search, search requests such as navigation-oriented and transaction-oriented are frequently used, and in order to provide a web search service, it is possible to present useful search results to the top for such search requests. It is an important element.
Fujita, S .; (2005). A decade after TREC-4: NTCIR-5 CLIR-J-experiments at Yahoo! Japan. In, Proceedings of NTCIR Workshop 5 Meeting, (pp. 130-137). Fujita, S .; (2007). Technology survey and invalidity search: a comparable study of differential tasks for Japan patent document retrial. Information Processing and Management: an International Journal, 43 (5), 1154-1172.

  However, even in the Web search, there is an information-oriented search request similar to a normal information search. In this case, the reproduction rate of the search result is also an important factor. It is disclosed in Non-Patent Documents 1 and 2 that pseudo feedback processing can be an effective technique.

  Accordingly, an object of the present invention is to provide a search processing apparatus that selectively applies a pseudo feedback process to an information-oriented search request.

  The present invention provides the following solutions.

  (1) A search processing device that executes a search of a Web document by a search engine, and a receiving unit that receives an input query that is an input to the search engine, and the received input query is based on a predetermined condition A determination unit configured to determine whether the request is an information-oriented search request for acquiring a plurality of Web documents related to an input query as a search result, and based on a result determined by the determination unit, a plurality of A selection unit that selects different search processes, a search execution unit that applies the search process selected by the selection unit, and executes a search for a Web document according to the input query by the search engine, and the search execution unit Search result acquisition means for acquiring a searched Web document as a search result, wherein the search execution means includes at least a pseudo-feature. Back search processing and normal search processing by the search engine, and the selection means obtains a plurality of Web documents related to the input query as search results as a result of the judgment by the judgment means. The pseudo-feedback search process is selected, and the search execution means applies the pseudo-feedback search process to the Web corresponding to the input query by the search engine. A search processing apparatus for performing a search for a document.

  According to the configuration of (1), the search processing device accepts an input query that is an input to the search engine, and based on a predetermined condition, the accepted input query uses a plurality of Web documents related to the input query as search results. A search engine that determines whether the request is an information-oriented search request that needs to be acquired, selects a plurality of different search processes based on the determined result, and applies the selected search process Thus, a search for a Web document corresponding to the input query is executed, and the searched Web document is acquired as a search result. The search processing apparatus includes at least a pseudo feedback search process and a normal search process by the search engine. As a result of the determination, the input query acquires a plurality of Web documents related to the input query as a search result. If the request is an information-oriented search request, the pseudo-feedback search process is selected, and the search execution means applies the pseudo-feedback search process and executes a search for the Web document according to the input query by the search engine. .

  Thus, the search processing device can execute a search process according to the search request for the input query by the search engine, and thus can provide the user with a search result according to the search request for the input query. In addition, the search processing device applies a pseudo-feedback search process and executes a search using the input query and the added feature terms as a query, thereby obtaining a search result with a high reproducibility for an information-oriented search request. Can be provided.

  (2) When the selection unit determines that the input query is a search request that does not require a plurality of Web documents related to the input query as search results, the normal search is performed. The search processing device according to (1), wherein a search is selected, and the search execution unit applies the normal search process and executes a search for a Web document according to the input query by the search engine. .

  According to the configuration of (2), the search processing device performs normal search when the input query is a search request that does not require a plurality of Web documents related to the input query to be acquired as search results. A process is selected, a normal search process is applied, and a search engine executes a Web document search according to the input query.

  As a result, the search processing device performs a search by applying the normal search process, thereby obtaining suitable search results for non-information-oriented search requests (navigation-oriented and transaction-oriented search requests). Can be provided.

  (3) storage means for storing a query; extraction means for extracting a predetermined query from the storage means; machine learning means for performing machine learning of a classifier using the extracted query; A classifying unit that classifies the extracted query by a classifier; and a query storage unit that stores the classified query. The determination unit includes the classified query and the query as the predetermined condition. The search processing apparatus according to (1), wherein the determination is made based on whether or not the received input query matches.

  According to the configuration of (3), the search processing device classifies the query by the machine-learned classifier, and determines whether or not the classified query matches the received input query as a predetermined condition. Do.

  Thereby, the search processing device can determine whether or not the input query is an information-oriented search request based on the classified query.

  (4) The determination unit is a user who inputs the input query as the predetermined condition, out of the received input query and a Web document included in a search result searched by the search engine according to the input query. The search processing device according to (1), wherein a determination is made based on whether or not a value calculated based on a Web document selected by a terminal device operated by the user exceeds a predetermined threshold.

  According to the configuration of (4), the search processing device inputs an input query as a predetermined condition among the received input query and the Web document included in the search result searched by the search engine according to the input query. The determination is made based on whether or not the value calculated based on the Web document selected by the terminal device operated by the user exceeds a predetermined threshold.

  Thereby, the search processing apparatus can determine whether or not the input query is an information-oriented search request based on whether or not the calculated value exceeds a predetermined threshold.

  (5) A search processing apparatus that executes a search for a Web document by a search engine, and accepting means for accepting an input query that is an input to the search engine; and accepting the search by the search engine by applying normal search processing A first search execution means for executing a search for a Web document according to an input query; a first search result acquisition means for acquiring the Web document searched by the first search execution means as a first search result; and a pseudo feedback search. A second search execution unit that executes a search of a Web document according to the received input query by the search engine by applying a process, and acquires the Web document searched by the second search execution unit as a second search result Second search result acquisition means, a search result including a predetermined number of Web documents among the acquired first search results, and the capture A search result merging unit that merges the search results of the predetermined number of Web documents with the search results excluding the duplicate Web document search results into one search result. A search processing apparatus characterized by that.

  According to the configuration of (5), the search processing device receives an input query that is an input to the search engine, and executes a search for a Web document according to the input query received by the search engine by applying a normal search process. The retrieved Web document is acquired as the first search result, the pseudo-feedback search process is applied, the Web document is searched according to the input query received by the search engine, and the searched Web document is second searched. As a result, a search result including a predetermined number of Web documents among the acquired first search results and a Web document that overlaps with a search result of the predetermined number of Web documents among the acquired second search results. The search results excluding the search results are merged into one search result.

  Thereby, the search processing device can provide the user with the search result obtained by the normal search process for the search result of the first page, which is highly likely not to be an information-oriented search request, for example. The search processing device can provide the user with the search results obtained by the pseudo feedback search process for the search results for the second and subsequent pages that are likely to be information-oriented search requests.

  (6) A search processing method for executing a search for a Web document by a search engine, wherein a reception step of receiving an input query that is an input to the search engine, and the received input query is based on a predetermined condition A determination step for determining whether or not an information-oriented search request for acquiring a plurality of Web documents related to an input query as a search result, and a plurality of determinations based on the result determined by the determination step A selection step of selecting a different search process, a search execution step of applying a search process selected in the selection step and executing a search of a Web document according to the input query by the search engine, and the search execution step A search result acquisition step of acquiring the searched Web document as a search result, The line step includes at least a pseudo feedback search process and a normal search process by the search engine. As a result of the determination in the determination step, the input query includes a plurality of Web documents related to the input query. In the case of an information-oriented search request that requires acquisition as a search result, the pseudo feedback search process is selected, and the search execution step applies the pseudo feedback search process to the search engine by the search engine. A search processing method characterized by executing a search for a Web document according to an input query.

  (7) A computer program executed by a computer of a search processing apparatus that executes a search for a Web document by a search engine, based on a reception step for receiving an input query that is an input to the search engine, and a predetermined condition A determination step for determining whether or not the received input query is an information-oriented search request for requesting acquisition of a plurality of Web documents related to the input query as a search result; and the determination step A selection step of selecting a plurality of different search processes based on the results of the search, and a search execution in which the search engine is applied to search the Web document according to the input query by applying the search process selected in the selection step And a Web document searched by the search execution step A search result acquisition step to be acquired, wherein the search execution step includes at least a pseudo feedback search process and a normal search process by the search engine, and the selection step is a result determined by the determination step. , If the input query is an information-oriented search request that requires a plurality of Web documents related to the input query to be acquired as search results, the pseudo feedback search process is selected, and the search execution step includes: A computer program that causes a computer to execute search of a Web document according to the input query by the search engine by applying the pseudo feedback search processing.

  According to the present invention, it is possible to provide a search processing device that selectively applies a pseudo feedback process to an information-oriented search request.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[Definition of terms]
“Web document” means a text document composed of text on the WWW (World Wide Web), represented by a Web page, an image not including text, a mixed document in which images and text are mixed, video, audio , And the like.

The “information-oriented search request” is a search for requesting acquisition of one or a plurality of Web documents related to a specific topic. For example, when it is desired to collect various information related to “automobile”, this search request is made.
The “navigation-oriented search request” is a search requesting to reach a specific Web document (or a typical Web document of a certain object). For example, a search for the purpose of reaching a certain university homepage corresponds to this search request.
A “transaction-oriented search request” is a search requesting to reach a Web document (online shopping, various services mediated by the Web, etc.) that involves interaction. For example, a search for searching for a hotel that meets one's request from hotels in Tokyo and making a reservation corresponds to this search request.

[Functional configuration of search system 1]
FIG. 1 is a block diagram showing a search system 1 according to an embodiment of the search processing apparatus of the present invention. The search server (search processing device) 10, the terminal device 20, and the content server 30 are connected via a communication line 40. A user of the terminal device 20 accesses the search server 10 and acquires a search result by inputting a query (search keyword) for reaching a desired Web document to a predetermined search engine. The user selects a URL listed in the search result and browses a Web document managed by the content server 30.

  The terminal device 20 is typified by a terminal device such as a personal computer, a mobile phone, a PHS (Personal Handy-phone System: registered trademark), a mobile, a notebook PC, or a PDA (Personal Digital Assistant), and performs wired or wireless communication. Any information can be used as long as the information transmitted from the search server 10 can be referred to via the communication line 40 to be used. Further, the search server 10 may be configured to include a Web server, a DB server, and an application server as necessary, and may be configured with one server or different servers.

[Hardware Configuration of Search Server 10]
FIG. 2 is a diagram illustrating an example of a hardware configuration of the search server 10 according to an embodiment of the present invention. As shown in FIG. 2, a CPU 11 (a plurality of CPUs such as a CPU 12 may be added in a multiprocessor configuration), a communication I / F (I / F: interface) 13, a main memory 14, a display device 15, an I / O The search server 10 is configured by a controller 16 and an input device 17 such as a keyboard and a mouse connected by a bus line BUS.

  The CPU 11 (CPU 12) is a part that controls the search server 10 in an integrated manner. The CPU 11 (CPU 12) cooperates with the above-described hardware by appropriately reading and executing various programs stored in the hard disk 18a, which will be described later. Various functions are realized.

  The communication I / F 13 is a network adapter when the search server 10 receives various input information from the terminal device 20 via the communication line 40 and transmits display contents to the display device 15. The communication I / F 13 may include a modem, a cable modem, and an Ethernet (registered trademark) adapter.

  The display device 15 includes a display device such as a cathode ray tube display device (CRT) or a liquid crystal display device (LCD). The I / O controller 16 can be connected to a storage device 18 such as a hard disk 18a and a semiconductor memory 18b. The input device 17 accepts input by the administrator of the search server 10. The hard disk 18a stores various programs for causing the hardware to function as the search server 10, a program for executing the functions of the present invention, a table to be described later, and the like. The search server 10 can also use a hard disk (not shown) separately provided as an external storage device.

  In the above example, the search server 10 has been mainly described. However, the functions described above can also be realized by installing a program in a computer and operating the computer as a server device. Therefore, the functions realized by the search server 10 described as an embodiment in the present invention can be realized by executing the above-described method by the computer, or by introducing the above-described program into the computer and executing it. It is feasible.

<First Embodiment>
[Search server functional configuration]
FIG. 3 is a functional block diagram of the search server 10 which is an embodiment of the search processing apparatus of the present invention. The search server 10 includes a click data storage unit 111, a determination condition storage unit 112, an extraction unit 118, a machine learning unit 119, a classification unit 120, a query table creation unit 121, an entropy calculation unit 122, and a query reception. Unit 123, determination unit 124, selection unit 125, search execution unit 126, search result acquisition unit 127, and search result transmission unit 128.

  The click data storage unit 111 and the determination condition storage unit 112 are mainly configured by the storage device 18. The query table creation unit 121, entropy calculation unit 122, query reception unit 123, determination unit 124, selection unit 125, search execution unit 126, search result acquisition unit 127, and search result transmission unit 128 are mainly executed by the CPU 11 (CPU 12). Composed.

  The click data accumulating unit 111 selects (clicks) a URL selected (clicked) by the terminal device 20 operated by the user among the query and a URL (Uniform Resource Locator) of the search result searched by the search engine using the query as a search keyword. (Hereinafter referred to as “click URL”) in association with each other and stored as history data. The history data is stored accumulatively as a click data table 111a, for example, as shown in FIG. The click data table 111a illustrated in FIG. 4 stores a query, a click URL, a search rank of the click URL, an IP address of a terminal that has selected the click URL, a selected time, and a session in association with each other.

  The determination condition storage unit (predetermined condition) 112 stores a query table 113 and a threshold value 114 that are used for the processing of the determination unit 124. FIG. 5 shows an example of the query table 113. The query table 113 stores a query and a number for organizing the query. For example, the number 1 is the query “total”. The number 2 is the query “A × Z”. Number 3 is the query “menu”. Number 4 is the query “Ma Δ”. The query table 113 is created by the query table creation unit 121 and stores a predetermined query that is an information-oriented search request.

  The extraction unit 118 refers to the click data table 111a, calculates a frequency indicating the number of times the click URL has been selected, and extracts a query with a high frequency. Then, the extraction unit 118 extracts a learning feature from the extracted query.

The machine learning unit 119 performs machine learning of the classifier using a part of the query extracted from the click data table 111a by the extraction unit 118.
The classification unit 120 automatically classifies all the remaining queries extracted from the click data table 111 a by the extraction unit 118 using the classifier learned by the machine learning unit 119. At this time, the classification unit 120 classifies whether or not all the remaining queries extracted using the classifier are information-oriented search requests.

  The query table creation unit 121 stores a query that is an information-oriented search request among the queries classified by the classification unit 120 as the query table 113 in the determination condition storage unit 112.

  The entropy calculation unit 122 refers to the click data table 111a, and when there is a query that matches the input query received by the query reception unit 123, the query (input query) and the click associated with the query Based on the URL, the entropy of the click URL for the query is calculated.

  For example, the probability P (http: //xxx.yyy) that a certain URL “http: //xxx.yyy.zzz” is clicked by aggregating history data for a specific query “car” in the click data table 111a. .Zzz | car).

ただし、０・ｌｏｇ０＝０とする。ここで、ｑはクエリ、ｕはクエリｑのクリックＵＲＬである。また、クリックＵＲＬがただ一つである場合には、エントロピーＨ（Ｕ）は、０になり、全てのクリックＵＲＬが等確率でクリックされた場合、ｌｏｇ（｜Ｕ｜）になる。 In this way, the entropy of the click URL for the input query can be obtained from the probability P (u | q) that a certain click URL (u) is clicked for the input query (q). And the entropy calculating part 122 calculates entropy H (U) based on the definition of the normalized entropy shown to the following formula | equation (1).

However, 0 · log0 = 0. Here, q is a query, and u is a click URL of the query q. In addition, when there is only one click URL, entropy H (U) becomes 0, and when all click URLs are clicked with equal probability, it becomes log (| U |).

  The query receiving unit 123 receives an input query, which is an input to the search engine, input by a user who operates the terminal device 20 via the communication line 40.

  Whether the determination unit 124 is an information-oriented search request for requesting that the input query received by the query reception unit 123 be acquired as a search result from a plurality of Web documents related to the input query based on a predetermined condition. Make a decision.

  Specifically, the determination unit 124 determines whether the query stored in the query table 113 of the determination condition storage unit 112 matches the input query received by the query reception unit 123 as a predetermined condition. .

  In addition, the determination unit 124 inputs an input query among the input query received by the query receiving unit 123 and the Web document included in the search result searched by the search engine according to the input query as a predetermined condition. The determination may be made based on whether or not the value of entropy H (U) calculated by the entropy calculation unit 122 exceeds the threshold value 114 based on the Web document selected by the terminal device 20 operated by the user. Good.

  The selection unit 125 selects a plurality of different search processes based on the result determined by the determination unit 124. Specifically, when the determination unit 124 determines that the input query is an information-oriented search request that requires a plurality of Web documents related to the input query to be acquired as search results, as a result of the determination by the determination unit 124. Selects pseudo feedback search processing. In addition, when the determination unit 124 determines that the input query is not an information-oriented search request that requires a plurality of Web documents related to the input query to be acquired as a search result, the selection unit 125 performs normal search. Select a process.

  The search execution unit 126 has a search engine and applies the search process selected by the selection unit 125 to execute a search for a Web document corresponding to the input query. Specifically, when the selection unit 125 selects the pseudo feedback search process, the search execution unit 126 applies the pseudo feedback search process and executes the search of the Web document according to the input query by the search engine. .

  In addition, when the normal search process is selected by the selection unit 125, the search execution unit 126 applies the normal search process and executes a search for the Web document according to the input query by the search engine. A publicly known search engine can be used.

The search result acquisition unit 127 acquires the Web document searched by the search execution unit 126 as a search result.
The search result transmission unit 128 transmits the search result acquired by the search result acquisition unit 127 to the user terminal device 20 via the communication line 40. Thereby, the user who operates the terminal device 20 can acquire the search result according to the query.

[Query table creation process]
FIG. 6 is a flowchart showing the flow of processing for creating the query table 113.

  Step S1: The extraction unit 118 refers to the click data table 111a, calculates a frequency indicating the number of times the click URL has been selected, and extracts learning features from about one million queries having a high frequency. Specifically, for example, the extraction unit 118 applies the following (a) to (c) as learning features. (A) The average number of click pages per session for the same session (cut by an appropriate time interval for the click operation of the same user). The significance is that the number of clicked pages is considered to be large in information-oriented search requests. (B) Search URL position distribution of the click URL for the query. The significance of this is that in information-oriented search requests, it is considered that there are many pages that have been searched and clicked down to the lower rank of the search rank. (C) Whether there is access to a specific domain. The significance is that, for example, when a URL described in the Wiki encyclopedia of Wikipedia (http://wikipedia.org/) is clicked, there is a high possibility that the request is an information-oriented search request.

  Step S2: The machine learning unit 119 performs machine learning of the classifier using a part of the query extracted in step S1. Specifically, for example, the machine learning unit 119 manually performs machine learning of the binary classifier using about 10,000 of the extracted queries as training data. The binary classifier is a classifier for automatically determining whether the request is an information-oriented search request, and performs machine learning using an SVM (support vector machine) or the like.

  Step S3: The classification unit 120 automatically classifies all the remaining queries extracted in Step S1 by the binary classifier that performed machine learning in Step S2. Thereby, it is determined whether or not all the remaining queries extracted in step S1 are information-oriented search requests.

  Step S4: The query table creation unit 121 creates a query table 113 by listing queries that are information-oriented search requests among the queries automatically classified in step S3, and stores them in the determination condition storage unit 112. .

  In this way, by creating a query table 113 by listing the queries that are information-oriented search requests, the search server 10 is based on the query table 113 and the input query is an information-oriented search request. It can be determined whether or not.

[Main processing]
FIG. 7 is a flowchart showing a flow of main processing of the search server 10 of the present embodiment.
Step S <b> 11: The query receiving unit 123 receives an input query input by a user operating the terminal device 20 via the communication line 40.

  Step S12: The determination unit 124 determines whether or not the query stored in the query table 113 of the determination condition storage unit 112 matches the input query received by the query reception unit 123, ie, the input query received in step S11. It is determined whether or not the query is an information-oriented search request. If the query stored in the query table 113 matches the input query received by the query receiving unit 123 (Yes), the process proceeds to step S13. On the other hand, if the query stored in the query table 113 does not match the input query received by the query receiving unit 123 (No), the process proceeds to step S14.

Step S13: Since the input query is determined to be an information-oriented search request in step S12, the selection unit 125 selects a pseudo feedback search process as a search engine process.
Step S14: Since it is determined in step S12 that the input query is not an information-oriented search request, the selection unit 125 selects the normal search process as the search engine process.

Step S15: The search execution unit 126 applies a pseudo feedback search process described with reference to FIG. 8 to be described later, and executes a search for a Web document according to the input query by the search engine.
Step S16: The search execution unit 126 applies a normal search process described with reference to FIG. 9 described later, and executes a search for a Web document according to the input query by the search engine.

Step S17: The search result acquisition unit 127 acquires the Web document searched in step S15 or step S16 as a search result.
Step S18: The search result transmitting unit 128 transmits the search result acquired in step S17 to the terminal device 20 operated by the user via the communication line 40.

  As described above, when the input query is an information-oriented search request, the search server 10 executes a search by applying the pseudo-feedback search process. When the input query is not an information-oriented search request, the search server 10 Perform search processing. Therefore, the search process according to the input query search request can be executed by the search engine, so that the search result according to the input query search request can be provided to the user.

  Further, the search server 10 determines whether or not the input query is an information-oriented search request based on whether or not the query automatically classified by the classifier with high classification accuracy learned by machine matches the input query. Therefore, the input query can be suitably determined.

[Pseudo feedback search processing]
FIG. 8 is a flowchart showing a flow of processing when executing the pseudo feedback search processing.
Step S21: The search execution unit 126 uses the search engine to execute a preliminary Web document search (hereinafter also referred to as a pilot search) using the input query received in step S1 of FIG.
Step S22: The search execution unit 126 acquires the search result of the Web document searched in step S21.

Step S23: The search execution unit 126 extracts a characteristic term (characteristic term) from the search result acquired in step S22.
Step S24: The search execution unit 126 adds the feature terms extracted in step S23 to the input query.

Step S25: The search execution unit 126 executes a search for a Web document according to the query by the search engine using the input query and the feature term added in step S24 as a query.
Step S26: The search execution unit 126 acquires the search result obtained by executing the search in step S25, and ends the process of this flowchart.

  In this way, the search server 10 executes a search using the input query and the added feature term as a query by applying the pseudo feedback search process, thereby performing a search with a high reproducibility for an information-oriented search request. Results can be provided.

[Normal search processing]
FIG. 9 is a flowchart showing the flow of processing when executing normal search processing.
Step S31: The search execution unit 126 is executed by the search engine using the input query received in step S1 of FIG.
Step S32: The search execution unit 126 acquires the search result obtained by executing the search in step S31, and ends the process of this flowchart.

  As described above, the search server 10 executes the search by applying the normal search process, so that the search server 10 is mainly the highest for non-information-oriented search requests (navigation-oriented or transaction-oriented search requests). It is possible to provide a search result with high search accuracy.

<Second Embodiment>
Next, a second embodiment of the present invention will be described. About 2nd Embodiment, a different point from 1st Embodiment is mainly demonstrated, the same code | symbol is attached | subjected about the structure similar to 1st Embodiment, and description is abbreviate | omitted. The search server 10 in the second embodiment has the same functional configuration as that of the search server 10 in the first embodiment, and a description thereof will be omitted.

  Next, the flow of processing of the search server 10 in the second embodiment will be described with reference to FIG.

  FIG. 10 is a flowchart showing a flow of main processing of the search server 10 of the present embodiment. In the flowchart shown in FIG. 10 of the second embodiment, Step S41 and Steps S44 to S49 are the same as Step S1 and Step S13 to Step S18 of the first embodiment, and thus description thereof is omitted.

  Step S42: The entropy calculation unit 122 refers to the click data table 111a, and if there is a query that matches the input query received in step S41, this query, the click URL associated with this query, and the expression Based on (1), the entropy H (U) of the click URL for the input query is calculated.

  Step S43: The determination unit 124 determines whether or not the value of the entropy H (U) calculated in step S42 exceeds the threshold value 114. If the value of entropy H (U) exceeds the threshold value 114 (Yes), the process proceeds to step S44. On the other hand, when the value of entropy H (U) does not exceed the threshold value 114 (No), the process proceeds to step S45. The threshold value 114 can be adjusted to an optimum value while actually operating. Further, when the value of entropy H (U) is 1.0 or less, there is a high possibility that it is a navigation-oriented search request, and when it is 2.0 or more, it is an information-oriented search request. Probability is high.

  As described above, when the input query is an information-oriented search request, the search server 10 executes a search by applying the pseudo-feedback search process. When the input query is not an information-oriented search request, the search server 10 Perform search processing. Therefore, the search process according to the input query search request can be executed by the search engine, so that the search result according to the input query search request can be provided to the user.

  Further, the search server 10 determines whether or not the input query is an information-oriented search request depending on whether or not the value of entropy H (U) exceeds the threshold value 114. Therefore, it is possible to suitably determine whether or not the input query is an information-oriented search request by appropriately adjusting the threshold value 114.

<Third Embodiment>
Next, a third embodiment of the present invention will be described. About 3rd Embodiment, a different point from 1st Embodiment and 2nd Embodiment is mainly demonstrated, the same code | symbol is attached | subjected about the structure similar to 1st Embodiment and 2nd Embodiment, and description is abbreviate | omitted. To do. The third embodiment is different from the first embodiment and the second embodiment in that a search is executed by a search engine using both a normal search process and a pseudo feedback search process.

  FIG. 11 is a functional block diagram of the search server 10 in the third embodiment. The search server 10 includes a query reception unit 123, a search result acquisition unit 127, a search result transmission unit 128, a search execution unit 131, and a search result merging unit 132.

  The query reception unit 123, the search result acquisition unit 127, and the search result transmission unit 128 have the same configurations as those of the first embodiment and the second embodiment shown in FIG.

  The search execution unit 131 executes a search by the search engine by applying a normal search process and a pseudo feedback search process to the input query received by the query reception unit 123.

  The search result merging unit 132 merges the search results obtained by applying the normal search process and the pseudo feedback search process acquired by the search result acquisition unit 127 into one search result. Specifically, the search result merging unit 132 acquires a search result including a predetermined number of Web documents among the search results to which the normal search process acquired by the search result acquisition unit 127 is applied, and the search result acquisition unit 127. Among the search results to which the pseudo feedback search process is applied, the search results of a predetermined number of Web documents and the search results excluding the search results of overlapping Web documents are merged into one search result.

FIG. 12 is a flowchart showing a flow of main processing of the search server 10 of the present embodiment.
Step S51: The query receiving unit 123 receives an input query input by a user who operates the terminal device 20 via the communication line 40.

Step S52: The search execution unit 131 executes a search by the search engine by applying a normal search process to the input query received in step S51. The normal search process is the same as the process shown in steps S31 to S32 in FIG.
Step S53: The search execution unit 131 executes a search by the search engine by applying a pseudo feedback search process to the input query received in step S51. Note that the pseudo feedback search process is the same as the process shown in steps S21 to S26 of FIG.

Step S54: The search result acquisition unit 127 acquires the Web document searched in step S52 as a search result.
Step S55: The search result acquisition unit 127 acquires the Web document searched in step S53 as a search result.

  Step S56: The search result merging unit 132 merges the search results acquired in Step S54 and Step S55. Specifically, out of the search results acquired in step S54, the search results of a predetermined number of Web documents (for example, the search result of the first page displayed on the web browser of the terminal device 20) and the search results acquired in step S55. Of the search results, the search results excluding a part that overlaps the search results of the predetermined number of Web documents that have been searched in step S52 (for example, the search results for the second and subsequent pages displayed on the Web browser of the terminal device 20). The search results of the first page and the search results that are the same as those of the first page are excluded).

  Step S57: The search result transmission unit 128 transmits the search result merged in Step S56 to the terminal device 20 operated by the user via the communication line 40.

  As described above, for example, for the search result of the first page that is highly likely not to be an information-oriented search request, the search server 10 can provide the user with the search result obtained by the normal search process. The search server 10 can provide the user with the search results obtained by the pseudo feedback search process for the search results of the second and subsequent pages that are likely to be information-oriented search requests. In addition, since the search results for the second and subsequent pages exclude the search results that overlap with the search results for the first page, even if the search results are applied with different search processing, the duplicate search results may be provided to the user. Absent.

  In the third embodiment, the search server 10 transmits the search result obtained by applying the normal search process to the terminal device 20 of the user who has input the input query. When there is a request for a Web document exceeding a predetermined number (for example, a request for a search result for the second and subsequent pages), a search may be performed for the input query by applying a pseudo feedback search process.

  As mentioned above, although embodiment of this invention was described, this invention is not restricted to embodiment mentioned above. The effects described in the embodiments of the present invention are only the most preferable effects resulting from the present invention, and the effects of the present invention are limited to those described in the embodiments of the present invention. is not.

1 is a block diagram showing a search system 1 according to an embodiment of a search processing device of the present invention. It is a figure which shows an example of the hardware constitutions of the search server 10 which is one Embodiment of the search processing apparatus of this invention. It is a functional block diagram of the search server 10 which is one Embodiment of the search processing apparatus of this invention. It is a figure which shows the click data table 111a. It is a figure which shows the query table. 10 is a flowchart showing a flow of processing for creating a query table 113. It is a flowchart which shows the flow of the main process of the search server 10 of 1st Embodiment. It is a flowchart which shows the flow of a process when performing a pseudo feedback search process. It is a flowchart which shows the flow of a process when performing a normal search process. It is a flowchart which shows the flow of the main process of the search server 10 of 2nd Embodiment. It is a functional block diagram of the search server 10 in 3rd Embodiment. It is a flowchart which shows the flow of the main process of the search server 10 of 3rd Embodiment.

Explanation of symbols

1 Search System 10 Search Server 20 Terminal Device 30 Content Server 40 Communication Line

Claims (4)

A search processing device for executing a search for a Web document by a search engine,
Accepting means for accepting an input query which is an input to the search engine;
Judgment condition storage means for storing a query which is an information-oriented search request for requesting acquisition of a plurality of Web documents related to an input query as search results;
If the input query received by the receiving unit matches the query stored in the determination condition storage unit, the query and the URL of the search result searched by the search engine using the query as a search keyword An entropy calculating means for calculating the entropy of the click URL for the input query based on the click URL that is the URL selected by the terminal device of the user and the probability that the click URL is selected for the input query When,
By determining whether or not the entropy value calculated by the entropy calculating unit exceeds a predetermined threshold, the input query received by the receiving unit uses a plurality of Web documents related to the input query as search results. A determination means for determining whether the request is an information-oriented search request to obtain,
Selection means for selecting a plurality of different search processes based on the result determined by the determination means;
Search execution means for applying a search process selected by the selection means and executing a search for a Web document according to the input query by the search engine;
Search result acquisition means for acquiring the Web document searched by the search execution means as a search result,
The search execution means includes at least a pseudo feedback search process and a normal search process by the search engine,
Said selecting means, by the value of the entropy calculated by said entropy calculation means is determined to exceed a predetermined threshold value by the determination unit, the input query is determined to the the information-oriented search request If the selects the pseudo feedback search process,
The search execution unit applies the pseudo feedback search process selected by the selection unit , executes a search of a preliminary Web document using the input query, and executes a search of the Web document searched. A characteristic term is extracted from a search result, the extracted characteristic term is added to the input query, and a Web document is searched by the search engine using the input query and the added characteristic term as a query . It performs search processing unit. The selection unit selects the normal search process when the input query is a search request that does not require a plurality of Web documents related to the input query to be acquired as a search result as a result of the determination by the determination unit. And
The search executing means, the search processor according to 請 Motomeko 1 to run a search for Web documents in response to the input query by the search engine by applying the normal search process. A search processing method executed by a computer of a search processing device that executes a search of a Web document by a search engine,
Accepting an input query that is an input to the search engine;
A determination condition storing step for storing a query which is an information-oriented search request for requesting acquisition of a plurality of Web documents related to an input query as a search result;
When the input query received by the receiving step matches the query stored in the determination condition storing step, the query and the URL of the search result searched by the search engine using the query as a search keyword An entropy calculating step of calculating an entropy of the click URL for the input query based on a click URL that is a URL selected by the user terminal device and a probability that the click URL is selected for the input query When,
By determining whether or not the entropy value calculated by the entropy calculation step exceeds a predetermined threshold, the input query received by the reception step uses a plurality of Web documents related to the input query as search results. A determination step for determining whether or not the information-oriented search request requires acquisition;
A selection step of selecting a plurality of different search processes based on the result determined by the determination step;
A search execution step of applying a search process selected in the selection step and executing a search of a Web document according to the input query by the search engine;
A search result acquisition step of acquiring the Web document searched by the search execution step as a search result,
The search execution step includes at least a pseudo feedback search process and a normal search process by the search engine,
Said selection step, by the value of the entropy calculated by said entropy calculation step by the determining step is determined to exceed a predetermined threshold value, the input query is determined to the the information-oriented search request If the selects the pseudo feedback search process,
The search execution step applies the pseudo-feedback search process selected in the selection step , executes a preliminary Web document search using the input query by the search engine, A characteristic term is extracted from a search result, the extracted characteristic term is added to the input query, and a Web document is searched by the search engine using the input query and the added characteristic term as a query . A search processing method for causing a computer to execute . A computer program executed by a computer of a search processing apparatus that executes a search for a Web document by a search engine,
Accepting an input query that is an input to the search engine;
A determination condition storing step for storing a query which is an information-oriented search request for requesting acquisition of a plurality of Web documents related to an input query as a search result;
When the input query received by the receiving step matches the query stored in the determination condition storing step, the query and the URL of the search result searched by the search engine using the query as a search keyword An entropy calculating step of calculating an entropy of the click URL for the input query based on a click URL that is a URL selected by the user terminal device and a probability that the click URL is selected for the input query When,
By determining whether or not the entropy value calculated by the entropy calculation step exceeds a predetermined threshold, the input query received by the reception step uses a plurality of Web documents related to the input query as search results. A determination step for determining whether or not the information-oriented search request requires acquisition;
A selection step of selecting a plurality of different search processes based on the result determined by the determination step;
A search execution step of applying a search process selected in the selection step and executing a search of a Web document according to the input query by the search engine;
A search result acquisition step of acquiring the Web document searched by the search execution step as a search result,
The search execution step includes at least a pseudo feedback search process and a normal search process by the search engine,
Said selection step, by the value of the entropy calculated by said entropy calculation step by the determining step is determined to exceed a predetermined threshold value, the input query is determined to the the information-oriented search request If the selects the pseudo feedback search process,
The search execution step applies the pseudo-feedback search process selected in the selection step , executes a preliminary Web document search using the input query by the search engine, A characteristic term is extracted from a search result, the extracted characteristic term is added to the input query, and a Web document is searched by the search engine using the input query and the added characteristic term as a query . A computer program that causes a computer to execute.

Images