JP7614705B2 - Information processing system, information processing method, and program - Google Patents

Description

The present invention relates to an information processing system, an information processing method, and a program.

As the number of digitized documents within a company increases, the importance of document search systems for efficiently searching for documents required for business is growing. A document search system presents a user with a collection of documents related to the search criteria entered by the user. In such a system, the user cannot always obtain the desired document with a single search, and there are many cases where a large number of documents related to the entered search criteria are hit. In such cases, the user will repeatedly try to narrow down the search results by adding additional keywords to the search criteria, but they may not be able to think of the appropriate keywords to narrow down the search results and will spend a lot of time before finding the document they need.

While it is difficult to create appropriate search criteria, it is relatively easy to determine whether the documents presented as search results are what the user wants (hereafter, relevant documents) or not (hereafter, non-relevant documents). Taking advantage of this property, a method called "relevance feedback" is being researched that aims to improve search performance based on users' evaluations of search results (Non-Patent Document 1).

In addition to the Rocchio algorithm, a representative method of relevance feedback, Non-Patent Document 1 mentions an algorithm that uses relational learning.

Patent Document 1 focuses on the problem of the heavy burden placed on users when determining whether multiple documents are relevant or not when providing relevance feedback, and describes a method of calculating relevance that is devised so that appropriate feedback can be obtained even when a single relevant document is specified. This method extracts words related to the search criteria and expands the search criteria by treating documents that appear at the top of the search results as pseudo-relevant documents.

Seiji Yamada, Takashi Onoda (2003). Document Retrieval by Relevance Feedback Journal of the Department of Chemistry and Information Science, Chemical Society of Japan, Vol. 21 (2003) No. 2, 32

JP 2004-178421 A

As described in Non-Patent Document 1, many relevance feedback techniques are based on the assumption that the user will evaluate multiple relevant and non-relevant documents. Although it depends on the purpose of the search, it can be very time-consuming for the user to evaluate the relevance of multiple documents every time they perform a search. To solve this problem, one possible method is to treat the top documents in the search results as pseudo-relevant documents, as described in Patent Document 1. However, if the majority of the top documents in the search results are non-relevant documents, this could actually reduce search accuracy. Another problem is that selecting relevant documents is relatively difficult compared to evaluating non-relevant documents.

The present invention aims to provide a search system that makes it easier for users to find the documents they want.

The information processing system of the present invention is characterized by comprising: a receiving means for receiving feedback from a user regarding documents searched based on a search request received from the user; an adjustment means for adjusting the search score of the searched document based on the content of the feedback received by the receiving means; and a display means for displaying the search results based on the search score adjusted by the adjustment means.

The present invention makes it possible to provide a search system that makes it easier for users to find the documents they want.

1 is a diagram illustrating an example of a system configuration of a document search system according to an embodiment of the present invention. 1 is a block diagram showing an example of a hardware configuration of a document search system according to an embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a document DB in the embodiment of the present invention. 13 is an example of a list of tf-idf values by document and by keyword in the embodiment of the present invention. FIG. 11 is a diagram showing an example of a search result in the embodiment of the present invention. FIG. 11 is a diagram showing an example of a search result with characteristic words in the embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a user interface in the embodiment of the present invention. 11 is a flowchart illustrating an example of a search score adjustment process in the embodiment of the present invention. FIG. 11 is a diagram illustrating an example of a score adjustment rule for relevant documents in the embodiment of the present invention. FIG. 11 is a diagram illustrating an example of a score adjustment rule for non-relevant documents in the embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a user interface in the embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a user interface in the embodiment of the present invention. FIG. 4 is a diagram illustrating an example of a user interface in the embodiment of the present invention. 1 is a flowchart showing an overview of a search process in an embodiment of the present invention.

The following describes an embodiment of the present invention in detail with reference to the drawings.

Figure 1 is a diagram showing an example of the system configuration of a document search system (information processing system) in an embodiment of the present invention.

The document search system 100 consists of a document registration device 110, a document DB 120, a document search device 130, and a client terminal.

The document registration device 110 is a device for registering documents to be searched by users, and is composed of a document receiving unit 111, a keyword extraction unit 112, and a document registration processing unit 113.

The document receiving unit 111 is a device for accepting documents to be registered. A user can send any document to the document receiving unit 111 via a web browser on a client terminal. Alternatively, a crawler may be configured to mechanically collect and send documents.

The keyword extraction unit 112 is a device for extracting keywords that are candidates for characteristic words in a document received by the document receiving unit 111, and the frequency of their occurrence. Details of characteristic words will be described later. The keyword extraction process in the keyword extraction unit 112 uses a known morphological analysis technique. Here, the morphemes to be extracted may be limited to specific parts of speech, such as proper nouns, depending on the purpose of the search system. Also, character strings that match a predetermined pattern may be extracted as keywords without using morphological analysis.

The document registration processing unit 113 is a device that links the document received by the document receiving unit 111 with the keywords extracted by the keyword extraction unit 112 and stores them in the document DB 120.

Figure 3 is an example of document DB 120. Document DB 120 has fields for document ID 121 for uniquely identifying a document, and keyword: occurrence frequency 122 for storing values extracted by keyword extraction unit 112. For example, for record 123 in document DB 120, the document ID is "Document 1", there are four keywords, "A", "B", "C", and "G", and the occurrence frequency of each keyword in document 1 is "3", "1", "2", and "1", respectively. Note that the above two fields are shown as a minimum configuration for explaining this idea, but additional fields for use as a search system, such as the document title and a URL indicating the location of the document, may be included.

Returning to FIG. 1, the document search device 130 consists of a search processing unit 131, a feature word extraction unit 132, and a search result output processing unit 133.

The search processing unit 131 is a device that receives search requests from a user and searches the document DB for documents that correspond to the requests, and has a function of retrieving documents related to the search request in order of score. To achieve efficient search processing, the document registration processing unit 113 can create a transposed index, which is a known technique, and use it during searching. The search processing unit 131 also performs the search score adjustment process of the present invention using matching documents and non-matching documents selected by the user. The search score adjustment process will be explained in detail later using an example.

The characteristic word extraction unit 132 is a device that selects characteristic keywords (characteristic words) for each document from the document DB 120 for each document in the search results obtained by the search processing unit 131. The selection of characteristic words can be achieved by using tf-idf, which is one of the indices that represent the characteristic amount of a word.

Figure 4 shows the tf-idf values for each keyword in each document in document DB 120. For example, when selecting two feature words in document 1, it is possible to extract "C" and "B" as feature words in descending order of tf-idf value in document 1. The method of obtaining feature words using tf-idf values is merely one example, and other methods may be used to obtain feature words.

The search result output processing unit 133 is a device that returns the search results obtained by the search processing unit 131 and the characteristic words obtained by the characteristic word extraction unit 132 to the client that sent the search request. A user of the document search system 100 can check the search results and the characteristic words for each document through a web browser or the like.

Figure 2 is a block diagram showing an example of the hardware configuration of an information processing device that can be used as a client terminal or document search system 100 in an embodiment of the present invention.

As shown in FIG. 2, the information processing device is connected to a CPU (Central Processing Unit) 201, a ROM (Read Only Memory) 202, a RAM (Random Access Memory) 203, a storage device 204, an input controller 205, an audio controller 206, a video controller 207, a memory controller 208, and a communication I/F controller 209 via a system bus 200.

The CPU 201 provides overall control over each device and controller connected to the system bus 200.

The ROM 202 or external memory 213 holds the BIOS (Basic Input/Output System) and OS (Operating System), which are control programs executed by the CPU 201, computer-readable and executable programs for implementing this information processing method, and various necessary data (including data tables).

RAM 203 functions as the main memory, work area, etc. of CPU 201. CPU 201 loads programs and the like required for executing processing from ROM 202 or external memory 213 into RAM 203, and executes the loaded programs to realize various operations.

The input controller 205 controls input from input devices such as a keyboard 210 and a pointing device such as a mouse (not shown). If the input device is a touch panel, the user can give various instructions by pressing (touching with a finger, etc.) icons, cursors, or buttons displayed on the touch panel.

The touch panel may also be a touch panel capable of detecting positions touched by multiple fingers, such as a multi-touch screen.

The video controller 207 controls the display on an external output device such as a display 212. The display is also intended to include the display of a notebook computer integrated into the main body. Note that the external output device is not limited to a display, and may be, for example, a projector. In addition, for devices capable of accepting the aforementioned touch operations, an input device is also provided.

The video controller 207 can control a video memory (VRAM) for display control, and can use part of the RAM 203 as a video memory area, or can provide a separate dedicated video memory.

The memory controller 208 controls access to the external memory 213. The external memory can be an external storage device (hard disk) that stores the boot program, various applications, font data, user files, edit files, and various data, a flexible disk (FD), or a compact flash (registered trademark) memory connected to a PCMCIA card slot via an adapter.

The communication I/F controller 209 connects and communicates with external devices via a network, and executes communication control processing on the network. For example, communication using TCP/IP, telephone lines such as ISDN, and 4G and 5G lines for mobile phones are possible.

The CPU 201 can display the outline font on the display 212 by, for example, executing a process of expanding (rasterizing) the outline font in a display information area in the RAM 203. The CPU 201 can also allow the user to give instructions using a mouse cursor (not shown) on the display 212.

First, we will explain the overall process of the present invention using the flowchart in Figure 14.

In step S1401, a search request is received from a user via a client terminal.

In step S1402, based on the search request received in step S1401, the search processing unit 131 searches the document DB for a document that meets the request.

In step S1403, the results of the search process in step S1402 are displayed on the client terminal. An example of a screen displaying the search results is shown in FIG. 7.

In step S1404, it is determined whether feedback on the search results has been received from the user, and if feedback has been received (step S1404: YES), the process proceeds to step S1405.

In step S1405, the search score is adjusted according to the feedback received from the user. Details of the search score adjustment process will be described later using the flowchart in FIG. 8.

In step S1406, a search result screen based on the search score adjusted in step S1405 is displayed on the client terminal. Examples of the adjusted screen are shown in Figures 11, 12, and 13.

As shown in FIG. 11 etc., it is possible to receive further feedback on the screen after adjustment, and if feedback is received, the processing of steps S1405 and S1406 is executed again.

Next, the search score adjustment process in an embodiment of the present invention will be explained using Figures 5 to 13.

Figure 5 shows an example of a search result acquired by the search processing unit 131 when a request to search for documents containing keyword "A" as a search condition is received from the user in step S1401. The search result 500 is an array of elements including a document ID 501 and a score 502 indicating the degree of relevance to the search condition, and is sorted in descending order by the value of the score 502. For example, element 503 of the search result 500 has a document ID of "Document 1" and a score of "0.029". Here, in order to simplify the explanation, an example is shown in which the sum of the tf-idf values of the keywords included in the search condition is used as the score, and the top three results are displayed in descending order of score, but a known algorithm such as BM25 may be used. The number of search results acquired may also be increased or decreased according to the user interface that displays the search results.

Figure 6 shows an example of the result of adding feature words extracted by the feature word extraction unit 132 to each document in the search results 500. To simplify the explanation, the method of extracting feature words is exemplified by extracting two in descending order of tf-idf value, but the feature words may be categories of documents automatically classified using known document classification technology, or may be extracted according to predetermined rules. The number of feature words to be acquired may be increased or decreased according to the user interface. Furthermore, the feature words for each document may be calculated in advance and stored in association with the document DB 120, or the feature words acquired during the search process may be cached for a certain period of time, or other measures may be taken to speed up the process of acquiring feature words.

Figure 7 is an example of a user interface that the search result output processing unit 133 displays on the browser of the client terminal in step S1403. The user interface 700 is characterized by having a search condition input form 701, a search result summary 702, an area for displaying characteristic words 601 for each document in the search result, and a score up button 703 and a score down button 704 as interfaces for the user to provide feedback to the system on the suitability or non-suitability of documents. Figure 7 shows the state after a search for documents containing keyword "A" as a search condition has been performed, and the search result summary 702 indicates that there are a total of six documents containing keyword A, of which the top three with the highest scores are displayed on the screen.

The characteristic words 601 are values used in the search score adjustment process described below, and displaying the characteristic words has the effect of making it easier for the user to determine whether a document is a relevant document or a non-relevant document. Note that information about each document may also be displayed, such as the document title and text surrounding the keywords in the search criteria, to make the search results easier to understand.

By pressing the score up button 703, the user can provide feedback to the system that the document is a matching document. By pressing the score down button 704, the user can provide feedback to the system that the document is a non-matching document.

Next, the details of the processing in step S1405 will be explained using FIG. 8.

Figure 8 is a flow chart showing the process in which the CPU 201 of the document search system 100 reads and executes a specific control program, and which illustrates the process of adjusting the search score in the search processing unit 131. In steps S801 to S805, the process of adjusting the search score of each document that matches the search criteria is performed.

First, in steps S802 and S803, for each document that has been given feedback from the user as a relevant document, the search score of each document is adjusted according to the score adjustment rules for relevant documents.

Figure 9 shows an example of a search score adjustment rule for matching documents. Multiple rules can be defined as score adjustment rules 900 for matching documents, and each rule has application order 901, score adjustment conditions 902, and score adjustment value 903 as values. All rules are evaluated in the order of application, and if the document to be evaluated satisfies score adjustment conditions 902, the search score is updated using the value specified in the score adjustment value. Specific examples will be described later.

Returning to FIG. 8, in steps S804 and S805, for each document that has been given feedback from the user as a non-relevant document, the search score of each document is adjusted according to the score adjustment rules for non-relevant documents.

Figure 10 shows an example of a search score adjustment rule for non-relevant documents. Similar to the score adjustment rule 900 for relevant documents, the score adjustment rule 1000 for non-relevant documents allows multiple rules to be defined, and each rule has an application order 1001, a score adjustment condition 1002, and a score adjustment value 1003 as values. All rules are evaluated in the order of application, and if the document to be evaluated satisfies the score adjustment condition 1002, the search score is updated using the value specified in the score adjustment value.

The above-mentioned example will be used for further detailed explanation below. In the state of the user interface 700 in FIG. 7, when the user presses the score DOWN button for document 1 (designating document 1 as a non-relevant document), the search processing unit 131 uses document 1 as a non-relevant document for adjusting the search score. The characteristic words of document 1 are "C" and "B". According to the score adjustment rule 1000 for non-relevant documents, the search scores of documents (documents 1, 4, and 6) that contain 30% or more characteristic words, i.e., at least one of the keywords "C" and "B", are multiplied by 0.5, and the search scores of documents (documents 1 and 6) that contain 70% or more characteristic words, i.e., both keywords "C" and "B", are further multiplied by 0.5.

Figure 11 shows the user interface 700 displayed on the client terminal in step S1406 after the search score has been adjusted by non-matching document 1. If there is a document designated by the user as a matching document or a non-matching document, the search result output processing unit 133 displays a score adjustment label 750 in the user interface 700, indicating that the score has been adjusted by the document designated by the user. The score adjustment label 750 consists of a label 751 indicating the matching document or non-matching document, a document label 752, and a release button 753. The user can release the document from the feedback target by pressing the release button 753.

By using the user interface 700 of the present invention, the user can further provide feedback on the relevant and non-relevant documents while viewing the search results after the score adjustment. In the state of the user interface 700 in FIG. 11, when the user presses the score up button for document 5 (designating document 5 as a relevant document), the search processing unit 131 adjusts the search score by treating document 1 as a non-relevant document and document 5 as a relevant document. The characteristic words of document 5 are "F" and "E". According to the score adjustment rule 900 for relevant documents, the search scores of documents (documents 2, 3, 5, and 7) that contain 30% or more of the characteristic words, i.e., at least one of the keywords "F" and "E", are doubled, and the search scores of documents (documents 5 and 7) that contain 70% or more of the characteristic words, i.e., both of the keywords "F" and "E" are doubled.

Figure 12 shows the user interface 700 after search scores have been adjusted from the state shown in Figure 11 using non-relevant document 1 and relevant document 5. In addition to score adjustment label 750, which indicates that document 1 is being used to adjust its score as a non-relevant document, score adjustment label 760, which indicates that document 5 is being used to adjust its score as a relevant document, is displayed. In this way, all score adjustment labels that correspond to the documents specified by the user are displayed. In Figure 12, the documents displayed as search results are "Document 5," "Document 7," and "Document 2," which are different from the document set before feedback was provided.

As described above, the present invention is characterized by the fact that it does not expand the search criteria themselves through relevance feedback from the user, but rather improves the scores of documents similar to relevant documents selected by the user and decreases the scores of documents similar to non-relevant documents selected by the user, thereby adjusting the order in which the search results are displayed without expanding or reducing the set of documents output as search results.

In this way, by limiting the effect of relevance feedback to score adjustments only, documents unrelated to the search criteria will not be added to the search results even if undesirable features are extracted from the documents determined by the user. The user can search for the document they are looking for by repeating the simple operation of selecting one relevant or non-relevant document from the documents included in the presented search results. We have shown that by using this invention, the user can bring the documents appearing at the top of the search results closer to the documents they desire, through the simple operation of interactively selecting one relevant and one non-relevant document at a time.

The score adjustment using the score adjustment rule 900 for relevant documents and the score adjustment rule 1000 for non-relevant documents described above corresponds to adjusting the scores of documents similar to the fed back document. To simplify the explanation, the score adjustment condition is a simple one using the content rate of feature words, but a method of evaluating the similarity of documents may be used that vectorizes documents and calculates the distance between vectors, or a method of estimating the category of documents and considering documents belonging to the same category as similar documents. In addition, the score adjustment value may not only be defined by multiplying a constant value or adding a constant value, but may also be dynamically adjusted using the score of the search results (such as performing score adjustment to the extent that documents designated as non-relevant documents are excluded from the top of the search results (for example, by setting the score to zero)). Since multiple score adjustment rules can be defined, these multiple methods may be combined.

The score adjustment labels are intended to clarify documents for which the user has provided feedback during the search process. To make it easier for the user to understand what kind of score adjustment has been performed, a function may be provided that displays a summary of the document, including feature words, by hovering or clicking on the score adjustment label, as shown in Figure 13. If the document category has been estimated, this information may also be displayed.

The present invention can be embodied, for example, as a system, device, method, program, or recording medium. Specifically, it may be applied to a system made up of multiple devices, or to a device made up of a single device.

The program of the present invention is a program that enables a computer to execute the processing methods of the flowcharts shown in Figures 8 and 14, and the storage medium of the present invention stores a program that enables a computer to execute the processing methods of Figures 8 and 14. The program of the present invention may be a program for each processing method of each device in Figures 8 and 14.

As described above, it goes without saying that the object of the present invention can be achieved by supplying a recording medium on which a program that realizes the functions of the above-mentioned embodiments is recorded to a system or device, and having the computer (or CPU or MPU) of that system or device read and execute the program stored on the recording medium.

In this case, the program read from the recording medium itself realizes the novel functions of the present invention, and the recording medium on which the program is recorded constitutes the present invention.

Recording media for supplying the program may include, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a DVD-ROM, a magnetic tape, a non-volatile memory card, a ROM, an EEPROM, a silicon disk, etc.

Furthermore, it goes without saying that not only are the functions of the above-mentioned embodiments realized by the computer executing a program it has read, but also that the functions of the above-mentioned embodiments can be realized by an operating system (OS) or the like running on the computer carrying out some or all of the actual processing based on the instructions of the program.

Furthermore, it goes without saying that this also includes cases where a program read from a recording medium is written into a memory provided on a function expansion board inserted into a computer or a function expansion unit connected to a computer, and then a CPU or the like provided on the function expansion board or function expansion unit performs some or all of the actual processing based on the instructions of the program code, thereby realizing the functions of the above-mentioned embodiments.

The present invention may be applied to a system made up of multiple devices, or to a device made up of a single device. Needless to say, the present invention can also be applied to cases where the effects of the present invention are achieved by supplying a program to a system or device. In this case, the effects of the present invention can be enjoyed by reading a recording medium that stores a program for achieving the present invention into the system or device.

Furthermore, by downloading and reading a program for achieving the present invention from a server, database, etc. on a network using a communication program, the system or device can enjoy the effects of the present invention. Note that the present invention also includes configurations that combine the above-mentioned embodiments and their variations.

100 Document search system (information processing system)
110 Document registration device 120 Document DB
130 Document Search Device

Claims (6)

a display control means for controlling display of search target data searched for using search criteria related to a search request received from a user in order of a search score, which is a degree of relevance to the search criteria;
a receiving means for receiving from the user a selection as to suitability of the search result for the search target data searched based on the search criteria related to the search request received from the user;
a control means for adjusting the search score related to the search target data in accordance with the selection received by the receiving means;
Equipped with
The control means adjusts the search score of non-selected search target data, which is search target data for which the selection of suitability has not been accepted, in accordance with the content of the selection accepted by the acceptance means, for the non-selected search target data in which a plurality of keywords characterizing the search target data for which the selection of suitability has been accepted matches a plurality of keywords characterizing the non-selected search target data at a predetermined rate or more ;
The information processing system further comprises: a display control means for controlling display of the searched search target data in order of the search score adjusted by the control means. The information processing system according to claim 1, characterized in that the accepting means accepts a selection of suitability of the search results for each piece of search target data. the display control means controls to display a button for accepting a selection of suitability of the search result for each of the search target data;
3. The information processing system according to claim 2 , wherein the accepting unit accepts a selection of suitability as a search result by accepting an operation on the button from a user. 4. The information processing system according to claim 1, wherein the search target data is document data. a display control step of controlling a display control means of the information processing system to display search target data searched for using search criteria related to a search request received from a user in order of a search score, which is a degree of relevance to the search criteria;
a receiving step in which a receiving means of the information processing system receives from a user a selection of suitability of search results for search target data searched based on search conditions related to the search request received from the user;
A control means of the information processing system adjusts the search score related to the search target data based on the content of the selection accepted in the accepting step;
a control step of adjusting the search score of non-selected search target data, which is search target data for which the selection of suitability has not been accepted, in accordance with the content of the selection accepted in the accepting step, for the non-selected search target data in which a plurality of keywords characterizing the search target data for which the selection of suitability has been accepted matches a plurality of keywords characterizing the non-selected search target data at a predetermined rate or more ;
a second display control step of controlling a display control means of the information processing system to display the searched search target data in the order of the search scores adjusted by the control step;
An information processing method comprising: A program for causing a computer to function as each of the means according to any one of claims 1 to 4 .

Images