JP2006285656A - Document search system, recording medium, program and document search method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document search system that facilitates the location of desired content and increases a user's search efficiency. <P>SOLUTION: The document search system for searching search object documents on an input query and outputting search results comprises a displayed result display registration means for recording a history of search results having been displayed on screen out of the search results, and means for, when a search on a new query produces search results, excluding part of the new search results equal to the history part of search results recorded in the displayed result display registration means or moving the equal part to the end of the new search results. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a document search system, a recording medium, a program, and a document search method, and in particular, a document search system and a record that make it easy to find a target search target document and can improve user search efficiency. The present invention relates to a medium, a program, and a document search method.

  Google (registered trademark) or Yahoo! Search sites such as (registered trademark) provide a keyword search system that can search a large number of Web pages using keywords. On the other hand, consumer manufacturers provide a natural text search system that can search a collection of frequently asked questions and answers (also referred to as contents) in natural text. In this specification, Web pages and contents are collectively referred to as search target documents.

  Users of these search systems usually repeat the search several times because the target search target document is not always found by a single search. When the search result includes a result display of a search target document other than the purpose of repetition (in this specification, the subject of the search target document, a summary of the search target document, and a preview are collectively referred to as a result display). There is a problem that it becomes difficult for a person to find a target search target document.

  As a method for solving the above problem, there is a conventional “information search system and method” (for example, see Patent Document 1). In the information search system and method of Patent Literature 1, a technique is described in which a search result history is taken and an icon is displayed beside a result display searched in the past. By using such a technique, it is easy to find the target search target document because the display of the result of the search in the past can be immediately known.

Further, there is a “document search apparatus, program, and recording medium” of the prior art (see, for example, Patent Document 2). In the document search device, program, and recording medium of Patent Document 2, there is a technique that takes a browsing history of a search target document and displays a search result in a separate area for a browsed search target document and an unread search target document. Are listed. If such a technique is used, the target search target document can be found efficiently by paying attention to the area of the unread search target document.
JP-A-8-44759 JP 2004-234284 A

  By the way, when the document search system displays a large amount of result display, it is divided into a plurality of pages, or is output in one page in anticipation of the user scrolling the screen or changing the size. .

  In the prior art of Patent Document 1, an icon is also attached to a result display that is included in past search results but is not displayed on the screen. Therefore, there is a problem that it is difficult for the user to find the target search target document when the target search target document is overlooked even though it is not displayed on the screen. The same can be said for the prior art of Patent Document 2.

  Further, it is considered that the user looks at the result display list, determines that there is no target search target document, and searches again. Even after a re-search, even if a list of result displays similar to that before the re-search is displayed, it is considered that it is difficult for the user to find the target search target document. In the prior arts of Patent Literature 1 and Patent Literature 2, the similarity of a search target document having a result display similar to the result display displayed on the screen cannot be reduced. Similarly to the result display, the similarity of the search target document similar to the search target document browsed before the re-search cannot be reduced.

  In addition, if the user's search purpose changes during a session (a series of operations performed on the search system for a user's purpose), the target search target document after changing to the search result history before the change If entered, there is a problem that it becomes difficult for the user to find the target search target document.

  The present invention has been made to solve such a problem. The object of the present invention is to make it easy to find the target search target document, to improve the search efficiency of the user, and to achieve the target search target document. It is an object to provide a document search system, a recording medium, a program, and a document search method that can make it difficult to overlook.

The present invention has been made to solve the above-described problem, and the document search system of the present invention is a document search system that searches a search target document based on an inputted question sentence, and outputs a search result. Among the search results, the displayed result display registration means for recording the history of the search results displayed on the screen before the previous time and the new search result when the search result is obtained by searching with a new question sentence. And a means for excluding a portion common to the history portion of the search result recorded in the displayed result display registration means, or a means for moving the common portion to the end of a new search result. To do.
In the document search system configured as described above, the information of the portion displayed on the screen among the search results of the question text input before the previous time is stored as history information. Either remove the part that is common to the history part that was previously displayed on the screen, or move the common part to the end of the new search results. As a result, only the result display seen by the user can be excluded from the search result or moved to the end of the search result, and it is possible to easily find a search target document other than already displayed.

The document search system of the present invention also includes search interface means for controlling the size of the screen for displaying the search results and the scroll range of the screen, and the screen display information of the search results received from the search interface means. And a displayed result display registration means for recording a history of search results displayed on the screen.
In the document search system configured as described above, information on the size of the screen on which the search result is displayed and the scroll range of the screen is collected, and the portion of the search result displayed on the screen is determined based on this information. Then, the portion of the search result displayed on the screen is recorded as history information. As a result, only the result display seen by the user can be excluded from the search result or moved to the end of the search result, and it is possible to easily find a search target document other than already displayed.

Further, the document search system of the present invention includes a means for outputting the search results together with the similarity to the question sentence in order from the highest similarity, and a search target document displayed on the screen before the previous time among the search results. When a search result is obtained by performing a search using a new question sentence, a displayed result display registration unit that stores information on the result display history, history information on the search target document browsed before and Among the new search results, the similarity of the search target document having a result display similar to the result display of the search target document recorded in the displayed result display registration unit and the recorded result display registration unit Means for reducing the similarity of a search target document similar to the search target document browsed before the previous time.
In the document search system configured as described above, when a search is performed using a new question sentence, a search target document having a result display similar to the result display displayed before the previous time, or a search similar to the browsed search target document Lower the similarity of the target document. As a result, it is possible to lower the display order of the search target document displayed before or last time, the search target document browsed, and the search target document similar in result display or similar to the browsed search target document. Search efficiency is further improved.

Further, the document search system of the present invention compares the search result of the previous question sentence with the search result of the current question sentence, and when the ratio of the common search results is small, the search purpose is considered to be changed and the displayed A means for deleting the history information of the result display recorded in the result display registration means is provided.
In the document search system configured as described above, whether or not the search purpose has changed is determined from the search result history. That is, the previous search result and the current search result are compared, and when the ratio of the common search results is small, it is considered that the search purpose has changed, and the result display history is deleted. As a result, when the search purpose changes, the process of excluding or moving the displayed result display from the search result is invalidated, so that it becomes difficult to overlook the target search target document.

In addition, the document search system of the present invention searches the search target document by both the means for storing the previous question text, the previous question text, and the current question text, and compares both search results. Means.
In the document search system configured in this way, in order to determine whether the search purpose has changed, both the previous question sentence and the current question sentence are searched simultaneously, and the ratio of the common search results is When the number is small, it is considered that the search purpose has changed, and the history of the result display is deleted. As a result, it is possible to determine that the search purpose has changed without saving the search result of the previous question sentence.

Further, the document search system of the present invention is a document search system that searches a search target document by an inputted question sentence, and outputs a search result together with a similarity to the question sentence in order from a high similarity. Among the search results, the display result display registration means for storing the history information of the search target document displayed on the screen before the previous time, the history information of the search target document browsed before the previous time, a new A search target document having a result display similar to the result display of the search target document recorded in the displayed result display registration unit among the new search results when a search result is obtained by performing a search using a question sentence. And a means for lowering the similarity of the search target document similar to the search target document viewed before the previous time recorded in the displayed result display registration means. .
In the document search system configured as described above, when a search is performed using a new question sentence, a search target document having a result display similar to the result display displayed before the previous time, or a search similar to the browsed search target document Lower the similarity of the target document. As a result, it is possible to lower the display order of the search target document displayed before or last time, the search target document browsed, and the search target document similar in result display or similar to the browsed search target document. Search efficiency is improved.

Further, the document search system of the present invention reduces the importance of words included in the result display history recorded in the displayed result display registration unit, and also browses the searched search recorded in the displayed result display registration unit. It is characterized by comprising means for reducing the similarity by lowering the importance of words contained in the target document.
In the document search system configured as described above, the result display history displayed on the screen before the previous time is used to reduce the importance of words included in the result display. Also, the importance of words included in the search target document browsed before the previous time is lowered. As a result, the similarity of a search target document having a result display similar to the displayed result display or a search target document similar to the browsed search target document can be lowered and the display order can be lowered. Search efficiency is improved.

Further, the document search system of the present invention compares the search result of the previous question sentence with the search result of the current question sentence, and when the ratio of the common search results is small, the search purpose is considered to be changed and the displayed A means for deleting the history information of the result display recorded in the result display registration means is provided.
In the document search system configured as described above, it is determined from the search result history whether the search purpose has changed. That is, the previous search result and the current search result are compared, and when the ratio of the common search results is small, it is considered that the search purpose has changed, and the result display history is deleted. As a result, when the search purpose changes, the process of excluding or moving the displayed result display from the search result is invalidated, so that it becomes difficult to overlook the target search target document.

In addition, the document search system of the present invention searches the search target document by both the means for storing the previous question text, the previous question text, and the current question text, and compares both search results. Means.
In the document search system configured in this way, in order to determine whether the search purpose has changed, both the previous question sentence and the current question sentence are searched simultaneously, and the ratio of the common search results is When the number is small, it is considered that the search purpose has changed, and the history of the result display is deleted. As a result, it is possible to determine that the search purpose has changed without saving the search result of the previous question sentence.

Further, the document search system of the present invention is connected to a terminal device having a browser for inputting the question text and displaying the search result through a communication network, receives information on the question text from the terminal device, and retrieves the search result. A Web server that transmits the information to the terminal device.
In the document search system configured as described above, a browser is mounted on a user's PC (personal computer), and a search is performed by accessing a Web server of the document search system. Thereby, the document search system of this invention can be used for a search site.

  The computer-readable recording medium of the present invention searches a document to be searched based on an inputted question sentence and outputs a search result to a computer in the document search system, and the search result is displayed on a screen before the previous time. A step of recording the history of the displayed search results in the displayed result display registration means; and when a search result is obtained by performing a search using a new question sentence, the displayed result display registration means is obtained from the new search result. A computer-readable record recording a program for excluding a part common to the history part of the search result recorded in the step or moving the common part to the end of the new search result It is a medium.

  Further, the computer program of the present invention searches the search target document by the inputted question sentence, and the search result displayed on the screen before the previous time among the search results on the computer in the document search system that outputs the search result. In the displayed result display registration means, and when a search result is obtained by searching with a new question sentence, the new search result is recorded in the displayed result display registration means. This is a program for executing a step of excluding a portion common to a history portion of a search result or moving the common portion to the end of a new search result.

  The document search method of the present invention is a document search method in a document search system that searches a search target document based on an inputted question sentence and outputs a search result, wherein the computer in the document search system Of the results, the step of recording the history of the search results displayed on the screen before the previous time in the displayed result display registration means, and when the computer performs a search with a new question and obtains the search results, Excluding a part common to the history part of the search result recorded in the displayed result display registration means from the new search result, or moving the common part to the end of the new search result; It is characterized by including.

  Next, the best mode for carrying out the present invention will be described with reference to the drawings.

[First Embodiment]
FIG. 1 is a diagram showing a network configuration example of a document search system according to the present invention. The FAQ search system (document search system) 1 includes a terminal device 10 and a FAQ search system server 20 connected to the terminal device 10 via the Internet 2. The user uses the FAQ search system 30 via the Web server 21 using the browser 11 installed in the terminal device 10. When the question text is passed to the FAQ search system 30 via the browser, the search result is displayed on the browser 11.

  FIG. 2 is a diagram illustrating a configuration example of the FAQ search system according to the first embodiment of this invention. As shown in FIG. 2, the FAQ search system (document search system) 1 includes a terminal device 10 and an FAQ search system server 20, and a user searches an FAQ search site in a browser 11 of the terminal device 10. When visiting, the software for the search I / F unit 12 (for example, Java (registered trademark) applet or dedicated software) is loaded. The search I / F unit 12 communicates with the FAQ search system 30 via the Web server 21.

  FIG. 3 shows an example of a browser loaded by the search I / F unit 12 and a content display browser. On the browser, a question sentence input unit 14, a search result display unit 15, a search condition setting unit 16, a search button 17. The search result display unit 15 displays the search results on a plurality of pages (see FIG. 5B).

  Also, when the software of the search I / F unit 12 is loaded into the browser, an ID for managing a session (hereinafter referred to as a session ID) is received from the Web server 21. When the user scrolls the screen, switches pages, etc. and sees the search result, the result display ID (hereinafter, result display ID) displayed on the screen is stored in the memory (hereinafter, result display ID array).

  Further, after the user presses the search button 17 on the browser 11, the question sentence, the result display ID array, the search condition, and the session ID are passed to the FAQ search system 30. The search result is received from the FAQ search system 30 and displayed on the search result display unit 15. When the user clicks on the result display displayed on the search result display unit 15, the content display browser 13 displays the content.

  The FAQ search system server 20 includes a Web server 21 and an FAQ search system 30. The session ID creation unit 22 in the Web server 21 creates a session ID and passes it to the search I / F unit 12 of the terminal device 10 when the user visits the FAQ search site.

The control unit 31 in the FAQ search system 30 controls each part of the FAQ search system 30.
As the search processing, the control unit 31 passes the result display ID array received from the search I / F unit 12 of the terminal device 10 to the displayed result display registration unit 35. Further, the question sentence received from the search I / F unit 12 is transferred to the question sentence vector creating unit 33, and the question sentence vector created by the question sentence vector creating unit 33 is received.
Further, the question sentence vector and the session ID are passed to the similar content search unit 34, and the search result is received. The search result is passed to the search I / F unit 12 of the terminal device 10.

  The control unit 31 receives the result display ID and the session ID from the content display browser 13 of the terminal device 10 as the content acquisition process, and passes them to the content acquisition unit 36. Further, the content is received from the content acquisition unit 36 and transferred to the browser 11.

  The morpheme analysis unit 32 refers to the morpheme analysis dictionary 37 for the sentence received as input, decomposes it into words, stores it in an array (hereinafter, word array), and outputs the word array. The question sentence vector creation unit 33 passes the question sentence to the morphological analysis unit 32 and receives the word arrangement. Then, a question sentence vector is created using the weighting dictionary 38 (see FIG. 4A), and this question sentence vector is passed to the control unit 31.

  FIG. 4A is a diagram illustrating an example of the weighting dictionary 38, which is a table in which weighting numerical values are recorded for each predetermined word.

  The similar content search unit 34 receives the session ID and the question sentence vector from the control unit 31 and searches the content DB 39 for content similar to the question sentence. The search result is passed to the control unit 31.

The displayed result display registration unit 35 receives the session ID and the result display ID array from the control unit 31, and registers the displayed result display history in the displayed result display table.
The displayed result display DB 40 has a displayed result display table. The displayed result display table stores a history of result display IDs displayed on the screen during the session. FIG. 5A shows an example of a displayed result display table. In the displayed result display table, the session ID and the result display ID are recorded in association with each other.

  The content acquisition unit 36 receives the result display ID, acquires the content from the content DB 39, and passes it to the control unit 31. The content DB (see FIG. 4B) 39 is a DB storing content collections. FIG. 4B is a diagram illustrating an example of the content DB 39, which is a DB in which a display result, a document vector, and content details are registered using a result display ID as an index.

FIG. 6 is a flowchart showing the flow of processing of the search I / F unit. In FIG. 6, “Start” indicates a point in time when the user visits the FAQ search site. In “communication initialization (step S <b> 11)”, the software of the search I / F unit 12 is loaded into the browser, and the session ID is received from the Web server 21.
In “input: question sentence (step S12)”, the question sentence input by the user and the search condition setting (for example, content that has already been displayed is not displayed) are passed to the FAQ search system 30. When re-searching, the result display ID array is also passed.
In “FAQ search (step S13)”, the FAQ search system 30 searches for content similar to a question sentence. Then, the search result is passed to the search I / F unit 12 (see FIG. 9).
“Output: search result (step S14)” is a process in which the search result display unit 15 (see FIG. 3) of the browser 11 receives the search result and displays it on the screen.

The “displayed result display ID storage (step S15)” stores the result display ID drawn on the screen in the result display ID array when the user changes the browser window size, scrolls, or switches the search result page. This is a process of storing (see FIG. 7).
“Content display (step S16)” is a process of starting the browser and displaying the content when the user clicks on the result display.
The “re-search (step S17)” condition is no when the user closes the browser or opens a page other than the FAQ search site. Otherwise, yes.

FIG. 7 is a flowchart showing the flow of processing in which the search result display unit 15 of the browser stores the displayed result display ID.
First, a search result is input to the search result display unit 15 (step S21). “Generation of drawing command (step S22)” is a process for determining whether or not a drawing command has been issued due to a change in browser window size, a change in scroll position, or the like.

  When the drawing command is generated, the search result display ID, the search result display, and the similarity are drawn (step S23), and the drawn result display ID is added to the result display ID array (step S24). “Re-search (step S25)” is yes when the search button is pressed. Otherwise it is no. Then, a result display ID array is obtained as an output (step S26).

FIG. 8 is a flowchart showing in detail the flow of processing from step S23 (drawing result display ID, result display, similarity) to step S24 (adding the result display ID drawn in the search ID array) in FIG. It is. It is assumed that the search result data (search result ID, result display, similarity) is read in advance from the memory.
In FIG. 8, first, the drawing area of the search result display part of the browser is acquired (step S31). Next, the background of the search result display unit is drawn (step S32). Then, the font size for drawing the search result ID, the result display, and the similarity is obtained (step S33).

  Further, the number of characters that can be drawn in one line is calculated from the drawing area width and the font width (step S34). From the height of the drawing area and the height of the font, the position of the result display data for starting drawing is obtained (step S35). From the height of the drawing area and the height of the font, the number of lines (lm) that need to be drawn is obtained (step S36). Further, the line number counter (I) is set to 1 and initialized (step S37).

Next, when the row number counter (I) is “I <= lm” (step S38), the result display data is drawn one by one. Using the number of characters that can be output in one line, drawing is performed while turning back (step S39). I is increased by the number of lines used to draw the result display data one by one (step S40). Then, the result display ID drawn in the result display ID array is added (step S41). In steps S39, S40, and S41, the line number counter (I) is “I = lm”.
It repeats until it becomes (step S38).

FIG. 9 is a flowchart showing the processing flow of the control unit of the FAQ search system.
In the FAQ search processing flow of the control unit shown in FIG. 9, first, a session ID, a question sentence, a search condition value, and a result display ID array are input as inputs (step S51). In the control unit, the result display ID is registered in the displayed result display table (step S52). Then, a question sentence vector is created from the question sentence (step S53), similar contents are searched, and the search result is output (step S55).

  In “question sentence vector creation” in step S53, a question sentence vector is created from the question sentence, and a document search is performed using the vector space method. This vector space method is generally widely used for document retrieval. Each term is dimensioned, and a question and a document are expressed in a vector space. The term frequency (TF) and document frequency (DF) are more similar. (See, for example, literature (Hiroshi Nakagawa, “Introduction to Natural Language Processing: Realistic Benefits: Information Retrieval”)).

  Hereinafter, two “similar content search” methods in the document search system of the present invention will be described. One is a flow for excluding the displayed result display from the search result (see FIG. 10), and the other is a flow for moving the displayed result display to the end of the search result (see FIG. 14). Both are processing flows performed by the similar content search unit 34 in the FAQ search system 30.

  FIG. 10 is a flowchart (excluding displayed result display) showing the flow of processing of the similar content search unit. In FIG. 10, first, a question text vector and session ID information are input as inputs (step S61). Next, a vector search is performed based on the question sentence vector (step S62). The vector space method described above is used for this vector search.

  Next, the display result display ID is acquired (step S63). In this “displayed result display ID acquisition”, the displayed result display ID is acquired from the displayed result display table using the session ID as a key.

  Next, the displayed result display portion is excluded from the search result (step S64). Here, the display of the displayed result is excluded from the vector search result, and the similar content search is output. Examples of results are shown in FIGS. 11, 12, and 13. FIG.

  In FIG. 11, “reinstall” is input as a question sentence (step S 71), and the information shown in FIG. 11 is obtained as a search result, and only the part surrounded by the thick line indicated by the symbol A in the browser is It is assumed that it is displayed on the screen (step S72). Next, it is assumed that “OS recovery” is input as a question sentence (step S73).

  When “OS recovery” is input as a question sentence (step S73), information on search results shown in FIG. 12 is obtained. In this information, the information indicated with an underline is information displayed on the screen in FIG. 11, and the information indicated with the underline is excluded from the search result (step S74). As a result, the search result shown in FIG. 13 is obtained.

  Next, FIG. 14 shows a flowchart showing a flow of processing for moving the displayed result display to the end of the search result. The flowchart shown in FIG. 14 is basically the same as the flowchart shown in FIG. In “move the displayed result display portion from the search result to the end of the search result” in step S64a, a process of moving the displayed result display portion of the vector search result to the end of the search result is performed. Examples of implementation results are described in FIGS. 15, 16, and 17.

  In FIG. 15, “reinstall” is input as a question sentence (step S 81), and the information shown in FIG. 15 is obtained as a search result, and only the part surrounded by the bold line indicated by the symbol A in the browser It is assumed that it is displayed on the screen (step S82). Next, it is assumed that “OS recovery” is input as a question sentence (step S83).

  When “OS recovery” is input as a question sentence (step S83), information on search results shown in FIG. 16 is obtained. In this information, the information indicated with an underline is the information displayed on the screen in FIG. 15, and the information indicated with the underline is moved to the end of the search result (step S84). As a result, the search result shown in FIG. 17 is obtained.

  As described above, the first embodiment of the present invention has been described. As a result, only the result display seen by the user is excluded from the search result or moved to the end of the search result. Can make content easier to find.

[Second Embodiment]
Next, a second embodiment of the present invention will be described. According to the second embodiment of the present invention, among search results, a search target document having a result display similar to the result display of the search target document displayed on the screen before the previous time, and a search target document browsed before the previous time are displayed. This is an example of lowering the similarity of similar search target documents. That is, the history of the result display is used to reduce the importance of words included in the result display. Further, this is an example of lowering the importance of words included in the browsed search target document.

  The configuration diagram of the system is basically the same as the configuration of FIG. 2 shown in the first embodiment, except that a viewing flag is added to the displayed result display table.

  FIG. 18 shows an example of a displayed result display table in the second embodiment. When the user browses the content, the history of the browsed content is registered in the displayed result display table. When the search I / F unit 12 of the browser displays the content on the content display browser 13, the content acquisition unit 36 receives the session ID in addition to the result display ID from the control unit 31, and displays the viewing flag in the displayed result display table. Register the browsed content history with a field value of 1. In this way, by adding the browsing flag field, it is possible to distinguish which result display is displayed on the screen and which content is browsed.

  The processing flow in the second embodiment is basically the same as the processing flow in the first embodiment, and the search I / F unit and the similar content search processing flow are only partially different. First, the flow of the search I / F unit will be described.

FIG. 19 is a flowchart showing a flow of processing of the search I / F unit in the second embodiment. The flowchart shown in FIG. 19 is basically the same as the flowchart shown in FIG. 6 in the first embodiment, and only differences from the first embodiment will be described.
That is, in “content display” in step S16a, the control unit 31 passes the result display ID and the session ID to the content acquisition unit. After acquiring the content from the content DB 39, the content acquisition unit 36 registers the history of the browsed content in the displayed result display table.

  Next, the processing flow of the similar content search unit shown in FIG. 20 will be described. In FIG. 20, first, information on a question sentence vector and a session ID is input as input (step S101). Next, the question sentence vector is changed (step S102). In the question sentence vector change, the inputted question sentence vector is changed in order to reduce the importance of the displayed result display and the words included in the browsed content (see FIG. 21). Then, a vector search is performed using the changed question sentence vector (step S103). Then, a search result is obtained as an output (step S104).

  FIG. 21 is a flowchart showing the flow of the question sentence vector change process. In FIG. 21, the question text vector and the session ID information are input as inputs (step S111). Then, information on the displayed result display is acquired, and information on the browsed content is acquired (step S112). In the acquired result display acquisition / browsing content acquisition in step S112, the result display ID is searched from the displayed result display table using the session ID as a key, and from the content DB using the searched result display ID as a key, Get the result display. If the content has been browsed, the content is acquired. Then, the displayed result display and the browsed content are morphologically analyzed and broken down into words (step S113).

  Then, the element value of the question sentence vector related to the decomposed word is set to −1 (step S14). As a result, the importance of words included in the result display and the content can be lowered. For example, as shown in FIG. 22, the similarity of content can be reduced by using the changed question sentence vector in which the element value of word 2 is set to -1.

  Note that “decreasing the importance of a word” has the same meaning as “decreasing the similarity of content”. The question vector with the importance of the word lowered is output.

  Next, a supplementary explanation will be given on how to create a question sentence vector. By the way, the method of creating the question sentence vector described here is to set the element value associated with the word included in the question sentence to 1, and set the element value associated with the word not included to 0 (for example, literature (Nakagawa) See Hiroshi, “Introduction to Natural Language Processing: Realistic Benefit: Information Retrieval”).

Example 1) It is assumed that the search target document includes word 1, word 2, and word 3, and the question sentence includes word 1 and word 3, and is as follows.
(Word 1, word 2, word 3) = (1, 0, 1)

There are other ways to create a question text vector. An example is a method in which an element value associated with a word is changed to a TF * IDF value of the word.
Example 2) (Word 1, Word 2, Word 3) = (0.52, 0, 1.9)

Next, another question text vector changing method will be described. As a method for changing the question sentence vector, a method of multiplying the element value associated with the decomposed word by −1 is also conceivable.
Example 3) (word 1, word 2, word 3) = (0.52, 0, 1.9),
(Word 1, word 2, word 3) = (− 0.52, 0, 1.9),
* Words 1 and 2 are included in the decomposed words. The element value of word 2 is 0 even if it is multiplied by -1.

In the above method, since the element value of word 2 is 0, it is 0 even if it is multiplied by -1. In addition to this, when the element value of the word is 0, a method in which the IDF value of the word is multiplied by −1 can be considered.
Example 4) (Word 1, Word 2, Word 3) = (0.52, 0, 1.9),
→ (Word 1, Word 2, Word 3) = (− 0.52, −1.2, 1.9),
* Assume that the IDF value of word 2 is 1.2.

Next, an example of the result of the second embodiment will be described. FIG. 23, FIG. 24, and FIG. 25 are diagrams for explaining an example of the results of the second embodiment.
In FIG. 23, “reinstallation” is input as a question sentence (step S121), and the search result information shown in FIG. 23 is obtained as a search result, and only the portion surrounded by the bold line indicated by the symbol A in the search result is obtained. Is displayed on the browser screen (step S122). Next, it is assumed that “OS recovery” is input as the question text (step S123).

  When “OS recovery” is input as a question sentence (step S123), information on search results shown in FIG. 24 is obtained. In this information, the word indicated with an underline is a word included in the search result displayed on the screen in FIG. 23, and the importance of the underlined word is lowered (step S124). As a result, the search result shown in FIG. 25 is obtained.

  As described above, the second embodiment of the present invention has been described. By this, the content having a result display similar to the displayed result display or the similarity of the content similar to the browsed content is reduced, thereby displaying the content. The search efficiency of the user is further improved.

[Third Embodiment]
Next, a third embodiment will be described. The third embodiment is a function for determining whether the user's search purpose has changed and for enabling / disabling the functions of the first and second embodiments.

  FIG. 26 is a system configuration diagram of the third embodiment. The system configuration diagram shown in FIG. 26 is obtained by adding a question sentence DB 41 to the system configuration diagram of the first embodiment shown in FIG.

  FIG. 27 is a diagram illustrating an example of a question sentence table. The question sentence DB 41 includes a question sentence table. The question sentence at the time of the previous search is registered in the question sentence vector field. This registration is performed by the similar content search unit 34. The processing flow is basically the same as that in the first embodiment, and only the “similar content search” flow is different.

  FIG. 28 is a flowchart showing the flow of processing in the similar content search unit. In FIG. 28, a question sentence vector and a session ID are input (step S131), a vector search is performed (step S132), and a search purpose change is determined after the vector search (steps S133 and S134). When the search purpose has changed (when “changed = yes” is yes), a “displayed result display history deletion (step S135)” process is performed. In this “deleted displayed result display history”, the displayed result display history is deleted from the displayed result display table using the session ID as a key. In the “question sentence table update (step S138)”, the question sentence vector field value of the question sentence table is set to the input question sentence vector.

  In step S134, if the search purpose has not changed (if "changed = yes" is no), a displayed result display ID is acquired (step S136), and the displayed result display portion is excluded from the search result. Is performed (step S137). Further, after the question sentence table is updated (step S138), the search result is output (step S139).

  FIG. 29 is a flowchart showing the flow of processing for determining whether the search purpose has changed. In FIG. 29, the session ID and the search result are input (step S141), and the previous question sentence vector is acquired (step S142). In “previous question sentence vector acquisition” in step S142, the question sentence vector at the time of the previous search is acquired from the question sentence table using the session ID as a key.

  Next, a search is performed using the question sentence vector at the previous search, and a search result is acquired (step S143). Then, a common part between the previous search result and the current search result is obtained (step S144), the current search result number is RC, and the common part search result number is CC (step S145).

  Then, “CC / RC <0.5” is determined (step S146). If the ratio of the previous search results included in the current search result is less than 0.5, the search purpose is changed. Regarded (step S147). If it is 0.5 or more, it is considered that the search purpose has not changed (step S148). Here, 0.5 is used as the threshold value, but the threshold value may be tuned by each search system.

  As described above, the third embodiment of the present invention has been described. By this, when the search purpose is changed, the process of excluding the displayed result display from the search result or invalidating the process is invalidated. , Makes it difficult to overlook the target content.

  As described above, the first, second, and third embodiments of the present invention have been described, but when using the FAQ search system, the user ID is used to display all the past data, not only during the session. You can do the same using the result display history.

  Further, instead of the search I / F unit 12 of the browser 11 passing the result display ID array to the FAQ search system 30, the displayed page number and the range of the displayed search result (for example, the first to fifth of the search results) ). In this way, the network traffic can be reduced. A configuration diagram in this case is as shown in FIG. The question sentence at the previous search is registered in the question sentence DB 41. When the search I / F unit 12 passes the page number and result display range to the FAQ search system 30, the FAQ search system 30 uses the question sentence registered in the question sentence DB 41 to search by the search result acquisition part 42. Reproduce the results. By doing this, the ID of the result display displayed on the page and the ID of the result display included in the search result range can be known. Thereafter, the same processing can be performed by performing the same processing as in the first, second, and third embodiments.

[Fourth Embodiment]
In the first, second, and third embodiments described above, the contents of the present invention have been described using an example using the FAQ search system 30 implemented as a Web application. Other examples include the following.

FIG. 31 is a configuration diagram of a document retrieval system using natural sentences constructed in a local PC. The natural text document search system is configured, for example, by removing the Web server 21 from the FAQ search system 30 implemented as the Web application shown in FIG. The natural sentence search system 20a is implemented as a stand-alone program. The search I / F unit 12a and the content display unit 13a become a user I / F of a stand-alone program.
The processing flow in the search engine 30a is the same if the network communication via the Web server, which was performed by the FAQ search system implemented as a Web application, is omitted.
As a result, the document search system of the present invention can be realized as a document search system using natural sentences built in the local PC.

[Fifth Embodiment]
FIG. 32 is a configuration diagram of a keyword search system constructed in the local PC. The keyword search system 20b shown in FIG. 32 removes the question sentence vector creation unit 33 and the morpheme analysis unit 32 from the natural sentence search system built in the local PC shown in FIG. The search engine 30b is configured in the form of 43. The content DB 39 has a keyword search index instead of a document vector.

The search I / F unit 12a receives a keyword as an input from the user. The keyword search unit 43 uses the keyword search index to search the content DB 39 for content including the keyword received from the search I / F unit 12a.
Thereby, the document search system of the present invention can be realized as a keyword search system built in the local PC.

  As mentioned above, although the embodiment of the document search system of the present invention has been described, in the document search system of the present invention, only the result display seen by the user is excluded from the search result, or moved to the end of the search result. This makes it easier to find content that is not already displayed.

  In addition, when displaying search results (contents) based on a new question sentence, it is possible to reduce the similarity of content that has a result display similar to the displayed result display or content similar to the viewed content. Since the ranking is lowered, the search efficiency of the user is further improved.

  Furthermore, when the search purpose is changed, the process of excluding or moving the displayed result display from the search result is invalidated, so that it is difficult to overlook the target content.

  The FAQ search system server 20, the natural sentence search system 20a, and the keyword search system 20b described above have a computer system therein.

  A series of processes related to the above-described process is stored in a computer-readable recording medium in the form of a program, and the above-described process is performed by the computer reading and executing this program.

  That is, in the FAQ search system server 20, the natural sentence search system 20a, and the keyword search system 20b, a central processing unit such as a CPU reads the above program into a main storage device such as a ROM or a RAM, This is realized by executing processing and arithmetic processing.

  Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program.

  The embodiment of the present invention has been described above. However, the document search system of the present invention is not limited to the illustrated example described above, and various modifications can be made without departing from the scope of the present invention. Of course.

The figure which shows the network structural example of the document search system of this invention. The figure which shows the structural example of the FAQ search system of the 1st Embodiment of this invention. The figure which shows the example of the browser and content display which a search I / F part loads. The figure which shows the example of a weighting dictionary and content DB. The figure which shows the example of a display result display table and a search result. The flowchart which shows the flow of a process of a search I / F part. The flowchart which memorize | stores the result display ID by which the browser was displayed. The flowchart which shows the flow of a drawing process and the addition process of result display ID. The flowchart which shows the flow of a process of the control part of a FAQ search system. The flowchart which shows the flow of the process which excludes a display result display. The figure which shows the example of an implementation result which excludes a displayed result display. The figure which shows the part following FIG. The figure which shows the part following FIG. The flowchart which shows the process which moves the displayed result display to the end. The figure which shows the example of an implementation result which moves the displayed result display to the end. The figure which shows the part following FIG. The figure which shows the part following FIG. The figure which shows the example of the displayed result display table in 2nd Embodiment. 9 is a flowchart illustrating a flow of processing of a search I / F unit according to the second embodiment. The flowchart which shows the flow of a process of a similar content search part. The flowchart which shows the flow of a process of a question sentence vector change. The figure which shows the example of a change of a question sentence vector. The figure which shows the example of an implementation result of 2nd Embodiment. The figure which shows the part following FIG. The figure which shows the part following FIG. The system block diagram of 3rd Embodiment. The figure which shows the example of the question sentence table of 3rd Embodiment. The process flowchart of the similar content search part of 3rd Embodiment. The flowchart which shows the flow of the process which determines the change of the search objective. The system block diagram in the case of passing a page number etc. instead of a result display ID array. The block diagram of the natural-text document search system constructed | assembled in local PC. The block diagram of the keyword search system constructed | assembled in local PC.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 FAQ search system (document search system), 2 Internet 10 Terminal device, 11 Browser, 12, 12a Search I / F part 13 Content display browser, 13a Content display part 14 Question sentence input part, 15 Search result display part, 16 Search Condition setting unit 17 Search button, 20 FAQ search system server 20a Natural sentence search system, 20b Keyword search system 21 Web server, 22 Session ID creation unit, 30 FAQ search system 30a, 30b Search engine, 31 Control unit, 32 Morphological analysis unit 33 Question sentence vector creation unit, 34 Similar content search unit 35 Displayed result display registration unit, 36 Content acquisition unit 37 Morphological analysis dictionary 38 Weighting dictionary, 39 Content DB
40 Displayed result display DB, 41 Question sentence DB
42 search result acquisition unit, 43 keyword search unit

Claims (13)

A document search system that searches a search target document based on an inputted question sentence and outputs a search result,
Among the search results, a displayed result display registration means for recording a history of search results displayed on the screen before the previous time;
When a search is performed using a new question sentence and a search result is obtained, a part common to the history part of the search result recorded in the displayed result display registration unit is excluded from the new search result, or And a means for moving the common part to the end of a new search result. Search interface means for controlling the size of the screen for displaying the search results and the scroll range of the screen;
5. Displayed result display registration means for receiving search result screen display information from the search interface means and recording a history of search results displayed on the screen based on the screen display information. The document search system according to 1. Means for outputting the search results together with the similarity to the question sentence in order from the highest similarity;
Among the search results, information on the history of the display of the search target document displayed on the screen before the previous time, and a displayed result display registration means for storing information on the history of the search target document browsed before the previous time,
When a search is performed using a new question sentence and a search result is obtained, a search having a result display similar to the result display of the search target document recorded in the displayed result display registration unit among the new search results 2. A means for reducing similarity of a target document and a similarity of a search target document similar to a search target document browsed before the previous time recorded in the displayed result display registration unit. Or the document search system of Claim 2. The search result by the previous question sentence is compared with the search result by the current question sentence. When the ratio of the common search results is small, the search purpose is considered to be changed, and the result display recorded in the displayed result display registration means is displayed. The document search system according to claim 1, further comprising: means for deleting the history information. Means for memorizing the previous question,
The document search system according to claim 4, further comprising: means for searching for a search target document using both the previous question sentence and the current question sentence, and comparing both search results. A document search system that searches a search target document by an inputted question sentence, and outputs a search result together with a similarity degree to the question sentence, in order from a high similarity degree,
Among the search results, information on the history of the display of the search target document displayed on the screen before the previous time, and a displayed result display registration means for storing information on the history of the search target document browsed before the previous time,
When a search is performed using a new question sentence and a search result is obtained, a search having a result display similar to the result display of the search target document recorded in the displayed result display registration unit among the new search results A document search system comprising: a similarity of a target document; and a means for reducing the similarity of a search target document similar to a search target document viewed before the previous time recorded in the displayed result display registration unit . Decrease the importance of the word included in the history of the result display recorded in the displayed result display registration means, and reduce the importance of the word included in the browsed search target document recorded in the displayed result display registration means. The document search system according to claim 3, further comprising a means for reducing the similarity by lowering. The search result by the previous question sentence is compared with the search result by the current question sentence. When the ratio of the common search results is small, the search purpose is considered to be changed, and the result display recorded in the displayed result display registration means is displayed. The document search system according to claim 6 or 7, further comprising means for deleting the history information. Means for memorizing the previous question,
The document search system according to claim 8, further comprising means for searching for a search target document using both the previous question sentence and the current question sentence, and comparing both search results. A Web that is connected to a terminal device having a browser for inputting the question text and displaying a search result through a communication network, receives information on the question text from the terminal device, and transmits information on the search result to the terminal device The document search system according to claim 1, further comprising a server. Search for documents to be searched using the entered question text, and output the search results to a computer in the document search system.
Of the search results, a history of search results displayed on the screen before the previous time is recorded in a displayed result display registration unit;
When a search is performed using a new question sentence and a search result is obtained, a part common to the history part of the search result recorded in the displayed result display registration unit is excluded from the new search result, or And a step of moving the common part to the end of a new search result. Search for documents to be searched using the entered question text, and output the search results to a computer in the document search system.
Of the search results, a history of search results displayed on the screen before the previous time is recorded in a displayed result display registration unit;
When a search is performed using a new question sentence and a search result is obtained, a part common to the history part of the search result recorded in the displayed result display registration unit is excluded from the new search result, or And a step of moving the common part to the end of a new search result. A document search method in a document search system for searching a search target document by an inputted question sentence and outputting a search result,
The computer in the document search system records, in the displayed result display registration unit, a history of search results displayed on the screen before the previous search results,
When the computer performs a search using a new question sentence and a search result is obtained, a part common to the history part of the search result recorded in the displayed result display registration unit is excluded from the new search result. Or a step of moving the common part to the end of a new search result.

Images