JP5104329B2 - Document search system - Google Patents

Description

  The present invention relates to a document search system in which a terminal and a server are connected via a network, and the server presents related keywords of search keywords received from the terminal.

  In a general document search system, a target document is searched for a keyword input by a user, and a storage location of the search result document is presented. Here, the document is, for example, a text file, an HTML (Hyper Text Markup Language) file, an XML (Extensible Markup Language) file, an image file, other document layout software, word processing software, spreadsheet software, etc. Includes files.

  Also, recently, with reference to a keyword entered by a user (referred to as keyword 1), a past search history (search history of all users) is referred to, and a keyword used together with keyword 1 or keyword 1 is selected. There is a service that presents keywords that include them. For example, when the user inputs “curry”, “curry recipe”, “how to make curry”, and “curry udon” are presented as the top three search keywords in the past to support the user's search operation. However, when referring to past search histories, effective keywords cannot be presented when documents are updated daily.

On the other hand, a service that refers to a search target document and presents related words of a search keyword has also been proposed (Patent Document 1). In Patent Document 1, in order to present highly accurate words as related words, combinations of related words are arranged in the order of “noun, parallel particle, noun” from documents to be searched.
JP 2004-94388 A

  However, in the mechanism of Patent Document 1, there is a possibility that the number of related words that can be presented is reduced because the accuracy of related words is pursued. That is, there are many cases where related words cannot be presented for the input search keyword. In addition, the presented related words have a very high degree of relevance, so that they are not spread. For example, when the search target document is facility information, the user often decides “what he wants to do”, “where he wants to go”, “what he wants to eat”, etc. during the search process. In such a case, in the mechanism of Patent Document 1, only words that have a grammatical parallel relationship are used as related words, and thus the user's search operation cannot be fully supported.

  The present invention has been made in view of the above-described problems, and the object of the present invention is included in a search target document with respect to an input search keyword, and has a broad (words in a grammatical parallel relationship). It is to provide a document search system that presents related keywords (not limited to only) and can support a user's search work.

  In order to achieve the above object, a first invention is a document search system in which a terminal and a server are connected via a network, and the server presents a related keyword of a search keyword received from the terminal, The server defines a word type as a word attribute, defines a combination of the word attributes in which the word attribute appears within a certain range in the search target document as an appearance pattern, and calculates a word group relevance Means for holding word attribute appearance pattern information in which the weight for each appearance pattern is defined as the appearance pattern weight, and searching for a word group that matches the appearance pattern in the search target document, Relevance calculating means for calculating the word group relevance based on the distance and the appearance pattern weight, and the relevance calculating means. Means for holding the calculated word group relevance as word relevance information, and related keyword presenting means for presenting a related keyword of the search keyword with reference to the word relevance information when the search keyword is received from the terminal And a document search system characterized by comprising: In the document search system according to the first aspect of the present invention, it is possible to support a user's search work by presenting related keywords that are included in the search target document and have a broader range for the input search keyword.

  The word group relevance in the first invention is, for example, the sum of the product of the appearance pattern weight and the reciprocal of the distance of the word group for all previously registered words and all the appearance patterns. It is.

  In addition, it is desirable that the appearance pattern weights can be set to values in both cases where the appearance order of the registered word groups is identical and in other cases. Thus, if an appropriate value is set, the accuracy of the related keywords to be presented can be improved (here, high accuracy means that, for example, there is a high probability that search operations can be supported for many users) Say.)

  The word attribute appearance pattern information is preferably updatable. As a result, even when the amount and content of the search target document change, the accuracy of the related keywords can be maintained by appropriately updating the word attribute appearance pattern information.

  The related keyword presenting means preferably presents the related keyword together with a search result based on the search keyword and / or a word attribute related to the related keyword. Thus, the user can determine whether or not to make a search request again by comparing the search result based on the search keyword input by himself with the presented related keyword. In addition, the user can automatically obtain broad related keyword information (without thinking by the user himself / herself) for the input search keyword.

  A second invention is a server that is connected to a terminal via a network and presents a related keyword of a search keyword received from the terminal, wherein a word type is defined as a word attribute, and the word attribute is included in a search target document. Word attribute appearance pattern information in which a combination of the word attributes appearing within a certain close range is defined as an appearance pattern, and a weight for each appearance pattern when calculating the word group relevance is defined as an appearance pattern weight Relevance calculating means for searching for a word group that matches the appearance pattern in the search target document, and calculating a word group relevance by the distance of the searched word group and the appearance pattern weight; A means for holding the word group relevance calculated by the relevance calculation means as word relevance information; Upon receiving the keyword, a server, characterized by comprising, a related keyword presentation means for presenting the related keywords of the search keyword by referring to the word relevance information.

  In a third invention, a terminal and a server are connected via a network, the server defines a word type as a word attribute, and the word attribute appears within a certain range in the search target document. A combination of word attributes is defined as an appearance pattern, word attribute appearance pattern information is defined in which the weight for each occurrence pattern when calculating the word group relevance is defined as an appearance pattern weight, and the server receives from the terminal A document search method for presenting related keywords of a search keyword, wherein the server searches a word group that matches the appearance pattern in the search target document, and the distance between the searched word group and the appearance pattern weight And calculating the word group relevance by the server and calculating the word group relevance by the server. Holding the word group relevance as word relevance information, the terminal transmitting the search keyword to the server, and the server referring to the word relevance information and receiving the search And a step of presenting a keyword related to the keyword.

  A fourth invention is a program for causing a computer to function as the server described in the second invention.

  According to the present invention, it is possible to provide a document search system capable of supporting a user's search operation by presenting a wide range of related keywords included in a search target document with respect to an input search keyword. it can. Such a document search system can sufficiently support a user's search work when the search target document is facility information, for example.

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

  First, a schematic configuration of a document search system 1 according to an embodiment of the present invention will be described with reference to FIG.

  FIG. 1 is a diagram showing a schematic configuration of a document search system 1. As shown in FIG. 1, the document search system 1 includes an administrator terminal 3, an index creation server 5, a search server 7, a web server 9, a user terminal 11, and the like connected via a network (not shown). The

  The index refers to data that is created in order to quickly search for related words related to related keywords presented to the user. In the embodiment of the present invention, an index is created, and related word information (related information with a search word, storage location information, related word attribute information, etc.) is divided into a plurality of files and held.

  The administrator terminal 3 is a terminal used by the administrator of the document search system 1 and communicates with the index creation server 5. The administrator transmits data used in the document search system 1 to the index creation server 5 via the administrator terminal 3. The administrator may directly input data used in the document search system 1 to the index creation server 5.

  The index creation server 5 is a server that creates an index of related keywords, and communicates with the administrator terminal 3, the search server 7, and the web server 9.

  The search server 7 is a server that holds related keyword data, and communicates with the index creation server 5 and the web server 9.

  The web server 9 is a server that responds to a search request transmitted from the user terminal 11, and communicates with the index creation server 5 and the search server 7.

  The user terminal 11 is a terminal used by a user of the document search system 1 and communicates with the web server 9. When the document search system 1 is implemented, the user terminal 11 basically does not need to have a special function if a commercially available OS (Operating System) software and web browsing software are installed.

  The index creation server 5, the search server 7, and the web server 9 may be configured by a single device. The index creation server 5, the search server 7, and the web server 9 may be provided with a plurality of devices having the same function in consideration of load distribution, fault tolerance, and the like.

  Hereinafter, when the index creation server 5, the search server 7, and the web server 9 are not distinguished or collectively referred to, they are simply referred to as “servers”. Further, when the administrator terminal 3 and the user terminal 11 are not distinguished or collectively referred to, they are simply referred to as “terminals”.

  Next, the hardware configuration of each device will be described with reference to FIG.

FIG. 2 is a hardware configuration diagram of a computer that realizes a terminal and a server. Note that the hardware configuration in FIG. 2 is an example, and various configurations can be adopted depending on the application and purpose.
A computer that realizes a terminal and a server includes a control unit 21, a storage unit 23, a media input / output unit 25, a communication control unit 27, an input unit 29, a display unit 31, a peripheral device I / F unit 33, and the like via a bus 35. Connected.

  The control unit 21 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like.

The CPU calls a program stored in the storage unit 23, ROM, recording medium or the like to a work memory area on the RAM and executes it, and drives and controls each device connected via the bus 35, which is performed by the terminal and the server. The processing described later is realized.
The ROM is a non-volatile memory and permanently holds a computer boot program, a program such as BIOS, data, and the like.
The RAM is a volatile memory, and temporarily stores programs, data, and the like loaded from the storage unit 23, ROM, recording medium, and the like, and includes a work area used by the control unit 21 for performing various processes.

The storage unit 23 is an HDD (hard disk drive), and stores a program executed by the control unit 21, data necessary for program execution, and the like. As for the program, a control program corresponding to the OS and an application program corresponding to processing to be described later are stored.
Each of these program codes is read by the control unit 21 as necessary, transferred to the RAM, read by the CPU, and executed as various means.

  The media input / output unit 25 (drive device) inputs / outputs data, for example, a CD drive (-ROM, -R, -RW, etc.), DVD drive (-ROM, -R, -RW, etc.), MO drive, etc. And other media input / output devices.

  The communication control unit 27 has a communication control device, a communication port, and the like, is a communication interface that mediates communication between the computer and the network 13, and controls communication with other computers via the network 13.

The input unit 29 inputs data and includes, for example, a keyboard, a pointing device such as a mouse, and an input device such as a numeric keypad.
An operation instruction, an operation instruction, data input, and the like can be performed on the computer via the input unit 29.

  The display unit 31 includes a display device such as a CRT monitor and a liquid crystal panel, and a logic circuit (such as a video adapter) for realizing a video function of the computer in cooperation with the display device.

  The peripheral device I / F (interface) unit 33 is a port for connecting a peripheral device to the computer, and the computer transmits and receives data to and from the peripheral device via the peripheral device I / F unit 33. The peripheral device I / F unit 33 is configured by USB, IEEE 1394, RS-232C, or the like, and usually has a plurality of peripheral devices I / F. The connection form with the peripheral device may be wired or wireless.

  The bus 35 is a path that mediates transmission / reception of control signals, data signals, and the like between the devices.

  Next, the software configuration of the server will be described with reference to FIG.

  FIG. 3 is a software configuration diagram of the server. As shown in FIG. 3, the server includes a related keyword index creation application 41, a related keyword search application 43, and the like.

  The related keyword index creation application 41 is installed in the index creation server 5 (however, a part input from the administrator terminal 3 is installed in the administrator terminal 3 if necessary), and word information registration 51, It has functions such as document registration 52, index creation 53, index distribution 54, and index creation history reference 56. Furthermore, the function of the index creation 53 executes processes such as an index update target document discrimination process 71, a morpheme analysis process 72, a word conversion process 73, a keyword assist use word selection process 74, a relevance calculation process 75, and an index creation process 76. To do. Details of the functions and processing will be described later with reference to FIGS.

  The keyword assist means that the related keyword of the search keyword is presented to assist the user's search operation.

  The related keyword index creation application 41 includes word attribute appearance pattern information 101, word attribute information 102, word conversion information 103, document information 104, updated document history information 105, index execution processing history information 106, update target document information 107, Morphological analysis result information 108, word conversion result information 109, relevance calculation target word information 110, word relevance information 111, index creation processing history information 112, keyword assist hash information 113, keyword assist search word information 114, keyword assist related words Data such as information 115 is temporarily or semi-permanently retained. Details of the data will be described later with reference to FIGS.

  The related keyword search application 43 is installed in the search server 7 (however, a part responding to a request from the user terminal 11 is installed in the web server 9), and has functions such as a related keyword search 55. Details of the function will be described later with reference to FIGS.

  Further, the related keyword search application 43 semipermanently holds data such as the keyword assist hash information 113, the keyword assist search word information 114, and the keyword assist related word information 115 distributed from the related keyword index creation application 41. Details of the data will be described later with reference to FIGS.

  Next, a work flow in the document search system 1 will be described with reference to FIG.

  FIG. 4 is a diagram showing an outline of a work flow in the document search system 1. As shown in FIG. 4, the work in the document search system 1 is performed in the order of word information definition 61, document collection / update 62, index creation 63, index distribution 64, search request acceptance 65, and the like.

  In the word information definition 61, the administrator performs registration / update / deletion work of the word attribute appearance pattern information 101, the word attribute information 102, the word conversion information 103, and the like. At this time, the function of the word information registration 51 included in the related keyword index creation application 41 is used. It is desirable to define the word information 61 in a batch before operating the system. Moreover, it is desirable for the administrator to periodically define the word information definition 61 even after the system is operated.

  In the document collection / update 62, for example, an administrator registers, updates, and deletes the document information 104 and the like. At this time, the function of the document registration 52 provided in the related keyword index creation application 41 is used. Further, for example, a computer may perform an automatic document collection operation.

  In the index creation 63, the index creation server 5 creates keyword assist hash information 113, keyword assist search word information 114, keyword assist related word information 115 and the like. At this time, the function of the index creation 53 provided in the related keyword index creation application 41 is used. The index creation 63 may be executed according to an instruction from the administrator, or may be executed periodically such as once a day by batch processing.

  In the index distribution 64, the index creation server 5 distributes the keyword assist hash information 113, the keyword assist search word information 114, the keyword assist related word information 115, and the like. At this time, the function of index distribution 54 included in the related keyword index creation application 41 is used.

  In the search request reception 65, the web server 9 performs a search request reception operation from the user terminal 11. Further, the search server 7 performs a search operation for the search request. At this time, the function of the related keyword search 55 provided in the related keyword search application 43 is used. The search request reception 65 may be performed during the work of document collection / update 62, index creation 63, and index distribution 64. However, if the document collection / update 62, index creation 63, and index distribution 64 have not been performed, the number of search results is zero.

  Although not shown, the administrator can also refer to the index creation history when working with the word information definition 61. At this time, the administrator refers to the index creation processing history information 112 using the function of the index creation history reference 56 provided in the related keyword index creation application 41. The index creation processing history information 112 includes history data of index creation processing executed in the past.

  Hereinafter, the functions, processes, and data shown in FIG. 3 will be described in detail for each operation shown in FIG. 4 with reference to FIGS.

(Definition 61 of word information)
With reference to FIGS. 5 to 7, functions related to the work of the word information definition 61 will be described. In the embodiment of the present invention, a case where the search target document is facility information will be described as an example.

  FIG. 5 is a diagram illustrating an example of the word attribute appearance pattern information 101. The appearance pattern is a type of what kind of word appears in what order in the document.

  The administrator defines a word type as a word attribute, defines a combination of the word attributes in which the word attribute appears within a certain range in the search target document as an appearance pattern, and determines the degree of association between words. The weight for each appearance pattern at the time of calculation is defined as the appearance pattern weight. That is, the word attribute appearance pattern information 101 includes data items such as an appearance pattern ID, a plurality of word attributes defined for each appearance pattern, and an appearance pattern weight. For example, up to 128 appearance patterns can be registered.

  The number of appearance patterns is the accuracy of the related keywords to be presented (here, high accuracy means, for example, that there is a high probability that search operations can be supported for many users), search processing performance, etc. Therefore, it is desirable to perform maintenance from time to time even after system operation. Specifically, when the contents of many documents are changed, or when the number of documents greatly increases or decreases, the administrator desirably updates the word attribute appearance pattern information 101. In the embodiment of the present invention, the word attribute appearance pattern information 101 is configured to be easily updated by an administrator.

  Types of words, that is, word attributes are, for example, “place name words” that are words related to place names, etc., “store name words” that are words related to store names, etc., “cooking words” that are words related to dish names, etc. “Food word” that is a word related to “station name”, “Station name word” that is a word related to the name of the station, etc. There are two or more word attributes arranged for each appearance pattern, for example, a maximum of eight.

  The significance of the appearance pattern weight is set to a high value for an appearance pattern in which the administrator browses a group of documents to be searched (or may be a request from the document creator) and has a high probability of specifying each document, for example. It is possible to improve the accuracy of related keywords to be presented. In addition, the appearance pattern weight can be set for both the case where the appearance order of the registered word groups is identical and the other case. This further increases the accuracy of the related keywords to be presented.

  As shown in FIG. 5, in the data with the appearance pattern ID “1”, the appearance pattern weights when the appearance order of the word attributes is “place name word” and “store name word” and the appearance order of the word groups all match. Is “1”, and the appearance pattern weight in other cases is “0.5”. In addition, in the data with the appearance pattern ID “2”, the appearance pattern weights when the appearance order of the word attributes is “store name word”, “cooking word”, and “food word” and the appearance order of the word groups all match. Is “1”, and the appearance pattern weight in other cases is “0.5”. In addition, in the data with the appearance pattern ID “3”, the appearance order of the word attributes is “place name word” and “station name word”, and the appearance pattern weight is “1” when the appearance order of the word groups all match. In other cases, the appearance pattern weight is “0.5”.

  FIG. 6 is a diagram illustrating an example of the word attribute information 102. The administrator browses a group of documents to be searched (or may be a request from the document creator) and determines a word to be used as a search keyword or a related keyword. Then, the administrator defines word attributes of words used as search keywords or related keywords. Basically, only the words defined here are used in the processing described later.

  The word attribute information 102 includes data items such as a word ID, a word, and a word attribute. As shown in FIG. 6, the data with the word ID “1” has the word “Hokkaido” and the word attribute “place name word”. The data with the word ID “4” has the word “McDonald (registered trademark)” and the word attribute “store name word”.

  FIG. 7 is a diagram illustrating an example of the word conversion information 103. A word generally has a plurality of character strings having the same meaning depending on the shaking of the notation, abbreviations, and the like. In order to handle these character strings having the same meaning as one word, the administrator desirably registers the word conversion information 103.

  The word conversion information 103 includes data items such as a pre-conversion word and a post-conversion word. As shown in FIG. 7, the data in which the pre-conversion word is “Mcdo (one of abbreviations for McDonald's (registered trademark))” is “Macdonald (registered trademark)”.

(Document collection / update 62)
Next, functions related to the work of document collection / update 62 will be described with reference to FIG.

  FIG. 8 is a diagram illustrating an example of data items of the document information 104. The administrator performs registration, update, and deletion of the document information 104 related to the document that the server finally presents to the user (more precisely, the server presents the document storage location). In addition, the computer may perform an automatic document collection operation, and may register, update, or delete the document information 104.

  As illustrated in FIG. 8, the document information 104 includes data items such as a document ID, a URI (Uniform Resource Identifier), a site ID, a title, a category, a keyword, a text, a last update date, a download time, and a status.

  The document ID is a unique ID number of the document. The URI is a document URL (Uniform Resource Locator). The site ID is the site name of the site where the document exists. The title is the title of the document. The category is a category name of the document. The keyword is a keyword added to the document. The text is the text of the document. The last update date is the last update date of the document. The download time is the time when the document is downloaded. The status is a numerical value indicating new or updated / no change / deleted.

  Data items used in the embodiment of the present invention are a document ID, a title, a text, a state, and the like.

(Index creation 63 and index distribution 64)
Next, functions and the like regarding the work of index creation 63 and index distribution 64 will be described with reference to FIGS.

  FIG. 9 is a diagram illustrating a data flow in the index update target document determination processing 71. As shown in FIG. 9, in the index update target document determination processing 71, the server creates update target document information 107 by referring to the update document history information 105 and index execution processing history information 106 based on the document information 104. Here, the update target document information 107 is not necessary when the next process (morpheme analysis process 72) is completed, and may be temporarily held.

  For example, when an administrator instructs creation of an index, whether to create a new index from all documents (new creation mode), or to update an existing index only for newly registered / updated / deleted differences ( Update mode) may be selected. In the case of batch processing, for example, the new creation mode may be set only on a specific day of the week or the last day of the month, and the update mode may be set on other days.

  The server refers to the selected mode, the updated document history information 105, and the index execution processing history information 106, extracts the document to be indexed from the document information 104, and creates the update target document information 107.

  FIG. 10 is a diagram showing a data flow in the morphological analysis process 72. As shown in FIG. 10, in the morphological analysis process 72, the server creates morphological analysis result information 108 based on the update target document information 107. Here, the morphological analysis result information 108 is not necessary when the next process (word conversion process 73) is completed, and may be temporarily held.

  The server generates the morpheme analysis result information 108 by decomposing the document title and body from the update target document information 107 into morpheme levels. As shown in FIG. 10, the server, for example, with respect to a character string “McDonald's (registered trademark) 3 minutes from Ebisu Station”, “Ebisu / Station / From / 3 / min / of / However, it is decomposed into the morpheme level as follows: / N / A / McDonald (registered trademark).

  FIG. 11 is a diagram illustrating a data flow in the word conversion process 73. As shown in FIG. 11, in the word conversion process 73, the server creates word conversion result information 109 based on the morphological analysis result information 108. Here, the word conversion result information 109 is not necessary when the next process (keyword assist using word selection process 74) is completed, and may be temporarily held.

  The server refers to the word conversion information 103 and converts the morpheme analysis result information 108 to create word conversion result information 109 in order to unify fluctuations in word notation and the like. As shown in FIG. 11, “Mac (one of abbreviations for McDonald's (registered trademark))”, “Mcdo (one of abbreviations for McDonald's (registered trademark))”, “Macdonald (registered trademark)”, “McDonald ' Since “s (registered trademark)” has the same meaning, that is, the name of the store “McDonald (registered trademark)”, it is converted to “McDonald (registered trademark)”.

  FIG. 12 is a diagram illustrating a data flow in the keyword assist using word selection process 74. “Example of part-of-speech classification 116 by a specific morphological analysis engine” shown in FIG. 12 is an example of a part-of-speech word used for keyword assist, and the part-of-speech classification is not limited to this. As shown in FIG. 12, in the keyword assist using word selection process 74, the server uses the word assist information 109 and the keyword assist exemplified by the word attribute information 102 and the part-of-speech classification example 116 by a specific morphological analysis engine. The relevance calculation target word information 110 is created by referring to the part-of-speech classification information of the word used for the above.

The server selects words used for keyword assist in order to perform keyword assist at high speed. Specifically, the server extracts a word from the word conversion result information 109 on the condition that the part-of-speech word shown in the part-of-speech 116 of the word used for keyword assist and the word registered in the word attribute information 102 are related. The degree calculation target word information 110 is created.
It should be noted that the part of speech of the word used for the keyword assist described above may include other than nouns depending on the content, usage, etc. of the document to be searched.

  FIG. 13 is a diagram illustrating a processing flow of the relevance calculation processing 75. FIG. 14 is a diagram illustrating a data flow in the relevance calculation process 75. Hereinafter, the relevance calculation process 75 will be described with reference to FIG.

  In the degree-of-association calculation process 75, the server searches a word group that matches the appearance pattern in the search target document, and calculates the degree of word group relevance based on the distance of the searched word group and the appearance pattern weight. In addition, the server holds the calculated word group association degree as word association degree information 111.

  As shown in FIG. 13, the control unit 21 of the server substitutes 1 for the subscript k (k is a natural number) of the document ID (step 1001), and substitutes 0 for the word extraction count n (n is 0 or a natural number). (Step 1002), 1 is substituted for the subscript m (m is a natural number) of the appearance pattern (Step 1003), and initialization processing is performed. Here, the subscript k of the document ID is a number in the column of the document ID in the table related to the relevance calculation target word information 110 shown in FIG. In FIG. 14, 1, 2,... 862 are illustrated as an example. Note that the value in the appearance position column of the same table indicates the number of words that the word appears in the document. Further, the word extraction number n is a value indicating the number of times the word extraction 81 is performed with the arrow of the word extraction 81 shown in FIG. Further, the subscript m of the appearance pattern is the head number of the table related to the word attribute appearance pattern information 101 shown in FIG. In FIG. 14, 1, 2, and 3 are illustrated as an example.

  Next, the control unit 21 of the server starts with the (n × P−n) -th (P is a natural number) word registered in the relevance calculation target word information 110 in the k-th document, P A word is extracted (step 1004). Here, the extraction number P is, for example, 100. Further, if the extraction number P is set to a large value, it is possible to extract the word of the entire document and execute subsequent processing for most documents. In FIG. 14, the word extracted in step 1004 is represented by the table indicated by the arrow of the word extraction 81.

  Next, the control unit 21 of the server calculates relevance levels for all word groups that match the mth appearance pattern (step 1005), and updates the word relevance information 111 (step 1006). Specifically, by repeating Step 1005 and Step 1006, the server control unit 21 calculates the product of the appearance pattern weight and the reciprocal of the distance of the word group for all the words registered in advance (however, the search target (Only words that appear in the document.) And all occurrence patterns are added together to obtain the word group relevance. As described above with reference to FIG. 5, the appearance pattern weights can be set to values in both cases where the appearance order of the registered word groups is identical and in other cases.

  The arrow of the relevance calculation 82 shown in FIG. 14 represents an example when step 1005 and step 1006 are executed only for two data (corresponding to data (1) and data (2) described later). Specifically, the control unit 21 of the server obtains data whose appearance pattern subscript m is “1” from the word attribute appearance pattern information 101. Next, the control unit 21 of the server extracts a word group whose word attributes are “place name word” and “cooking word”. In the example shown in FIG. 14, the words “Tokyo” (appearance position “10”) and “curry” (appearance position “87”) (data (1)), and the word “curry” (appearance position “ 87 ”) and“ Tokyo ”(appearance position is“ 928 ”) (data (2)). Next, the control unit 21 of the server calculates the product of the appearance pattern weight and the reciprocal of the distance. First, in the case of data (1), since the appearance order of word attributes related to the registered word group (first “place name word” and then “cooking word”) all match, the appearance pattern weight is “1”. Become. The distance is a value obtained by subtracting the appearance position “10” of the word “Tokyo” from the appearance position “87” of the word “curry”, for example. On the other hand, in the case of data (2), the appearance pattern weight is “0.5” because it does not coincide with the appearance order of the word attributes related to the registered word group. The distance is a value obtained by subtracting the appearance position “87” of the word “curry” from the appearance position “928” of the word “Tokyo”. This is expressed by an equation: 1 × (1 / (87−10)) + 0.5 × (1 / (928−87)) = 0.01358. Therefore, in the table related to the word association degree information 111 shown in FIG. 14, the record in which the word 1 is “Tokyo” and the word 2 is “curry” has the association degree “0.0136”.

  In the example shown in FIG. 14, the distance is calculated from the difference in appearance position, but other calculation methods may be used. For example, normalization may be performed using the total number of words included in each document.

Returning to the description of FIG. Next, the control unit 21 of the server substitutes m + 1 for the subscript m of the appearance pattern (step 1007), and checks whether m> M (M is the final number of the subscript of the appearance pattern) is satisfied (step 1008).
When the condition is not satisfied (No in Step 1008), the server control unit 21 repeats from Step 1005.
When the condition is satisfied (Yes in Step 1008), the server control unit 21 proceeds to Step 1009.

Next, the control unit 21 of the server substitutes n + 1 for the number n of word extractions (step 1009), and checks whether or not P words are extracted in step 1004 (step 1010).
If the condition is satisfied (Yes in step 1010), the control unit 21 of the server repeats from step 1003.
When the condition is not satisfied (No in Step 1010), the server control unit 21 proceeds to Step 1011.

Next, the control unit 21 of the server substitutes k + 1 for the subscript k of the document ID (step 1011), and checks whether k> K (K is the final number of the subscript of the document ID) is satisfied (step 1012).
When the condition is not satisfied (No in Step 1012), the control unit 21 of the server repeats from Step 1002.
If the condition is satisfied (Yes in step 1012), the control unit 21 of the server ends the process.

  FIG. 15 is a diagram showing a data flow in the index creation processing 76. As shown in FIG. 15, in the index creation processing 76, the server creates keyword assist hash information 113, keyword assist search word information 114, and keyword assist related word information 115 based on the word association degree information 111. Here, the keyword assist hash information 113, the keyword assist search word information 114, and the keyword assist related word information 115 are divided and held by dividing data corresponding to the index data 117 shown in FIG. 15 in order to speed up the search process. Yes.

  The functions related to the work of index distribution 64 shown in FIG. 4 are not particularly shown. The index distribution 64 is an operation of distributing the keyword assist hash information 113, the keyword assist search word information 114, and the keyword assist related word information 115 to an apparatus that installs the related keyword search application 43.

(Retrieval request reception 65)
Next, with reference to FIG. 16, functions and the like related to the work of the search request reception 65 will be described.

  FIG. 16 is a diagram showing a flow of processing related to the search request reception 65. In the processing related to the search request reception 65, the server receives a search word input as a search keyword by the user and searches for a related word.

  As shown in FIG. 16, the control unit 21 of the server performs trimming and character conversion of the search word (step 2001). Search word trimming and character conversion are (1) deleting blanks at both ends of the search word, (2) conversion from full-width alphanumeric characters to half-width alphanumeric characters, (3) conversion from half-width katakana to full-width katakana, (4 ) Conversion from uppercase letters to lowercase letters.

Next, the control unit 21 of the server searches the search word file position from the keyword assist hash information 113 using a predetermined hash function (step 2002), and checks whether the target data exists (step 2003). .
When the target data exists (Yes in Step 2003), the control unit 21 of the server proceeds to Step 2004.
When the target data does not exist (No in Step 2003), the server control unit 21 ends the process.

Next, the control unit 21 of the server searches the related word file position from the keyword assist search word information 114 (step 2004), and confirms whether the target data exists (step 2005).
When the target data exists (Yes in Step 2005), the control unit 21 of the server proceeds to Step 2006.
When the target data does not exist (No in Step 2005), the server control unit 21 ends the process.

Next, the control unit 21 of the server obtains a list of related words from the keyword assist related word information 115 (step 2006). Here, it is desirable to obtain the degree of association associated with the related word and the word attribute together with the related word. The acquired related word, relevance degree, and word attribute are transmitted to the user terminal by the control unit 21 of the server as related keywords. The related keywords to be transmitted may be all of the acquired related words or the like, may be those having a relevance level equal to or higher than a threshold value, and may be those having a high relevance level.
Note that the control unit 21 of the server may separately perform a search based on a search keyword input by the user, and present a related keyword together with the search result based on the search keyword. Thus, the user can determine whether or not to make a search request again by comparing the search result based on the search keyword input by himself with the presented related keyword. Moreover, you may make it the control part 21 of a server show to a user also including a word attribute, when acquiring a word attribute. As a result, the user can automatically obtain broad related keyword information (without thinking by the user himself / herself) for the input search keyword.

  Further, it is desirable that the control unit 21 of the server presents the terminal so that the terminal can transmit the search request again using the related keyword without inputting the related keyword from the terminal. Specifically, the control unit 21 of the server (as the web server 9), for example, “The display portion of the related keyword can be selected by the user using the input unit 29 of the terminal, and the user can input the input unit 29. When the display portion of the related keyword is selected via the terminal, an HTTP (HyperText Transfer Protocol) response in which the terminal transmits a search request based on the related keyword may be transmitted to the terminal. As a result, the user can perform the search operation smoothly.

  Next, examples according to the embodiment of the present invention will be described with reference to FIGS. The present embodiment is an example in which a keyword assist function is used as an auxiliary function of an in-site search function in an information portal site.

  FIG. 17 is a diagram illustrating a system configuration of the embodiment. As shown in FIG. 17, in this embodiment, the site viewer terminal 11a and the Web server (information portal site) 9a are connected via the Internet 19a, and for example, a keyword assist function index is created in a firewall (not shown). Server 5a, keyword assist function search server 7a, and Web server 9a are connected.

  FIG. 18 is a diagram illustrating main data used in the embodiment. As shown in FIG. 18, in this embodiment, the word attribute appearance pattern information 101a and the word attribute information 102a shown in FIG. 18 are used.

  The site administrator infers the word appearance pattern from the content of the content (= document to be searched) in the information portal site, and uses the word attribute appearance pattern information 101a and the word attribute information 102a as the keyword assist function index creation server 5a. Register with. Data registration is performed by a dedicated GUI (Graphical User Interface) application from the site administrator's terminal, for example.

  In addition, the site administrator instructs the Web server 9a to import the search target document into the keyword assist function index creation server 5a. For example, the document is imported from a site administrator's terminal using a dedicated GUI application. Similarly, the site administrator instructs the server about the creation of the index and the distribution of the index from the terminal of the site manager by the dedicated GUI application.

  FIG. 19 is a diagram illustrating input of a search keyword. The Web server 9a receives the search character string input to the in-site search form of the information portal site, that is, receives a search request from the site viewer terminal 11a. Then, the Web server 9a performs morphological analysis and breaks it down into words. The example of the search keyword shown in FIG. 19 is “curry / famous store / Tokyo”, and is decomposed into “curry / famous / store / Tokyo”.

  FIG. 20 is a diagram illustrating retrieval of related keywords. The Web server 9a extracts only the words included in the word attribute information 102a registered by the site manager from the decomposed words, and searches for related words for each. As shown in FIG. 20, when first searching for “curry”, the related word search results are “place name: India”, “place name: Ginza”, “food: beef”, and “food: seafood”. Next, when searching for “Tokyo”, the related word search results are “place name: Shinagawa”, “place name: Ginza”, “action: tourism”, and “food: seafood”.

  FIG. 21 is a diagram illustrating presentation of related keywords. The Web server 9a extracts a common word from the related word search result and presents the related keyword to the site viewer. In the example shown in FIG. 21, since the common words are “Ginza” and “Seafood”, “+ place name: [curry Tokyo Ginza]” including the word attribute is presented as the first related keyword. In addition, “+ ingredients: [Curry Tokyo Seafood]” including the word attribute is presented as the second related keyword. As described above, in this embodiment, related keywords that are broad with respect to the input search keyword can be presented. Further, it is possible to present related keywords that can more appropriately narrow down the search results for the document group to be searched.

  The preferred embodiments of the document search system and the like according to the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited to such examples. It will be apparent to those skilled in the art that various changes or modifications can be conceived within the scope of the technical idea disclosed in the present application, and these naturally belong to the technical scope of the present invention. Understood.

The figure which shows schematic structure of the document search system 1 Computer hardware configuration diagram Server software configuration diagram The figure which shows the outline | summary of the work flow in the document search system 1 The figure which shows an example of the word attribute appearance pattern information 101 The figure which shows an example of the word attribute information 102 The figure which shows an example of the word conversion information 103 The figure which shows an example of the data item of the document information 104 The figure which shows the data flow in the index update object document discrimination | determination processing 71. The figure which shows the flow of the data in the morphological analysis process 72 The figure which shows the flow of the data in the word conversion process 73 The figure which shows the flow of the data in the keyword assist utilization word selection process 74 The figure which shows the flow of a process of the relevance calculation process 75 The figure which shows the flow of the data in the relevance calculation process 75 The figure which shows the flow of the data in the index creation process 76 The figure which shows the flow of a process regarding reception 65 of a search request. The figure which shows the system configuration | structure of an Example. Figure showing the main data used in the examples Diagram showing search keyword input Diagram showing related keyword search Diagram showing related keyword presentation

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ......... Document search system 3 ......... Administrator terminal 5 ......... Index creation server 7 ......... Search server 9 ......... Web server 11 ......... User terminal 13 ......... Network 21 ... ... Control unit 23 ......... Storage unit 25 ......... Media input / output unit 27 ......... Communication control unit 29 ......... Input unit 31 ......... Display unit 33 ......... Peripheral device I / F unit 35 ......... Bus 41 ... …… Related keyword index creation application 43 ……… Related keyword search application 51 ……… Word information registration 52 ……… Document registration 53 ……… Index creation 54 ……… Index distribution 55 ……… Related keyword search 56 ... …… Index creation history reference 61 ……… Definition of word information 62 ……… Document collection / update 63 ……… Indices 64 ......... Distribution of index 65 ......... Reception of search request 71 ......... Index update target document discrimination processing 72 ......... Morphological analysis processing 73 ......... Word conversion processing 74 ......... Words using keyword assist Sorting process 75 ......... Relevance calculation process 76 ... ... Index creation process 101 ... ... Word attribute appearance pattern information 102 ... ... Word attribute information 103 ... ... Word conversion information 104 ... ... Document information 105 ... ... Update document history information 106 ......... Index execution process history information 107 ......... Update target document information 108 ......... Morphological analysis result information 109 ......... Word conversion result information 110 ......... Relation degree calculation target word information 111 ... ... Word relevance information 112 ......... Index creation processing history information 113 ......... Over the word assist hash information 114 ......... keyword assisted search word information 115 ......... keywords assist relevant word information 116 ......... particular example 117 ......... index data word class classification according to the morphological analysis engine

Claims (16)

A document search system in which a terminal and a server are connected via a network, and the server presents a related keyword of a search keyword received from the terminal,
The server
When defining a word type as a word attribute, defining a combination of the word attributes where the word attribute appears within a certain range in the search target document as an appearance pattern, and calculating the word group relevance Means for holding word attribute appearance pattern information in which the weight for each appearance pattern is defined as the appearance pattern weight;
Relevance calculating means for searching for a word group that matches the appearance pattern in the search target document, and calculating a word group relevance by the distance of the searched word group and the appearance pattern weight;
Means for holding the word group relevance calculated by the relevance calculation means as word relevance information;
When the search keyword is received from the terminal, related keyword presenting means for presenting the related keyword of the search keyword with reference to the word relevance information;
A document search system comprising:   The word group relevance is a product of the product of the appearance pattern weight and the reciprocal of the distance of the word group for all previously registered words and all the appearance patterns. Item 2. The document search system according to Item 1.   2. The document search system according to claim 1, wherein the appearance pattern weights can be set for both cases where the appearance order of the registered word groups all match and other cases.   The document search system according to claim 1, wherein the word attribute appearance pattern information is updateable.   The document search system according to claim 1, wherein the related keyword presenting unit presents the related keyword together with a search result based on the search keyword and / or a word attribute related to the related keyword. A server connected to a terminal via a network and presenting a related keyword of a search keyword received from the terminal,
When defining a word type as a word attribute, defining a combination of the word attributes where the word attribute appears within a certain range in the search target document as an appearance pattern, and calculating the word group relevance Means for holding word attribute appearance pattern information in which the weight for each appearance pattern is defined as the appearance pattern weight;
Relevance calculating means for searching for a word group that matches the appearance pattern in the search target document, and calculating a word group relevance by the distance of the searched word group and the appearance pattern weight;
Means for holding the word group relevance calculated by the relevance calculation means as word relevance information;
When the search keyword is received from the terminal, related keyword presenting means for presenting the related keyword of the search keyword with reference to the word relevance information;
A server comprising:   The word group relevance is a product of the product of the appearance pattern weight and the reciprocal of the distance of the word group for all previously registered words and all the appearance patterns. Item 7. The server according to item 6.   The server according to claim 6, wherein the appearance pattern weight can be set to a value for both cases where the appearance order of the registered word groups all match and other cases.   The server according to claim 6, wherein the word attribute appearance pattern information is updateable.   The server according to claim 6, wherein the related keyword presenting unit presents the related keyword together with a search result based on the search keyword and / or a word attribute related to the related keyword. A terminal and a server are connected via a network, the server defines a word type as a word attribute, and the word attribute appears in the search target document within a certain close range. It is defined as a pattern, holds word attribute appearance pattern information in which the weight for each appearance pattern when calculating the word group relevance is defined as the appearance pattern weight, and the related keyword of the search keyword received by the server from the terminal A document search method to be presented,
The server searches for a word group that matches the appearance pattern in the search target document, and calculates a word group relevance by the distance of the searched word group and the appearance pattern weight;
The server holds the word group relevance calculated by the step of calculating the word group relevance as word relevance information;
The terminal transmitting the search keyword to the server;
The server refers to the word relevance information and presents the received related keyword of the search keyword;
A document search method characterized by including:   The word group relevance is a product of the product of the appearance pattern weight and the reciprocal of the distance of the word group for all previously registered words and all the appearance patterns. Item 12. The document search method according to Item 11.   The document search method according to claim 11, wherein the appearance pattern weights can be set in both cases where the appearance order of the registered word groups is identical and in other cases.   The document search method according to claim 11, wherein the word attribute appearance pattern information is updateable.   12. The document search method according to claim 11, wherein the step of presenting the related keyword presents the related keyword together with a search result by the search keyword and / or a word attribute related to the related keyword.   A program that causes a computer to function as the server according to any one of claims 6 to 10.

Images