JP2008112310A - Retrieval device, information retrieval system, retrieval method, retrieval program and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a retrieval device, an information retrieval system, a retrieval method, a retrieval program and a recording medium, suitable to retrieve a document group wherein documents especially mutually have reference relation when automatically detecting a candidate of a document desired by a user from a lot of documents present in a document management system or the Internet. <P>SOLUTION: The document fit for a given retrieval condition is retrieved, adaptability of the retrieved document and the retrieval condition is calculated, calculation condition information showing a characteristic of the document included in the retrieved documents is extracted, the adaptability calculated about each document is weighted based on the calculation condition information to calculate a score, and the document group retrieved based on the calculated score is aligned and is provided to the user. Thereby, the high-reliability document fit for the input retrieval condition can be provided to the user. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a search device, an information search system, a search method, a search program, and a recording medium, and in particular, automatically detects candidate documents desired by a user from a large number of documents existing on the Internet or a document management system. In particular, the present invention relates to a search apparatus, an information search system, a search method, a search program, and a recording medium suitable for searching a document group in which documents have a reference relationship.

  When searching for content that matches a URL or query specified by the user, there is a method of detecting a conforming document by detecting several reference source documents and searching for documents linked to them. .

  In Patent Document 1, in order to collect more necessary WWW pages in a given time or storage area, when the search unit receives a search instruction from the input unit, the search starting URL is retrieved from the search information storage unit. Acquire, analyze the corresponding HTML text via the communication unit, store the link destination URL, anchor character string, etc. in the hyperlink storage unit, and store the keyword and character string, etc. stored in the search information storage unit And the evaluation value of the degree of search purpose match of the data specified by the link destination URL is calculated, and the link destination URL including “link” or the like in the character string or the like is given priority, and the link destination of the highest evaluation value is calculated. The URL is extracted, the HTML text of the extracted URL is acquired, the search is repeated until a predetermined end condition, and the display information creation unit refers to the search history and stores the text stored in the hyperlink storage unit Reordering columns in predetermined selection criteria, the information retrieval agent device that displays a list has been proposed on the display unit.

  In Patent Document 2, for example, a query such as “I want to acquire a homepage that includes one or more images, uses the keywords“ Kyoto ”and“ sightseeing ”and is described in“ Japanese ”,” There has been proposed an information search apparatus, an information search method, and a recording medium that can acquire only necessary content among content connected from a predetermined start page by giving it to a search robot.

  In Patent Document 3, the input of a search term is received by an input receiving unit of a search device, a storage unit stores a document, terms included therein, and a reference relationship between documents, and the search unit refers to the storage unit. Search for a document including a search term that has received an input, and further search for a document group including a plurality of documents in a reference relationship, each of which includes the search term, The calculation unit calculates a score for the document and the document group from the frequency at which the search term appears in the searched document and the frequency at which the search term appears in each of the documents included in the searched document group The department arranges the searched documents and document groups according to the calculated scores and outputs them as search results. In the score calculation, the type of reference relationship (“citation”, “description”, etc.) is used. The Rukoto, search apparatus can properly search for documents, such as documents with each other can have a reference relationship, the search method, and program are proposed.

In Patent Document 4, information on a plurality of documents that may be updated is collected, and a document score for assigning a score for each document to be used for extracting a specific document from the collected document group In the calculation method, attribute information representing the characteristics is extracted from each document, attribute information extracted at two or more time points is simultaneously held for each document, and based on a plurality of attribute information held for each document. Calculates the time-dependent change of attribute information, reflects the obtained time-dependent change of attribute information in the calculation of points, and also shows the relationship between the primary attribute included in the target document as the attribute information, the target document, and other documents. Document score calculation method, apparatus, and program capable of obtaining a result closer to a document required by a user when searching a large amount of documents by extracting next attribute information It has been.
JP 2000-90111 A JP 2002-07431 A JP 2004-272626 A JP 2004-227165 A

  However, the above invention has the following problems.

  A search technique is used that extracts documents that meet specified search conditions from documents stored in the Internet or a document management database, calculates a score as appropriate, and presents the documents in the order of the scores.

  In the score calculation, attributes (keywords, document update date, number of links, etc.) indicating the characteristics of each document are used. In particular, a link indicating a reference relationship between documents is considered to be an index indicating the reliability of the document, and a score calculation method using the number of links and the number of linked links has been devised.

  These attribute values are information given by the document creator. On the other hand, there is information given by a third party who is not the document creator.

  For example, comments and trackbacks, which are functions of blogs that have been popular in recent years (web logs, Web logs, abbreviated as “blogs” hereinafter), are one of them. A blog generally means a type of Web site that is updated daily, and has recently been known as an information providing service on the Internet.

  For example, an article such as news or a web page saved by the founder of the blog (hereinafter referred to as “blog founder”) and published on the Internet may be a viewer of the article (hereinafter referred to as “blog viewing”). A web site in which comments such as personal opinions, criticisms and explanations of the blog viewer are added, and the added comments are stored in time series.

  This blog is different from a bulletin board for the purpose of exchanging opinions among a plurality of persons in that the main purpose is for an information sender to express an opinion. In the blog, for each article saved (written) by the blog establishment person, a blog viewer can input a comment, and has a function of adding the comment to the article. Blog creators and viewers can see comments added to an article. In addition, the blog system provides a function for a blog creator to input, save, and publish his / her article by quoting an article of another blog creator.

  On the other hand, a link to the article of the blog creator who cited the article is added to the article of the other person who established the blog. In the past, the HTML (Hyper Text Markup Language) format document used for publishing a document on a website was used to set a link referring to another HTML format document. Will be linked to each other's article and the cited article. This technique is hereinafter referred to as “Track Back”. By using the trackback, a blog creator and a viewer can check whether or not a certain article is quoted by another blog creator, and if so, the link of the article to which it is cited.

  Examples of the blog system include Movable Type (URL: http://www.movabletype.org/) and Coco Log (URL: http://www.cocolog-nifty.com/). The technical specifications of Movable Type trackbacks are: Movable Type Users Manual: TRACKBACK (URL: http://www.movabletype.org/docs/mttrackback.html), TRACKBACK Modules for RSS 1.0 / 2.0 (URL: http: / /madskills.com/public/xml/rss/module/trackback/).

  Therefore, the present invention searches for a document that matches a given search condition, calculates the degree of matching between the searched document and the search condition, and calculates calculation condition information indicating the characteristics of the document included in the searched document. Retrieval apparatus and information for extracting and weighting the fitness calculated for each document based on calculation condition information to calculate a score, and arranging a group of documents searched based on the calculated score and providing them to the user The object is to propose a search system, a search method, a search program, and a recording medium.

  According to the first aspect of the present invention, there is provided a presentation unit that searches and presents a document group that matches an input search condition, an input unit that receives the search condition, and a document that includes the search condition received by the input unit. In a search device having a search means for searching, the degree of match calculation means for calculating the degree of match between the document searched by the search means and the search condition, and the characteristics of the document searched by the search means Extraction means for extracting calculation condition information, and a score for calculating a score by weighting the fitness of each document calculated by the fitness calculation means based on the calculation condition information extracted by the extraction means A calculation unit; an alignment unit that aligns the document group based on the score calculated by the score calculation unit; Characterized by presenting the documents that have been aligned by serial alignment means to the user.

  According to a second aspect of the present invention, in the search device according to the first aspect, when the search condition received by the input unit is a reference relationship, the matching degree calculation unit is configured to display the reference source document and the reference destination document. The degree of fit with is calculated.

  According to a third aspect of the present invention, in the search device according to the second aspect, when the fitness calculated by the fitness calculation means is lower than a preset threshold, the score calculation means excludes the reference relationship. A score is calculated based on the calculation condition information.

  According to a fourth aspect of the present invention, in the search device according to the second or third aspect, the score calculation means includes the number of the search conditions included in each of the reference source document and the reference destination document, and the document The score is weighted based on the average of the numbers including the search condition.

  According to a fifth aspect of the present invention, there is provided an information retrieval system in which the retrieval device according to any one of the first to sixth aspects, a document database, and at least one client terminal requesting retrieval of a document are connected via a network. It is characterized by that.

  According to the sixth aspect of the present invention, there is provided a presentation unit that searches and presents a document group that matches an input search condition, an input unit that receives the search condition, and a document that includes the search condition received by the input unit. In a search method of a search device having a search means for searching, a fitness calculation step for calculating a fitness between the document searched by the search means and the search condition, and a search for the document searched by the search means An extraction step of extracting calculation condition information indicating characteristics, and a score by weighting the fitness of each document calculated by the fitness calculation step based on the calculation condition information extracted by the extraction step A score calculating step for calculating, an aligning step for aligning the document group based on the score calculated by the score calculating step, and the presenting means Characterized by a step of presenting the documents that have been aligned by the alignment process to the user.

  According to a seventh aspect of the present invention, in the search method according to the sixth aspect, when the search condition received by the input means is a reference relationship, the fitness calculation step includes a reference source document and a reference destination document. The degree of fit with is calculated.

  According to an eighth aspect of the present invention, in the search method according to the seventh aspect, if the fitness calculated by the fitness calculation step is lower than a preset threshold value, the score calculation step excludes the reference relationship. A score is calculated based on the calculation condition information.

  The invention according to claim 9 is the search method according to claim 7 or 8, wherein the score calculation step includes the number of the search conditions included in each of the reference source document and the reference destination document, and The score is weighted based on the average of the numbers including the search condition.

  The invention according to claim 10 is a presentation unit that searches and presents a document group that matches the input search condition, an input unit that receives the search condition, and a document that includes the search condition received by the input unit. In a search program of a search device having a search means for searching, a fitness calculation process for calculating a fitness between the document searched by the search means and the search condition, and a search for the document searched by the search means Based on the extraction process for extracting calculation condition information indicating characteristics, and the calculation condition information extracted by the extraction process, a score is obtained by weighting the fitness of each document calculated by the fitness calculation process A score calculation process to calculate; an alignment process to align the document group based on the score calculated by the score calculation process; Display means, characterized by having a process of presenting the documents that have been aligned by the alignment process to the user.

  According to an eleventh aspect of the present invention, in the search program according to the tenth aspect, when the search condition received by the input means is a reference relationship, the fitness calculation processing is performed by referring to a reference source document and a reference destination document. The degree of fit with is calculated.

  According to a twelfth aspect of the invention, in the search program according to the eleventh aspect, when the fitness calculated by the fitness calculation processing is lower than a preset threshold, the score calculation processing excludes the reference relationship A score is calculated based on the calculation condition information.

  According to a thirteenth aspect of the present invention, in the search program according to the eleventh or twelfth aspect, the score calculation processing includes the number of the search conditions included in each of the reference source document and the reference destination document, and The score is weighted based on the average of the numbers including the search condition.

  A recording medium according to a fourteenth aspect is characterized in that the search program according to any one of the tenth to thirteenth aspects is recorded.

  The present invention searches for documents that match a given search condition, calculates the degree of matching between the searched document and the search condition, and extracts calculation condition information indicating the characteristics of the document included in the searched document. The score is calculated by weighting the fitness calculated for each document based on the calculation condition information, and the group of documents retrieved based on the calculated score is arranged and provided to the user. A highly reliable document that meets the search conditions can be provided to the user.

  Hereinafter, the configuration and operation of a search device according to an embodiment of the present invention will be described.

  First, an outline of a network of an information search system having a search device according to the present embodiment will be described with reference to FIG.

  The information retrieval system includes client terminals 110 and 120 that request information retrieval, a retrieval device 130 that is a retrieval server, a database 140 that manages documents to be retrieved, and a network 100 that connects them.

  Next, a schematic configuration of a typical information processing apparatus that can function as the search apparatus 130 according to the present embodiment will be described with reference to FIG.

  The information processing apparatus 101 is controlled by a CPU (Central Processing Unit). When the information processing apparatus 101 is turned on, the CPU 102 executes an IPL (Initial Program Loader) stored in the ROM 103.

  The IPL reads and executes an OS (Operating System) program stored in a recording medium such as the hard disk 104, the FD attached to the FD drive 110, and the CD-ROM attached to the CD-ROM drive 111. It is a program.

  After the OS is started, the CPU 102 executes an application stored in the hard disk 104 or the like in accordance with a user instruction input from the keyboard 105 or the mouse 106 or in accordance with the contents of a setting file previously described in the hard disk 104 or the like. Run the program.

  By executing the application program, the information processing apparatus 101 functions as the search apparatus 130. The application program is recorded on a storage medium such as the hard disk 104, an FD mounted on the FD drive 110, and a CD-ROM mounted on the CD-ROM drive 111. The CPU 102 uses the RAM 107 as a temporary work storage area when executing these programs. In addition, a register or cache provided in the CPU 102 is used as a temporary work storage area.

  As the program is executed, the result is reported to the user and the progress is displayed, so that the CPU 102 can display the information on the display device 108 such as a liquid crystal display or a CRT (Cathode Ray Tube). In the instruction operation using the mouse 106, the cursor displayed on the screen is moved by moving the mouse 106, and the menu item pointed to by the cursor can be selected by clicking the mouse 106.

  The search device 130 can communicate with a computer communication network such as the Internet via an interface 109 such as a NIC (Network Interface Card) or a modem. The document data received via the interface 109 can be processed, the processing result can be transmitted via the interface 109, the program received via the interface 109 can be executed, and the like.

  FIG. 3 is a schematic diagram showing a basic configuration of the search device according to the present embodiment. FIG. 4 is a flowchart showing a flow of processing executed by the search device according to the present embodiment. Note that the search device according to the present embodiment includes an input unit 210, a search unit 220, a calculation unit 230, a storage unit 240, and an alignment unit 250.

  First, the input unit 210 receives an input of search conditions (step S401). Search conditions include keywords, file size and number of characters for each document, number of nouns, language, reference relationship (link, linked, trackback) and table, comments, advertisements, images, presence / number of music, creation date and time for each document This is information indicating document characteristics given by a third party, such as information given by the period, URL, author, and blog establishment. Various techniques such as the following can be applied to the method of inputting search conditions.

(A) Inputting a character string or a symbol as a search condition.
For example, a word is input as a character string, and the input word is set as a search condition. Alternatively, a natural sentence is input, a morphological analysis is performed to obtain a word, and this is used as a search condition. Furthermore, normalization of words, different notation, and development of synonyms may be used as search conditions.
(B) Selection of search conditions from a list or the like presented in the search I / F.

  Therefore, the keyboard 105 and the mouse 106 function as the input unit 210 in cooperation with the CPU 102 and the display device 108.

  The search unit 220 refers to the storage unit 240 and searches for a document that meets the search condition for which the input has been accepted (step S402). Further, information (hereinafter referred to as calculation condition information) such as comments and reference relationships (link, linked, trackback) adopted in each document by the document creator is extracted from the extracted conforming documents (step S403). ). Therefore, the CPU 102 functions as the search unit 220 in cooperation with the hard disk 104.

  Then, the calculation unit 230 calculates the degree of matching between each document and the search condition from the number and frequency of search conditions included in the searched document, and further calculates the number and frequency of calculation condition information included in the searched document. Are used to calculate the score of each document. Note that a vector dot product method or the like is used to calculate the fitness based on the retrieved document and the retrieval condition. The weighting based on the retrieved document and the calculation condition information is performed by, for example, a method of making the weight of the document heavier than the weight of the other document when the trackback number of the certain document is larger than that of the other document. adopt.

  Then, the total score of each weighted document is calculated to calculate a score for the document group (step S404). In the present embodiment, the search condition included in each document, the reference relationship between each document, and the frequency of the search condition included in each document are stored in the storage unit 240.

Here, an example of a method for calculating the score of each document will be described. In general, in document search, the appropriateness of a keyword input as a search condition by a user and each document included in a search target document group is calculated as a numerical value called a score of each document for the search. In the present embodiment, a method for calculating the fitness using TF (Term Frequency) and IDF (Inverse Document Frequency) will be described. TF is the number of times a keyword appears in the document. The IDF is calculated by log (total number of documents / number of documents including the corresponding word) +1. The goodness of fit is calculated by the product of TF and IDF (TF * IDF), and the higher the value, the higher the possibility of selecting a specific document from the document group. Here, TF * IDF is adopted as the weight of the document vector component. The score So of each document di is calculated based on Equation 1. In Equation 1, kj is a keyword included in the search condition, and j = 1, 2, 3,.
So (di) = ΣTF (kj, di) * IDF (kj) (Equation 1)
When the score So of each document is calculated, the score Su calculated by weighting the fitness of the document is calculated based on the formula 2. In Equation 2, rl is a weight calculated from the calculation condition information, and l = 1, 2, 3,.
Su (di) = So (di) * rl (di) * (Equation 2)

  Further, the calculation information used by the calculation unit 230 for weighting has a reference relationship. The reference relationship includes the link that the document creator gives to the document, the trackback that the third party gives to the document, the linked link that gives the third party the link to the document, etc. Information (not limited to this, other reference relationships may be used). When the reference relationship is used as the calculation information information, the calculation unit 230 determines the weight for the fitness using the cross-citation by the reference relationship, the number of reference relationships, the reference frequency, and the assigned information interval.

Specifically, the weight rl (di) when the reference relationship (trackback number or the like) is used as the calculation condition information is calculated based on Equation 3. In Expression 3, dt is the number of trackbacks of the document di, and Nt is the average of the number of trackbacks in the search target document group.
rl (di) = dt / Nt (Formula 3)
In this case, the score Su for the document group is calculated based on Equation 4.
Su (di) = So (di) * dt / Nt (Formula 4)

  Finally, the aligning unit 250 arranges the retrieved documents and document groups according to the calculated score, and outputs them as a retrieval result (step S405). Accordingly, the CPU 102 functions as the alignment unit 250 in cooperation with the display device (when displaying the search result) 108 and the hard disk (when storing the search result) 104.

  As described above, when only one input of the search condition is accepted, all the documents included in the searched document group include the search condition. On the other hand, when receiving a plurality of input search conditions, the document searched in step S402 and the document having a reference relationship with the document include at least one of the plurality of search conditions. May be calculated.

  In addition, when a keyword is entered as a search condition, the similarity between the search condition and the document is determined by using Vector Space Model as a document search method as well as those that include the search condition specified from multiple documents. Is calculated by the inner product or the like, thereby detecting a highly compatible document. When the Vector Space Model is used as a document search method, the similarity between the search condition and the document is calculated by an inner product or the like. In this case, even a document with a high degree of similarity may not include any search condition, and a document with “high similarity to the search condition” is detected as a matching document.

  Further, since the calculation condition information is information given from a third party, there is a possibility that information that has nothing to do with the original document to which the information is given is described. If such information is used for calculation of compatibility with a search condition designated by the user, an adverse effect occurs. Therefore, the relevance is calculated using only necessary information so that information unrelated to the original document is not used as calculation condition information.

  For example, a blog may be tracked back to show a product description or advertisement such as “trackback spam”. Here, trackback spam is an example of information that has nothing to do with the original document. Therefore, when bibliographic information or a document such as a trackback destination URL or title is used as a search condition, the calculation unit 230 first calculates the degree of conformity with the trackback source document. As a result, when the document is completely unrelated to the content of the reference source (when the fitness level is lower than a certain threshold value), the trackback information should be excluded from the calculation condition information when weighting the fitness level. Thus, appropriate weighting is executed. For the calculation method of conformity of two documents, vector space method (Salton, G .: The Vector Space Model, Automatic Text Processing, Addison-Wesley Publishing, pp.312-325 (1989). Are defined, a vector space spanned by those phrases is defined, and a vector of representative phrases for each document is formed, and the similarity between the documents is calculated by calculating the inner product of the vectors. It can be calculated by calculating). In other words, morphological analysis is first performed on the trackback source document, keywords are extracted, the extracted keywords are regarded as search condition keywords in the above-described score calculation (TF * IDF), and the trackback spam destination document is further regarded as di. Perform the degree calculation. When the calculated fitness is low, no weighting is performed on the fitness. Alternatively, a list of information that is not added to the calculation condition information used for weighting may be created in advance, and the weighting for the degree of matching of each document may be executed by referring to the list. For example, when a trackback is targeted, a list of URLs not to be added to the calculation condition information is created in advance, and when a URL described in the list is trackbacked, when weighting is performed in consideration of the number of trackbacks. The URL is calculated without counting the trackback number.

  Here, a method for calculating the degree of matching between the trackback destination document and the trackback source document will be described. First, the number of information additions, the information addition frequency, and the information addition interval in each document are extracted from the addition information used as a search condition in document search. For example, when a comment is used as the grant information, a field in which a comment is described is extracted from each document, and the grant interval based on the number of comments and the comment update date and time is calculated as the fitness. In addition, for the information grant frequency, for example, the number of comments assigned to all blogs registered in advance in the blog portal site is extracted, and the degree of grant is calculated by comparing with the average of the number of comments given. Degree. If each of these matching levels is higher than the threshold value, the weight based on the calculation condition information is increased, and the document is detected as a document that matches the trackback source document.

  For example, the information application frequency is the number of times (frequency) that a certain trackback (calculation condition information) is applied in a certain period. In this case, the weight is calculated by rl (di) = dc / Nc. Note that dc is the number of trackbacks for one month before and after the date searched by the user. Nc is the average number of trackbacks per month in the search target document group. In this case, the score of the document di is calculated by Su (di) = So (di) * dc / Nc.

  By using the information assignment frequency for the calculation condition information, even if the number of trackbacks in the long term of one year is less than that of other documents, there are documents that are larger than other documents in a certain short period of time. It is considered a topic that has been attracting attention, and it is possible to present search results that take this viewpoint into consideration.

  The information grant interval is the average interval tracked back from the time the document was registered to the point of search or between the search time and the month before and after the search for a certain period (from the date of the first trackback in the fixed period to the end) Trackback period (number of days, time, etc./number of trackbacks in a certain period). The calculation unit 230 calculates an information addition interval, and calculates a score by weighting the quotient with the average trackback interval in the search target document group.

  The calculation condition information used for calculating the weight is arbitrarily selected by the user, the weight is calculated using the selected calculation information, and the score is calculated by weighting the fitness. Is also possible.

  Various information such as keywords, trackbacks, comments, etc. are used as search condition and calculation condition information, but it is also possible to calculate the suitability of each document for each specified type or type and display it as search results is there. Specifically, the score is calculated by calculation condition information used at the time of score calculation on the search device side in advance and a combination thereof, and each result is displayed as a search result. The search result for each calculation condition information is a search result in which a score is calculated using only the trackback number as calculation condition information or a score is calculated using only the number of linked links. The score calculation based on the combination is a search result obtained by calculating the cost using two pieces of information of the track back number and the number of linked links as calculation condition information.

It is the schematic which shows the structure of the information search system containing the search device which concerns on this embodiment. It is the schematic which shows the structure of the typical information processing apparatus with which the search device which concerns on this embodiment is implement | achieved. It is a block diagram which shows the outline of the search device which concerns on this embodiment. It is a flowchart which shows operation | movement of the search device which concerns on this embodiment.

Explanation of symbols

101 Information processing apparatus 102 CPU
103 ROM
104 Hard disk 105 Keyboard 106 Mouse 107 RAM
108 display device 109 interface 110 FDD
111 CDD
210 Input unit 220 Search unit 230 Calculation unit 240 Storage unit 250 Alignment unit

Claims (14)

A search unit including a search unit that searches and presents a document group that matches an input search condition, an input unit that receives the search condition, and a search unit that searches for a document including the search condition received by the input unit. In the device
Goodness-of-fit calculation means for calculating goodness-of-fit between the document searched by the search means and the search condition;
Extraction means for extracting calculation condition information indicating characteristics of the document searched by the search means;
Score calculating means for calculating a score by weighting the fitness of each document calculated by the fitness calculation means based on the calculation condition information extracted by the extraction means;
Alignment means for aligning the document group based on the score calculated by the score calculation means,
The presenting means presents the document group aligned by the aligning means to a user.   2. The matching level calculation unit calculates a matching level between a reference source document and a reference destination document when the search condition received by the input unit is a reference relationship. Search device.   The score calculation means calculates a score based on the calculation condition information excluding the reference relationship when the fitness calculated by the fitness calculation means is lower than a preset threshold value. Item 3. The search device according to Item 2.   The score calculation means weights the score based on the number of the search conditions included in each of the reference source document and the reference destination document and an average of the number of the document including the search conditions. The search device according to claim 2 or 3.   7. An information search system comprising: the search device according to claim 1; a document database; and at least one client terminal that requests a document search via a network. A search unit including a search unit that searches and presents a document group that matches an input search condition, an input unit that receives the search condition, and a search unit that searches for a document including the search condition received by the input unit. In the device search method,
A fitness calculation step of calculating a fitness between the document searched by the search means and the search condition;
An extraction step of extracting calculation condition information indicating characteristics of the document searched by the search means;
Based on the calculation condition information extracted by the extraction step, a score calculation step of calculating a score by weighting the fitness of each document calculated by the fitness calculation step;
An alignment step of aligning the document group based on the score calculated by the score calculation step;
The presenting means includes a step of presenting to the user the group of documents arranged in the alignment step.   7. The matching level calculation step calculates a matching level between a reference source document and a reference destination document when the search condition received by the input unit is a reference relationship. retrieval method.   The score calculation step calculates a score based on the calculation condition information excluding the reference relationship when the fitness calculated by the fitness calculation step is lower than a preset threshold value. Item 8. The search method according to Item 7.   The score calculating step weights the score based on the number of the search conditions included in each of the reference source document and the reference destination document and an average of the number of the documents including the search condition. The search method according to claim 7 or 8. A search unit including a search unit that searches and presents a document group that matches an input search condition, an input unit that receives the search condition, and a search unit that searches for a document including the search condition received by the input unit. In the device search program,
A fitness calculation process for calculating a fitness between the document searched by the search means and the search condition;
An extraction process for extracting calculation condition information indicating characteristics of the document searched by the search means;
Based on the calculation condition information extracted by the extraction process, a score calculation process for calculating a score by weighting the fitness of each document calculated by the fitness calculation process;
An alignment process for aligning the document group based on the score calculated by the score calculation process;
The presenting means includes a process of presenting the document group aligned by the alignment process to a user.   11. The matching level calculation process calculates a matching level between a reference source document and a reference destination document when the search condition received by the input unit is a reference relationship. Search program.   The score calculation process calculates a score based on the calculation condition information excluding the reference relationship when the fitness calculated by the fitness calculation process is lower than a preset threshold value. Item 12. The search program according to Item 11.   The score calculation process is characterized in that the score is weighted based on the number of the search conditions included in each of the reference source document and the reference destination document and an average of the number of the document including the search conditions. The search program according to claim 11 or 12.   The computer-readable recording medium which recorded the search program of any one of Claim 10 to 13.

Images