JP2008310811A - Information retrieval system, information retrieval method and information retrieval server - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information transmitting technique capable of outputting not a perfectly matched search output by a conventional search engine but a search result reflecting a searching user's search intention to the searching user, and enabling an information providing user to easily transmit information in a manner such that the transmitter thereof can be easily searched by an intended searching user by the transmission of information, whereby communications between the users can be smoothed. <P>SOLUTION: To a server accumulating a registration matrix including a category, a theme and a keyword, a search matrix of the same data structure as the registration matrix is transmitted from a user terminal connected thereto through a network, and collated with the registration matrix, and a search result is output based on the similarity between both the matrixes. According to this, the searching user can obtain an intended search result. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention is a technology effective when applied to information retrieval via a general-purpose network.

  As a search system on the Internet, which is a general-purpose network, a search server system called a search engine such as google.com is known.

In such a search engine, the display order of the list of information displayed as a search result is the number of links from other sites to the address (URL: Uniform Resorce Locator) where the information is stored, and the information is searched. It is said that it is determined comprehensively based on factors such as the frequency at which information is updated and the frequency at which information is updated, but most of these factors are not related to the degree of match with the search conditions expected by the search user. In many cases.

  There is no problem when a search user as an information receiver requests information (a popular site) that is frequently viewed, but information that meets certain special conditions is ranked higher even if it is not popular. If you want to display, you can't get enough satisfaction with the technology so far.

  In addition, as to whether or not the search target is displayed as a result of the search, when a simple keyword enumeration is performed, only the keywords that match all the keywords entered by the search user are displayed. It is.

  For example, when you enter three keywords, it is common that hundreds of information hits, but as soon as you increase one keyword to four, the hit information becomes zero. It is.

  However, as a result of adding one keyword, there is a high possibility that the information that the search user most demands is actually present in the information that is inconsistent and is out of the search result.

  Here, when specifying the search condition, it is possible to specify a search condition that is somewhat complicated by combining the keywords with an AND condition (logical product) or an OR condition (logical sum). However, in order to accurately describe these compound conditions, it is necessary to have a high level of logical ability to assemble a logical expression, and it is necessary to be proficient in searching even to an expert area called a searcher.

  In the first place, the information required by general search users is often not clear at the time of searching which keyword is combined and how to get to the information that is most expected. Even if you search by specifying a compound condition with an expression, the search result you originally wanted is not necessarily limited to exactly exactly the decryption condition, but it does not match some conditions. However, there should have been something that you do not want to exclude from the display.

  On the other hand, the characteristics of the information transmission side to be searched by the search engine must also be considered.

  In other words, with the spread of blogs and social networking services (SNS) in recent years, such as writing on electronic bulletin boards, the number of network users who send out information has increased compared to before. From the user's point of view, it's not that big.

  There may be an idea that network users do not send information themselves because they do not feel such a necessity, but it is difficult to summarize the information so that it can be disclosed to others by changing the way of view. It may be troublesome.

  Previously, if you wanted to send information on the Internet, you had to create a homepage and have your own site, but with the advent of blogs and SNSs, it is not always necessary for you to do the screen layout of your homepage. lost.

  However, apart from illustration and video sites, “text” is still an essential element for information transmission. Unless you are a person who has no resistance to writing, or a person who is confident that you can write a sentence that can be shown to others, it can be said that the reality is that it will be difficult to send information on the Internet.

Japanese Patent Application Laid-Open No. 2007-11438 (Patent Document 1) discloses a technique for performing a narrow search for narrowing down candidate characters every time a character to be searched from the beginning of a keyword is input. For the data that stores the keyword character string, the index is simply configured to have the corresponding number of data and the next candidate character information following the character for each character of the keyword. It was not a technique devised in consideration of search results.
JP 2007-11438 A

  The present invention has been made in view of the above-mentioned points, and for Internet users who tend to increase more and more, search users are not completely matched search outputs by conventional search engines, Provide a technology that can output search results that reflect the search user's search intentions. Information can be easily transmitted to information providing users, and the information can be easily searched by the search user intended by the caller. Therefore, it is a technical problem to provide an information transmission technology that facilitates communication between users.

  In order to solve the above problems, the present invention employs the following means.

  The present invention comprises a server for storing a registration matrix consisting of a combination of data, a network, and a search user terminal connected via the network, and a registration matrix consisting of a combination of data stored in the server. , Transmitting a search matrix composed of a combination of data input from user terminals connected via a network, and comparing the search matrix input in the above and the registration matrix in the server according to the similarity between the matrices An information search system that outputs search results and transmits the search results to the search user terminal.

  The information search system can also be realized as an information search server or an information search method in the information search server.

  Here, the registration matrix and the search matrix are composed of a genre, a theme, and two or more keywords related to the theme.

  However, it is not necessary to be limited to these three data structures. For example, a sub theme may be added under the theme, and a keyword may be placed under the sub theme.

  Further, in the matching between the registration matrix and the search matrix, themes having a large number of matching keywords are evaluated as having high matching points, and are judged to have high similarity.

  Further, the number of coincidence may be set as n, and the nth power of a predetermined parameter (number of elements in a matrix term) may be calculated as a coincidence point.

  Here, when considering the essence of the act of searching for information, it should be noted that the searching user does not have complete knowledge about the information to be searched at the time of searching. This is because, if you already have complete knowledge, there is no need to search for such information. In that sense, the knowledge at the time when the search user is searching and the content of the requested information are essentially impossible.

  The act of information search is to search for information closest to the request at that time based on the knowledge that the search user has at that time. The content of the information that is sought is not an equal (equal) relationship but an approximate (approximate) relationship. Nevertheless, in the conventional search technology, a method of narrowing down by matching logical conditions (an alternative of true or false in the end) has been adopted, so the information that the search user expects most is There was a limit to the function of searching. That is, the “match point” has a function as an index for measuring the rarity from the viewpoint of similarity between the “registration matrix” and the “search matrix”.

  As described above, the information search in the present invention is different from the conventional technique in that the knowledge at the time of search and the information to be searched are “similar” by the method of matching the “registration matrix” and the “search matrix”. It is a method that enables the highest priority, and even if registered information is viewed less frequently, it is easy to do so if there is a search user who happens to seek similar information. It is a method that makes it possible to find out.

  Since the search results are displayed in order of high similarity, this search method is referred to as an “OLL (Order that Looks Like) search method” in the following description.

  In this OLL search method, a search engine for generating a matrix may be an existing search engine (server search program) used in a general-purpose search site.

  According to the present invention, by realizing a so-called OLL search method that collates a “registration matrix” with a “search matrix”, the priority is given to “similar” between the knowledge at the time of search and the information that is sought. Search can be performed, and even if registered information is browsed less frequently, if there is a search user who happens to request similar information, it can be easily found.

  This embodiment is a member registration system information registration / retrieval system on the Internet.

  The configuration of the information registration / retrieval system according to the present embodiment will be described with reference to FIGS. 14 and 15.

  As shown in the figure, this system includes a server (SV), a network (NW), and user terminals (UT1, UT2) connected to the server (NW) via the network (NW). ing.

  The server (SV) is composed of a general-purpose network-connectable computer. Here, the server (SV) is described as having the functions of a Web server, an application server, and a database server. However, a server is separately provided for each of these functions. Needless to say, it may be constructed.

  As shown in FIG. 15, the server (SV) is connected to a central processing unit (CPU) and a main memory (MM) via a bus (BUS), and a hard disk device (HD) as a large-scale storage device, A keyboard (KBD) as an input device and a display device (DISP) as an output device are provided.

  The bus (BUS) is connected to the network (NW).

  In the hard disk device (HD), together with operating software (OS), a Web server program (to be described later), a user database (DB), an information registration program for registering this user database, and registered information from the user terminal Information search programs and the like (collectively “programs”) for searching by request are registered.

  The various programs are sequentially loaded into a central processing unit (CPU) via a bus (BUS) and a main memory (MM), thereby executing a server function.

  The user terminal connected via the network (NW) has almost the same hardware configuration, and thus the description thereof will be omitted. However, the hard disk device (HD) of the user terminal is used to access the server. A browser program or a predetermined program downloaded from the server is registered. As a user terminal, in addition to a general-purpose personal computer, a mobile phone (MP) that can access the network (NW) via a communicable PDA or a base station (BS) connected to the network (NW). Such mobile terminal devices can also be used.

  As described above, the server (SV) includes, for example, a Web server program and a database, and functions as a Web server and a database server. The Web server program stored in the hard disk device (HD) is for receiving a message by an HTTP request transmitted from a user terminal and performing processing corresponding to the message. Specifically, when the server determines that the message from the user terminal is a message for accessing the Web page specified by the URL, the server provides the source file and content data of the Web page to the user terminal. To do.

The Web page is data constituting a screen to be provided to a user terminal designed using a prescribed format such as HTML (Hyper Text Mark-up Language), and is stored in a database.

  The Web page is prepared in advance as static data and dynamically generated by a predetermined program.

The predetermined program is executed when the Web server software receives a specific message by HTTP, and executes a process according to the message so that a Web page according to a user instruction is dynamically generated. To work. In addition to registration information by members, the database also stores Web pages and image information to be provided to users.

  In this embodiment, the server has a function that a single device (server) may have all the functions, or a plurality of devices (a plurality of servers connected to a network). May be.

  Next, an information registration method using this system will be described.

  A user who registers information first accesses the server address (URL) from the user terminal via the network (NW), and personal information (address, name, credit card) on the screen provided by the server information registration program. Member registration is performed by inputting the number, and if this is approved by the server, a registered user ID for identifying the registrant is given. As a result, the user becomes a member of the service and can register information in the service system.

  After completing the registration as a member, a plurality of arbitrary character strings (strings) are registered in the form of a multidimensional array as information indicating the characteristics of the registrant or registered contents. All of these pieces of information are registered in a user database constructed in the hard disk device (HD) of the server.

The character string in the user database is assumed to be closer to a word rather than a sentence, and the maximum length is limited to about 10 to 20 characters. However, if it is within this number of characters, it is not required to be a “word” in a strict sense. In the following description, the registration information based on the multidimensional array of character strings is referred to as a “registration matrix”. Since the information as the “registration matrix” is a core part of the registration information, it is assumed to be essential input information for transmitting information.

  The server information registration program defines the number of array dimensions, the number of elements, and the maximum length of a character string (string), and allows registration according to the rules.

  The information registration user can perform registration within the range of the number of dimensions, the number of elements, and the string length. For example, as shown in FIG. 1, assuming a case of a three-dimensional array, suppose that the highest dimension is “genre”, the second dimension is “theme”, and the lowest dimension is “keyword”. To do. Also, suppose that the maximum number of elements of “genre” is 3, the maximum number of elements of “theme” is 4, the maximum number of elements of “keyword” is 5, and the string length is 10 characters.

  In this case, the information registration user can register up to three arbitrary character strings as “genres”, and can register up to four “themes” related to one “genre”. Thus, up to five “keywords” related to one “theme” can be registered.

    That is, a maximum of three “genre” names, a maximum of 12 “theme” names, and a maximum of 60 “keywords” can be registered in a structure limited in the form of a three-dimensional array.

  FIG. 2 shows an example of a screen for registering a “registration matrix”. FIG. 2 shows an example in which one “genre” can be registered per page, and a plurality of “genres” can be registered by preparing a plurality of pages.

  Here, “self-introduction” is registered as a genre.

  “Job change” is registered as theme 1, and “homepage”, “Web designer”, “video”, “IT”, and “accounting” are registered as keywords related to theme 1.

  In addition, “hobby” is registered as theme 2, and “writing manga”, “cooking”, “walking”, “game”, and “movie” are registered as keywords related to theme 2.

  In addition, the theme 3 is “favorite TV program”, and keywords related to this theme 3 are “Oath of Chang ○○”, “Hana ○ Naru ○ Family”, “Hana Yori ○ Ko”, “TR ○ CK” "XX rider kabuto" is registered.

  The theme 4 is “favorite game”, and keywords related to this theme 4 are “Dora ○ IV”, “Dora ○ II”, “Dora ○ V”, “F ○ IX”, “Sakura ○ "Battle" is registered.

  For the information registered in the user database, the “registration matrix” described above is indispensable, but it does not exclude registering other information. For example, e-mail addresses, registrant profiles, and text Arbitrary input information such as comments by can be registered. These optional input information items can be selected by the registrant for each item as to whether or not to disclose them.

  FIG. 3 shows a data structure of a “registration matrix” stored in the user database. In this way, genres → themes → keywords are registered in a tree-like manner.

  Here, as a feature of the present embodiment, a user who searches for information can search for information by specifying data having the same structure as the registered essential input information.

  In other words, the information search from the user terminal is characterized in that it is performed by systematically specifying a plurality of character strings made up of the same multidimensional array as when information was registered.

  In the following description, search information designated by the multi-dimensional array of character strings, that is, a combination of “genre”, “theme”, and “keyword” is referred to as “search matrix”.

  FIG. 4 shows an example of an information search screen for searching as a “search matrix”.

  Here, “self-introduction” is registered as a genre.

  Then, “job change” is registered as theme 1, and “IT”, “Web”, “system engineer”, “search technology”, and “video distribution” are registered as keywords related to theme 1.

  In addition, “hobby” is registered as theme 2, and “sleeping”, “walk”, “manga”, “movie”, and “novel” are registered as keywords related to theme 2.

  In addition, “* not specified” is the theme 3 and keywords related to this theme 3 are “Murakami Yuki”, “Miya Miyuki”, “Kes ○ Nur”, “○ Ma Taro”, “Oshi ○ Director”. It is registered.

The theme 4 is “* not specified”, and keywords related to the theme 4 are “
“Chang XX”, “Dora XX”, “XX Kunishi”, “XX Rider”, and “Deadly XX people” are registered.

  The relationship between the “registration matrix” registered on the information registration screen shown in FIG. 2 and the “search matrix” input on the information search screen shown in FIG. 4 will be described in detail later.

  The server's information search program displays the result of the information search by collating the "search matrix" data input from the user terminal with the "registration matrix" data in the information registered by the information registration program. Search for information to do.

  FIG. 7 shows the flow of search processing.

  In the information search, first, search matrix data is received from the user terminal (steps 701 to 702), the character strings registered as “genre” are repeatedly collated (steps 703 to 705), and there is matching data. In this case, the character string of “theme” registered as belonging to the “genre” is collated, and if there is further matching data, it is registered as related to the “theme”. The character string “keyword” is collated (step 704).

  Then, the display target data as a search result is sorted in descending order at the overall total matching points (step 706), and the network (NW) ) To send via.

  FIG. 8 shows a matrix collation procedure.

  Here, collation of the registration matrix and the search matrix is repeated for all of the genre (steps 804 to 805), the theme (steps 807 to 808), and the keyword (steps 810 to 811) (steps 813 to 817).

  Here, if there is matching data for the “keyword”, a “matching point” for comparing the degree of matching is given to the registered information (step 814).

In this way, all such “keywords” belonging to the same “theme” are collated, and “match points” are calculated according to the number of matching “keywords” (step 8).
18).

  Such collation and calculation of matching points are executed for all “themes” and all “genres”, and the total “matching points” obtained are set as “total matching points” of the entire registration information ( Step 818).

  However, the “total match point” is not simply adding the number of matches of “keywords”. For example, when there are only one “keyword” that matches the same “theme” and when there are three “keywords” , “Match points” are weighted by the number of matches, such as 9 times instead of simply 3 times.

  This is because, for the same “theme”, the probability that only one “keyword” matches and the probability that three “keywords” match will have a large difference far exceeding three times.

This can be explained by the "law of product" in probability theory. That is, the probability that a "1" will appear when a dice is shaken once is 1/6, but the probability that an all "1" will appear when shaken three times is 1/216. In this way, a search result with a high probability that the user desires can be obtained by referring not only to the keyword but also the matching of the theme.
As a method for calculating the matching point, the matching degree value when X out of the number Y of elements in the search matrix matches the “theme” of the database registration matrix is expressed as a function F (X), and the function F (X) It is assumed that Y is raised to the Xth power, F (X) = Y ^X , and an information search is performed using the F (X) value of each vector added for all the themes as the matching point value. Good. Alternatively, if the number of elements of the “theme” in the registration matrix is Z, and the X matches among the elements of the “theme” in the search matrix, the function F (X) is expressed as F (X) = Z ^X may be used.
Further, as a generalized matching point numerical value calculation method, a function F (X) having the following properties may be defined and used in order to increase the matching point numerical value as the number of matches increases.
F (X + 1) / F (X)> (X + 1) / X, X> 0, F (0) = 0, F (1) = 1
And,
F (X + 2) / F (X + 1)> F (X + 1) / F (X)
The function F (X) has the property that when the number of matches X increases by 1, F (X) increases more than 1 and when X + 1 increases by 1 more than when X increases by 1. It has characteristics that increase rate is large. As an example, a function F (X) = Y ^X (Y> 1) can be considered.

FIG. 9 shows an example of calculating coincidence points for one “genre”. In this example, if the number of matches of “keywords” in the same “theme” is “N” and the match point for that “theme” is “Y”, the match point is expressed by the formula “Y = 3 to the Nth power”. Calculation is performed.

  This calculation is performed for all “genres”, and those points are added to obtain matching points of the registered information.

  As described above, in the present embodiment, the “match point” is given from the viewpoint of the similarity between the “registration matrix” and the “search matrix” and is used as an index for measuring the rarity.

  In this way, the “registration matrix” and the “search matrix” are collated, and a search result that gives the highest priority to “similarity” between the knowledge at the time of the search and the information to be obtained can be obtained. Even if the registered information has a low frequency, if there is a search user who happens to seek similar information, it can be easily found.

  The information retrieval program of the server executes the above collation for all the registered information, and sequentially sends the registered information having a high “match point” as data in HTML format to the user terminal that has performed the information retrieval. At this time, the server information search program does not require that the character strings be completely equal in determining whether or not there is a match between the data. When the character string “matrix” is included, it is regarded as matching data.

The server information search program displays the registration information retrieved from the user database in the form of a list in descending order of “match points”.
When sending a search result from the server to the client, when the client sends a search matrix to the server, a search is performed by setting a lower limit of the total matching points in advance in order to avoid sending a lot of information with low similarity to the server. A matrix may be created and transmitted.

  FIG. 5 shows a screen image for displaying a list of search results.

  When this list is displayed, according to the registration unit identified by the registered user ID, identification information regarding the registered user is displayed, and by selecting one of them, a “registration matrix” or other optional input is further displayed. Detailed registration information including information is displayed.

  FIG. 6 shows a screen image for displaying detailed information.

If the registrant of the registered information that has been found out publishes the e-mail address or URL of the homepage, the search user can contact the registrant directly without going through this service system based on the published information. Can take. When the registrant does not disclose the e-mail address or URL, a system is adopted in which the registrant can be contacted using the message transmission function of this service system. In the following description, this message function is referred to as “contact message”.

A service that accepts a user's membership registration and collates the registration matrix with the search matrix based on the profile and provides profile information to the search user is called a “WMP (Word Matrix Profiling) service”.
The “WMP service” can be operated as an independent service, but can also be used as an additional function of an existing member registration type service such as a blog or SNS (social networking service).

  FIG. 10 shows the flow of the “WMP service”.

  In the figure, the data transmission / reception relationship among the user terminal of the information registration user, the server, and the user terminal of the information search user is shown in time series.

  In addition to the “WMP service” described above, a search service for searching websites by using a new method different from the conventional search service is performed using the “OLL search method” of the information search function of the “WMP service”. be able to.

In the following description, a Web site search service using the “OLL search method” is referred to as an “OLL search service”.
The “OLL search service” can be operated as an independent search service, but can also be used as an additional function of the conventional search service.

Conventional search engines perform keyword extraction and rearrangement processing for searching for a web page when or after registration, but the data structure of the arranged keywords is as follows: It was not based on the premise of comparing the degree of agreement.
In a search engine that employs the “OLL search service”, when generating keyword data for searching a Web site, data having a structure that allows comparison of the degree of coincidence is generated. Provide a mechanism that enables display in descending order of match.

  In the “OLL search service”, registration is performed by embedding a “registration matrix” having the same structure as the “WMP service” in the HTML of each page constituting the Web site. That is, the information of the “registration matrix” is described in the form that does not affect the original display or operation of the page for the HTML of each page of the Web. A mechanism that only the information search program of the “OLL search service” interprets is created.

Based on certain rules (rules) prescribed by the information search program of “OLL search service”, the creator of the Web page describes “registration matrix” information in a form according to the rules. However, since this information is used only for the “OLL search service”, it is meaningless or invalid for the Web browser so as not to affect the original display and operation as well as other search services. Define the rules in such a way that they are descriptive. Specifically, this is realized by the following method.

The user who creates the Web page embeds a description for executing a script having a specific name (for example, OLLmatrix.ja) immediately after the HTML tag in the HTML of the target page, and this “OLLmatrix.ja” "The file is empty (0 byte file). On the other hand, “registration matrix” data having the same structure as that of the “WMP service” is created in the XML file format whose file name is OLLmatrix.xls and stored in the same folder as the file of OLLmatrix.ja.

  Rules such as the number of dimensions of the “registration matrix” and the number of elements in the array are defined by the side providing the “OLL search service”.

  If a user who creates a Web page creates a “registration matrix” file according to such a rule, the user can create a page as a search target of the “OLL search service” without performing member registration or the like.

  At this time, the description of HTML and the creation of the XML file may seem awkward at first glance, but the side providing the “OLL search service” distributes this registration information free of charge by distributing this tool free of charge. The problem can be solved.

  This “registration information creation support tool” has almost the same mechanism as the information registration program of “WMP service” with respect to the user interface.

  In the “OLL search service”, the information search function has the same mechanism as the information search program of the “WMP service”.

  The information search user searches for registered information by inputting a “search matrix”. The difference from the “WMP service” is that the registration information to be searched is a general Web page.

Next, a system integrated with a conventional Web site search engine will be described.
In an environment where an information search server that can search for information including all of an arbitrary number of keywords is used, a one-dimensional search matrix M (1) having keywords as n elements is generated at the client terminal, and M (1) Generate a total of j sub-search matrices MS ₁ (1), MS ₂ (1), MS _j (1) having m or more elements, but n ≧ m elements. . However, j is calculated by the following mathematical formula.
j = ( _n C _m ) + ( _n C _{m + 1} ) +,, + ( _n C _n )
Here, _x C _y is a combination function and represents the total number of combinations to be extracted from x.
_n C _m = n! / (M! × (nm)!)
Information that matches all the element keywords of these j sub search matrices is sequentially sent to the information search server for information search, the search result information is temporarily stored, and all the j sub search matrices are searched. After that, an information search client terminal device that arranges and displays the search result information from the number n of matching keywords to m sequentially.

In the above description, the “OLL search service” in a form that completely adopts the “OLL search method” has been described. However, the “pseudo OLL” that employs only a part of the concept of the “OLL search method”.
A search service is also possible.

  In this method, a function of “OLL search method” is incorporated into a part of an existing search service.

  In the “pseudo-OLL search service”, the information registrant does not need to register any special information (“registration matrix”) at the time of registration.

  For one Web page, a keyword extracted by a conventional search engine mechanism is reconstructed as one-dimensional “registration matrix” data.

  For example, in the case of a search engine that extracts keywords from the full text of a Web page, a “registration matrix” having an infinite number of elements in one dimension is generated from the extracted keywords.

  For information retrieval, a plurality of keywords are specified in a single row in the same way as existing retrieval services, but these cannot be distinguished from AND conditions and OR conditions, and are interpreted as a one-dimensional “search matrix”. The

  In the information search program, the “registration matrix” and the “search matrix” are collated according to the “OLL search method”, and the search results are displayed in descending order of the total matching points.

  In the “pseudo-OLL search service”, the accuracy is reduced in terms of evaluation of “similarity” of information because the matrix is a one-dimensional array and there is no quantitative limitation on registered information. This is effective when the creator of the page wants to perform a search that takes into account the element of "similarity" even without any special registration.

  When inputting a registration matrix consisting of “genre”, “theme”, and “keyword”, a list of already registered character strings can be displayed and selected from the list.

FIG. 11 shows an example of a screen for performing input by selection from a list.
“Unspecified” is allowed for words in higher-order arrays such as “genre” and “theme”.

For example, if “No specification” is entered for “Theme”, the search process always treats “Theme” as a match, and if “Keyword” matches, a match point is added. The value of the matching point is set lower than when “theme” is specified and “keyword” also matches.
When a search matrix format with “not specified” is sent to the server, the matching point of the corresponding item is calculated as a matching point assuming that only one keyword matches in the search, and the total matching point is calculated. Alternatively, the total matching point may be calculated on the assumption that all the items match.

  When there is an information search user who wishes to contact a registrant of specific registration information obtained by information search, the “WMP service” uses a “contact message” to notify the registrant to that effect. "Has a mechanism.

However, in order to prevent spam-like usage, the number of registrants who can send “contact messages” within a certain period is limited on the condition that they are registered as members in order to make this notification. May be provided.

  The information registered as the “registration matrix” is, in principle, character string data that can be freely input by the registrant, but the “WMP service” itself can also use a specific word as a reserved word. For example, “match” as “genre name” is a “genre” reserved word, “self-introduction” as “theme name” belonging to the “genre name”, “request from other party”, etc. are set as “theme” reserved words. Registrants who wish to register in this genre need to enter keywords related to “self-introduction”, “request for others”, etc., but any other “theme name” can be specified. It is.

  In the case of “genre” or “theme” reserved by “WMP service”, the number of dimensions and the number of elements of “registration matrix” must also be set on the side of “WMP service” specific to the reserved keyword. A mechanism that can For example, in the example of “match” and “genre”, the number of elements unique to the “genre” can be set in such a way that the number of “themes” belonging to the “genre” is six, A number of dimensions specific to the “genre” may be set under the “theme” of “self-introduction”, such as “hobby” or “profession” as a “sub-theme”.

  When the “genre” reserved by the “WMP service” is registered, a registration screen dedicated to the “genre” different from the normal registration screen may be provided.

  FIG. 12 shows a registration screen image dedicated to “employee recruitment genre” as an example. In addition, FIG. 13 shows a registration screen image dedicated to “matching genre” as an example.

  Information on reserved "genre name" and "theme name" is exactly the same as the data format in the server by adding a special code that can identify the reserved word. Distinguishes from data when a word is registered by arbitrary input.

  Regarding the reserved “genre” and “theme”, it is possible to input “keyword” in the form of selection from a list of words prepared in advance in addition to normal arbitrary input. For example, in the example of FIG. 13, when “annual income” is prepared as a “sub theme”, the keyword input method is not a normal arbitrary input but a list prepared in advance (such as a drop-down list). “4 million yen or less”, “4 million yen to 5 million yen”, “5 million yen to 6 million yen”, “6 million yen to 7 million yen”, “7 million yen to 8 million yen”, “ Items such as “8 million yen to 9 million yen”, “9 million yen to 10 million yen”, “more than 10 million yen” are displayed, and even if you input by selecting the corresponding item from among them Good.

  When searching for a reserved “genre” or “theme”, it is possible to set a “required matching word” so that a specific keyword can always be displayed as a search result only when it matches.

For example, in the example of FIG. 13, when “gender” is prepared as the “sub-theme”, for the information search user, it is the most important matter whether the registrant is male or female. If they do not match, no matter how high the matching points are in other items, it becomes meaningless information for the information search user, so it is desirable to be able to exclude it from the search results.
In the search matrix format, if some vectors or elements are “essential match words” and the search for the specified term does not match, the match point of the term (vector) is set to 0, or the upper vector of the item If the total match point of the entire search matrix is 0 and multiple “mandatory match words” are specified, the search results are sent to the client including which terms of the “mandatory match words” did not match. Also good.
FIG. 16 illustrates an example of a three-dimensional database registration matrix MA (3).
FIG. 17 shows a configuration example of a classical music database using a three-dimensional database registration matrix. The registration matrix MA (3) includes V ₁ , V ₂ , and V ₃ as first dimension vectors, and V ₁ is the _{first one} . It consists of two-dimensional vectors V ₁₁ , V ₁₂ , V ₁₃ , V ₁₄ , and similarly V ₃ consists of second-dimensional vectors V ₃₁ , V ₃₂ , V ₃₃ , V ₃₄ . Each three total of 12 of each of the two-dimensional vector V _11, V ₁₂ ,,, V ₃₄ each register element E ₁₁₁ is, E ₁₁₂ ,,, E _343, consisting of a total of 36 elements. Here are given the name of each vector, is to grant the "Beethoven" of the composer, for example, V _1, the V ₁₁ have granted the "symphony". A search matrix for searching a database created with such a registration matrix can be described in a format similar to this registration matrix.
The search matrix is preferably a format similar to the registration matrix for convenience. The database search server sends a search matrix format similar to the registration matrix to the client (search user) but with blank elements. It is possible to search for information by a procedure of describing a keyword to be found in an element.
FIG. 18 illustrates an example of a registration matrix MA (3) including empty vectors and elements (empty vectors and keyword elements belonging to it). In this matrix, an empty vector is, for example, a case where it is clear that there is no genre belonging to it, a case where it is clear that an empty element has no keyword belonging to it, or the number of subdivisions is the same as that of other vectors. It can be applied to cases where Considering music as an example, Brahms has no opera works, and the vector or element of that part can be empty.
Also, among the first dimension vectors, V ₃ has only V ₃₁ as the second dimension vector. This means that V ₁ and V ₂ are subdivided into second dimension vectors, but V ₃ is the second dimension vector. If the second dimension vector is only V ₃₁ when it cannot be subdivided into dimension vectors, the database can be constructed with V ₃ identical to V ₃₁ .
FIG. 19 shows an example of a search matrix MB (3) including empty vectors and elements. In this example, the second dimension vectors V ₁₂ , V ₃₂ , V ₃₄ and the elements belonging to them are empty, and the vector is not empty, but among the vector elements, S _{132 in} V ₁₃ and S _{333 in} V ₃₃ are empty. Yes. These empty vectors and empty elements have a registration and search matrix of a rectangular shape, a rectangular parallelepiped shape, or a shape of a super rectangular parallelepiped (a rectangular parallelepiped in a coordinate system of four dimensions or more) when the dimension is four or more. Depending on the contents of the database, some vectors must be divided again. In general, some vectors cannot be further subdivided, or the number of subdivisions varies depending on the item. At this time, empty vectors and elements can be used to realize both of these situations in a matrix.

  The present invention can be used for a search system on the Internet.

The figure which shows the conceptual structure of the registration matrix which is embodiment of this invention Figure showing the registration matrix information registration screen Diagram showing the data structure of the registration matrix Diagram showing search matrix information inquiry screen Figure showing search result list display Figure showing the detailed information screen of search results Flow chart showing search processing procedure Flow chart showing matching procedure between registration matrix and search matrix Figure showing an example of calculating keyword matching points Sequence diagram showing the procedure of WMP service Figure showing the list reference screen Figure showing the reserved genre registration screen (1) Figure showing the reserved genre registration screen (2) The figure which shows the system configuration | structure of embodiment The figure which shows the hardware constitutions of the server of embodiment Example of 3D database registration matrix MA (3) Example of 3D database registration matrix MA (3) in music database Illustrative example of 3D database registration matrix MA (3) with empty vectors and empty elements Example of 3D database search matrix MB (3) in a music database with empty vectors and empty elements

Explanation of symbols

SV server UT User terminal NW Network CPU Central processing unit MM Main memory BUS Bus HD Hard disk device KBD Keyboard DISP Display device OS Operating system

Claims (10)

It consists of a server that stores a registration matrix consisting of a combination of data, a network, and a search user terminal connected via the network,
A search matrix consisting of a combination of data input from a user terminal connected via a network is sent to a registration matrix consisting of a combination of data stored in the server,
In the server, an information search system that collates the search matrix input in the above and the registration matrix, outputs a search result based on similarity between the matrices, and transmits the search result to the search user terminal.   The information search system according to claim 1, wherein the registration matrix and the search matrix include a genre, a theme, and two or more keywords related to the theme.   The information search system according to claim 1, wherein in the comparison between the registration matrix and the search matrix, a theme having a large number of matching keywords is determined to have high similarity by setting a high matching point.   The information search system according to claim 3, wherein the number of matches is n, and the nth power of a predetermined parameter is calculated as a match point. It consists of a server that stores a registration matrix consisting of a combination of data, a network, and a search user terminal connected via the network,
Setting a registration matrix comprising a combination of data stored in the server;
Transmitting a search matrix comprising a combination of data input from user terminals connected via a network;
In the server, collating the search matrix inputted in the above and the registration matrix and outputting a search result by similarity between the matrices;
Transmitting the search result to the search user terminal. A search processing program executed by the server, comprising a server for storing a registration matrix comprising a combination of data, a network, and a search user terminal connected via the network,
Setting a registration matrix comprising a combination of data stored in the server;
Transmitting a search matrix comprising a combination of data input from user terminals connected via a network;
In the server, collating the search matrix inputted in the above and the registration matrix and outputting a search result by similarity between the matrices;
An information search program executable on a server comprising the step of transmitting the search result information to the search user terminal. A server that accumulates a registration matrix consisting of a combination of data and accepts a search matrix from a search user terminal connected via a network,
In the server, a search matrix consisting of a combination of data input from a user terminal connected via a network is received for a registration matrix consisting of a combination of accumulated data,
In the server, an information search server that collates the search matrix input in the above and the registration matrix, outputs a search result based on similarity between the matrices, and transmits the search result to the search user terminal. A system for performing information retrieval according to a predetermined database structure from a database created in advance with a structure of a predetermined classification method, and the structure of the database is determined in the form of an N-dimensional matrix MA (N) , the n-size dimension matrix to the _{n 1 xn 2 x ,, n n} , V 1 a first dimension of the (n-1) dimension element vectors constituting the n-dimensional matrix, V ₂ ,,, V _n1 and then, then the respective (N-1) dimensional vector V _1, V ₂ ,,, V ₁₁ the first two-dimensional vector with respect to V ₁ constituting the _{V n1, V 12 ,,, V 1n2} , V with respect to V _{2 21} , V _22, ..., V _2n2, and similarly V _n11 , V _n12, ... V _n1n2 with respect to V _n1 , and similarly, each element constituting the (N−1) -dimensional vector V ₁ ,, ₁ E ₁ ,, _{11
, E 1,, 12 ,,, E} 1,, and _{1 nN,} hereinafter likewise _{V n1,, n (N-} 1) each element constituting E of _{n1,, n (N-1} ) 1, E n1, _{, n (N-1) 2,} ..., E _n1 ,, _{n (N-1) nN,} and using a similar N-dimensional search matrix MB (N) to search a database constructed in this way , The size of the N-dimensional search matrix MB (N) is m ₁ xm ₂ x ,, m _N , and each (N-1) -dimensional element vector of the first dimension constituting MB (N) is represented by U ₁ , U _2. ,,, and U _n1, then the (N-1) dimensional vector U _1, second constituting the U ₂ ,,, U _n1
The dimension vector is U ₁₁ , U ₁₂ ,, U _1n2 for U ₁ , U ₂₁ , U ₂₂ ,, U _{2n2 for} U ₂ , and so on, as well as U _n11 , U _n12 ,, U _n1n2 for U _n1. , and so the (N-1) dimensional vector U _1,, S ₁ the elements constituting the _{1,, 11, S 1,} , 12 ,,, S 1,, and _{1 nN,} Similarly U _{n1, , n (N-1)} is represented by S _n1 ,, _{n (N-1) 1} , S _n1 ,, _{n (N-1) 2} ,, S _n1 ,, _{n (N-1) nN} And quantifying the degree of matching of each element group for each vector of each (N-1) -dimensional vector group of the database classification N-dimensional matrix MA (N) and information retrieval N-dimensional matrix MB (N) An information retrieval device that sequentially outputs information retrieved together with a degree value. A system for performing information retrieval according to a predetermined database structure from a database created in advance with a structure of a predetermined classification system, the structure of the database being determined in the form of an N-dimensional registration matrix MA (N) The size of the N-dimensional registration matrix is n ₁ xn ₂ x ,, n _N, and each (N-1) -dimensional element vector of the first dimension constituting the N-dimensional registration matrix is represented by V ₁ , V ₂ ,. , and V _n1, then the respective (N-1) V ₁₁ with respect to dimensional vectors V _1, V ₂ ,,, V ₁ the second dimensional vectors forming the _{V n1, V 12 ,,, V 1n2} , V 2 and V _21, V ₂₂ ,,, V _2n2 respect, and so the _{V n11,} V _n12 ,,, V n1n2 respect V _n1, Similarly the (N-1) constituting the dimensional vector V ₁ ,, ₁ Each element is set to E ₁ ,, ₁₁ , E ₁ ,, _12, ..., E _{1,, 1nN,} and similarly, each element constituting V _{n1 ,, n (N-1)} is set to E _{n1 ,, n (N -1) 1} , E _n1 ,, _{n (N-1) 2} ,, E _n1 ,, _{n (N-1) Suppose} that _nN , some vectors of the (N-1) th dimension from the first dimension of the registration matrix are empty vectors, and some of the elements of the (N-1) th dimension are empty, like this When using a similar N-dimensional search matrix MB (N) to search the constructed database, the size of the N-dimensional search matrix MB (N) is m ₁ xm ₂ x,
, _{M N} and the first (N-1) dimensional element vectors constituting the MB (N) as U ₁ , U ₂ , U _n1, and then each (N-1) dimensional vector U ₁ the second dimension vectors forming the U ₂ ,,, U _n1 and U _11, U ₁₂ ,,, U _1n2 respect U _1, and U _21, U ₂₂ ,,, U _2n2 respect U _2, hereinafter likewise U and _{U n11,} U _n12 ,,, U n1n2 respect _n1, hereinafter likewise the (N-1) dimensional vector U _1,, S ₁ the elements constituting the _{1,, 11, S 1,} , 12 ,,, S _{1,, 1nN,} and similarly, each element constituting U _{n1 ,, n (N-1)} is represented by S _{n1 ,, n (N-1) 1} , S _{n1,, n (N-1) 2} , , S _n1 ,, _{n (N-1) nN,} and each of the (N-1) -dimensional vector groups of the database classification N-dimensional registration matrix MA (N) and the information search N-dimensional search matrix MB (N). The degree of coincidence of each element group for each vector was quantified without performing it on empty vectors and empty elements in the registration matrix MA (N), and searched together with the degree of coincidence value. Information retrieval apparatus for sequentially outputting the broadcast.   In claim 7, when a server that performs information retrieval according to a predetermined database structure from a database and a client that performs information retrieval are connected, a search matrix format based on a database registration matrix format is first sent to the client, An information search system for generating a search matrix using the received search matrix format and sending the generated search matrix to a server for information search.

Images