JP2008065417A - Associative word group search device, system, and content match type advertisement system - Google Patents

Abstract

A searcher is provided with a wide group of associative words related to a search keyword without troublesome work.
A PC 200 operated by a searcher, a website search support apparatus 100 for extracting an association word group highly relevant to a search keyword transmitted from the PC 200, and a DB 300 in which the website is stored And a search engine 400 for searching a website stored in the DB 300, a website search support apparatus 100, a PC 200, a network 500 such as the Internet for connecting the DB 300 and the search engine 400 to each other.
[Selection] Figure 1

Description

  The present invention relates to an associative word group search device, a system, and a content match type advertisement system, for example, an associative word group search device, a system, and a content match type advertisement system that support website search using a keyword matching type search engine. .

  Conventionally, when searching for a website, a searcher inputs a keyword related to information to be searched to a search engine such as a keyword matching type. This type of search engine searches the website containing the keyword by referring to the database storing the website based on the keyword, and outputs the search result to the searcher. Yes.

  However, if the keyword entered by the searcher is not appropriate, the search result may not include the website intended by the searcher. In such a case, the searcher is forced to perform troublesome work of finding a website containing necessary information by changing the keyword and repeating check and error.

  In order to avoid such an inconvenience, in order to acquire a keyword optimal for a website search, a homepage displayed by a browser, a keyword extraction unit that extracts a keyword associated with the homepage, and the keyword extraction unit An invention relating to software including keyword display means for displaying the keywords extracted in (1) is disclosed (Patent Document 1).

JP 2002-149668 A

  However, the invention disclosed in Patent Document 1 has the following problems.

  First, the searcher is forced to perform troublesome work. For example, the searcher needs to perform an operation such as “select a car as a classification when searching a homepage using automobile sales as a keyword” and an operation “input a search expression into the search expression input box 4”.

Secondly, the keyword extraction targets are limited. So, for example,
(1) Depending on the search conditions used in the search engine, it is difficult to provide a wide range of associative words to searchers,
(2) Even when the relationship between information that the searcher wants to obtain and the search keyword entered by the searcher is sparse, it is difficult to provide a wide range of associative words to the searcher.

In particular,
(1) When the search condition used in the search engine is to output a website including a search keyword and having a large number of links as a higher rank of the search result, or
(2) If the information that the searcher wants to obtain is “broad information necessary to open used car sales”, but the search keyword entered by the searcher is “car sales”,
As search results, a wide variety of information such as information on new car dealers and car dealers can be obtained, but it is difficult to obtain information specialized for “used car sales”. On the other hand, it is usually impossible to obtain information on opening funds, accounting, sales, etc. necessary for “opening”.

  Accordingly, an object of the present invention is to provide a wide range of associative words to a searcher without the searcher performing troublesome work.

In order to solve the above problems, the associative word group search device of the present invention is:
A search means for searching a group of websites including the input keyword;
Execution means for executing a process of dividing text data included in some websites of the website group searched by the search means;
Output means for outputting, as a group of associative words of the keyword, a word whose frequency of appearance on each website among the words obtained by the dividing process of the execution means is a predetermined number or more.

  That is, the present invention searches for an associated word group of input keywords using a plurality of websites as a corpus.

  In addition, the present invention may include a filter unit that filters the word based on the word having the predetermined number of appearances and the keyword and the keyword. Thereby, the association word group finally provided can be narrowed down.

The filter means includes
The number of hits of websites that include both the word and the keyword whose appearance count is greater than or equal to a predetermined number,
The number of hits on the website containing the keyword,
It is also possible to filter the words based on a statistical process using the number of hits of a website that includes words whose appearance count is greater than or equal to a predetermined number.

  Thereby, the narrowing-down precision of an associative word group can be improved.

  When the number of hits of a website including the word whose number of appearances is greater than or equal to a predetermined number is less than or equal to a predetermined threshold, it is preferable to have an excluding unit that excludes the word. This is because by reducing the number of words to be processed that are considered to have little contribution to the finally output associative word group, it is possible to reduce the load on the device required for searching the associative word group.

BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In addition, in each figure, the same code | symbol is attached | subjected to the same part.
(Embodiment 1)
FIG. 1 is a block diagram showing a schematic configuration of a website search support system including a website search support apparatus 100 which is an associative word group search apparatus according to Embodiment 1 of the present invention.

  FIG. 1 shows a personal computer (hereinafter referred to as “PC”) 200 operated by a searcher or the like, and a web for extracting a group of associative words highly relevant to a search keyword transmitted from the PC 200. Site search support apparatus 100, database (hereinafter referred to as “DB”) 300 in which a website is stored, search engine 400 that searches a website stored in DB 300, and website search support apparatus 100 And a network 500 such as the Internet that connects the PC 200, the DB 300, and the search engine 400 to each other.

  The website search support apparatus 100 includes a receiving unit 10, a searching unit 20, a storing unit 25, an extracting unit 30, a reading unit 40, an executing unit 50, a selecting unit 60, and an acquiring unit 70, which will be described below. A filter unit 80, a classification unit 85, a creation unit 90, and a transmission unit 95.

  The receiving means 10 receives a search keyword input to the PC 200 by a searcher or the like and then transmitted via the network 500. The search keyword received by the receiving unit 10 is stored in a memory (not shown) by the storage unit 25.

  Here, the administrator or the like of the website search support apparatus 100 searches the search web page for the web page searcher to input the search keyword, and the search result for inserting the search result of the web page and the advertisement corresponding thereto. Prepare a page.

The search means 20 searches the DB 300 using the search engine 400 for the website group A (generic name for the websites A1, A2, A3...) Including the search keyword received by the receiving means 10. . The search means 20 also acquires the number of websites (number of hits) including the search keyword at the time of the search. Similarly, the search means 20
(1) Associative word candidates a1, a2, a3,. . . Search for websites that each contain the number of hits associated with each,
(2) Associative word candidates b1, b2, b3,. . . Search for websites containing each of them, get the number of hits related to each,
(3) Associative word candidates a1, a2, a3,. . . A search for a group of websites including both the search keyword and the search keyword received by the receiving means 10, and obtaining the number of hits related to each
(4) Associative word candidates b1, b2, b3,. . . And a search for a website group including both the search keyword received by the receiving means 10 and the number of hits related to each search.

  Associative word candidates b1, b2, b3,. . . Are associated word candidates a1, a2, a3,. . . Therefore, the acquisition process shown in (3) above may be omitted.

  The storage unit 25 stores the search keyword received by the reception unit 10, the search result acquired by the search of the search unit 20, the number of hits, and the like in a memory (not shown).

  The extraction means 30 is provided by the websites A1, A2, A3. . . When the number of websites indicating the number exceeds a predetermined number, for example, the top 10 or 10 websites B that are thinned out or randomly selected (generic name for websites B1, B2,..., B10) are extracted. is there. Note that setting the number of extractions of the website B to 10 is an example, and the number of extractions may be, for example, 5 or 20. The extraction method will be described later.

  The reading means 40 actually reads the website B extracted by the extracting means 30 from the DB 300, and further reads the text data contained therein using a web wrapper or the like as necessary. is there.

  The execution unit 50 performs morphological analysis included in the division process on each text data read by the reading unit 40 using a probability model stored in a memory (not shown) or the like. .

  The selection means 60 provides an associative word candidate a that provides the user with words having a TF (Term Frequency) value or a DF (Document Frequency) value of 2 or more among the words obtained by the morphological analysis executed by the execution means 50. (Associative word candidates a1, a2, a3,...) Are selected.

  The acquisition unit 70 (1) associative word candidates a1, a2, a3,. . . Acquire associative word candidate b (generic name of associative word candidates b1, b2, b3,...) Obtained by excluding associative word candidates related to websites having, for example, 5000 hits or less from associative word candidate a. It is. The processing by the acquisition means 70 is performed by (3) associative word candidates a1, a2, a3,. . . In order to alleviate the search burden of the website group including both the search keyword and the search keyword, this is selectively performed and not necessarily performed. Note that the number of 5000 cases, which is an acquisition condition for acquiring the associative word candidate b from the associative word candidate a, is an example. The number of cases may be determined according to, for example, the number of websites stored in the DB 300, the number of hits of websites including the search keyword transmitted from the PC 200, and the like. As an example, when the number of hits of the website including the search keyword transmitted from the PC 200 is 500,000, the acquisition condition of the associative word candidate b is 5000, and when the number is 600,000, the acquisition condition is 6000. do it.

  The filter means 80 filters the associative word candidate b based on statistical processing such as mutual information or t-score using the number of hits stored in the storage means 25 to obtain an associative word group of search keywords. Is to do. However, the filtering process does not necessarily have to be executed when the CPU has sufficient processing capability for the number of associative word candidates b.

  The classifying unit 85 creates a mutual network of the associative word groups using the mutual information or the χ-square value for the associative word groups acquired by the filter unit 80, and performs processing such as classification using the Newman method. Is what you do. Note that the chi-square value can be calculated using, for example, “Feature Selection, Perceptron Learning Usability Case Study for Text Categorization, Proc of ACM SIGIR '97, 1997” by Kok Leong et al.

  The creation unit 90 creates a display screen (website) including the association word group classified by the classification unit 85.

  The transmission unit 95 transmits the display screen (web site) created by the creation unit 90 to the PC 200 via the network 500.

  FIG. 2 is a schematic explanatory diagram of the operation of the website search support apparatus 100 shown in FIG. First, when the search keyword is received, the website search support apparatus 100 uses the search engine 400 to search the website group A containing the search keyword from the websites stored in the DB 300. To do. As a result, the website search support apparatus 100 can acquire a website indicating the website group A or the like that includes the search keyword.

  This website usually includes websites A1, A2,. . . , An respectively specifying identification information (for example, URL) and the number of search keyword hits.

  The website search support apparatus 100 includes websites A1, A2,. . . , An are actually read from the DB 300, and text data in each website A1 and the like are further extracted. Thereafter, the website search support apparatus 100 extracts words such as nouns by performing morphological analysis or the like from the text data.

  The website search support apparatus 100 identifies a frequently used word among the words by n-gram processing or the like, and selects an associative word group of the search keyword based on the identified word and the search keyword. Then, the associative word group is classified (clustered) using a Newman method or the like.

  FIG. 3 is a flowchart showing the operation of the website search support apparatus 100 shown in FIG. Hereinafter, with reference to FIG. 3, a specific method for realizing the operation outline described with reference to FIG. 2 will be described.

  First, the website search support apparatus 100 transitions to a search keyword transmission waiting state from the PC 200 during operation (step S1).

  In this state, it is assumed that the user of the website search support apparatus 100 transmits a search keyword input by operating the PC 200 to the website search support apparatus 100 through the network 500. In this case, the website search support apparatus 100 receives the transmitted search keyword by the receiving unit 10 and outputs it to the storage unit 25 and the search unit 20.

  When the search keyword output from the receiving means 10 is input, the search means 20 uses the search engine 400 to search the website group A containing the search keyword from the DB 300. Processing is executed (step S2).

  As a result, the search means 20 can search the websites A1, A2, A3. . . It is possible to acquire a website in which specific information (for example, URL) for specifying each and the number of search keyword hits is described.

  Thereafter, the search means 20 extracts a description portion such as “about XX search results” based on a tag in the website or by text matching or the like. Thus, the search means 20 acquires the number of hits for the search keyword. However, the number of hits may be acquired using a search engine API (Application Interface). Then, the search unit 20 outputs the specific information to the extraction unit 30 and outputs the specific information and the number of hits to the storage unit 25.

  When the specific information is output from the search unit 20, the storage unit 25 executes a storage process in which the specific information is associated with the search keyword corresponding to the specific information and stored in the memory (step S3).

  The extraction unit 30 is configured to output the websites A1, A2, A3,. . . For example, the websites B1, B2,. . . , B10 specific information is extracted and output to the reading means 40 (step S4).

  As described above, when extracting the top 10 search results, the hyperlink includes “http” and the like from the website indicating the top 10 search results generated by the search engine 400. The letters attached to the websites B1, B2,. . . , B10 can be extracted as specific information. Websites B1, B2,. . . , B10 may be selected by arbitrarily extracting characters including “http” etc. from an arbitrary website indicating the search result.

  The reading means 40 includes the websites B1, B2,. . . , B10 based on the specific information of B10, actually reading out the website B from the DB 300, extracting each text data included in the website B using a web wrapper or the like, and outputting it to the execution means 50 Is executed (step S5).

  The text data is extracted from websites B1, B2,. . . , B10 may be all the pages constituting each of them, or some specific pages such as each top page.

  The execution unit 50 performs morpheme analysis on each text data output from the reading unit 40 using morpheme analysis software such as tea bowl and Japanese cloth bowl (both product names) (step S6). The execution unit 50 outputs the word obtained as a result to the selection unit 60. Here, by selecting the morpheme type prepared by the morpheme analysis software as appropriate, for example, it may be set not to extract suffixes, numbers, non-independent words, pronouns, etc. while extracting nouns and unknown words, for example. .

  Note that n-gram division or the like may be performed instead of or in combination with morphological analysis (for example, n = 5). In this case, compound word processing may be performed as necessary, for example, by handling words having an appearance degree of 0.7 or more as the same word.

  The selection unit 60 executes a selection process of the associative word candidate a in which a word having a TF value or a DF value of, for example, two or more output from the execution unit 50 is selected as the associative word candidate a to be provided to the user ( Step S7). In this way, words that appear with high frequency or low frequency are excluded from the associated word candidates.

  The search means 20 outputs the associative word candidates a1, a2, a3,. . . Are input from the DB 300 using the search engine 400, the associative word candidates a1, a2, a3,. . . The number-of-hits acquisition process of acquiring the number of hits related to each search result is performed by searching for websites that respectively include (step S8).

  And the search means 20 is the associative word candidate a1, a2, a3,. . . And the number of hits corresponding to these are output to the acquisition means 70 and the storage means 25 as a pair. The storage means 25 stores these pieces of information in the memory in association with the corresponding search keywords already stored.

  The acquisition unit 70 associates the associative word candidates a1, a2, a3,. . . Among these, the associated word candidates whose corresponding hit count is, for example, 5000 or less are deleted from the associated word candidate a, and the associated word candidates b1, b2, b3,. . . Is acquired (step S9).

  The acquisition means 70 acquires the acquired associative word candidates b1, b2, b3,. . . Is output to the search means 20. The search means 20 outputs associative word candidates b1, b2, b3,. . . , The associative word candidates b1, b2, b3,. . . The number-of-hits acquisition process of acquiring the number of hits related to each search result is performed (step S10).

  And the search means 20 is the associative word candidate b1, b2, b3,. . . Then, a hit number storing process is executed in which the hit numbers corresponding to these are stored in the storage means 25 as a pair (step S11).

  In addition, the search means 20 uses the associative word candidates b1, b2, b3,. . . The search engine 400 is used to search for a website that includes both of the above and the search keyword received by the receiving means 10 and stored in the storage means 25, thereby obtaining the number of hits related to each search result. The hit number acquisition process is executed (step S12).

  And the search means 20 performs the hit number storage process of storing each hit number in the storage means 25 (step S13).

As a result, the storage means 25 stores
(1) Associative word candidates b1, b2, b3,. . . And the number of hits relating to a website including both the search keyword received by the receiving means 10 and stored in the storage means 25;
(2) Associative word candidates a1, a2, a3,. . . The number of hits related to websites including
(3) Associative word candidates b1, b2, b3,. . . The number of hits related to websites including
(4) The number of hits related to the website including the search keyword received by the receiving unit 10 and stored in the storage unit 25 is stored.

The filter means 80 reads out the required number of hits stored by the storage means 25 and substitutes these into Equation 1 to calculate the mutual information amount. And the filter means 80 performs the filter process of filtering the associative word candidate b in order to take out, for example, the top 100 cases having a large numerical value from the calculation results (step S14).
here,
The right side denominator hit (associative word candidate b) in Equation 1 is the number of hits of the website when the associated word candidate b is searched, that is, when the associated word candidate b is searched as a query.

  The right denominator hit (search keyword) in Equation 1 is the number of hits on the website when the search keyword is used as a query.

  The right-side molecule hit (associative word candidate b, search keyword) in Equation 1 is the number of hits on the website when both the associative word candidate b and the search keyword are searched as queries.

N is a coefficient, and in this embodiment, for example, 100 × 10 ⁶ to 10 ⁷ .

  For this purpose, the filter unit 80 stores a calculation program using the mathematical formula 1 in a storage area (not shown), reads the program prior to the filter process, and calculates the required number of hits from the storage unit 25. Read and perform calculation using Equation 1. Then, the filter unit 80 arranges the calculation results, for example, in descending order of the numerical value, deletes the associated word candidates other than the top 100, and outputs the remaining as associated word groups to the classifying unit 85.

  The classification unit 85 classifies the associative word group acquired by the filter unit 80 by using the following method.

  4 and 5 are explanatory diagrams of the associative word grouping technique performed by the classifying unit 85 of FIG. FIG. 4 shows a state in which associative words are connected to each other via a network. The connection may be realized by a known method such as connecting the ones having close chi-square values. FIG. 5 shows a state where the associated word groups connected to the network are classified. The classification may be realized by a known method using a Newman method or the like.

  The creation unit 90 executes a screen creation process of creating a display screen for information indicating the associated word group output from the filter unit 80 (step S15).

  The transmission unit 95 executes a transmission process of transmitting the display screen created by the creation unit 90 to the PC 200 via the network 500 (step S16).

  As a result, the user of the PC 200 can acquire an associative word group that is highly related to the search keyword transmitted to the website search support apparatus 100. For this reason, even if the user does not obtain a desired search result by the search based on the search keyword, the user can easily obtain the desired search result by performing a refinement search based on the acquired associative word group. Become.

(Embodiment 2)
Next, the content match type advertising system 1000 according to the second embodiment of the present invention will be described. The content-matching advertising system 1000 advertises an advertisement having a high relevance to the contents included in the website in the website where the advertisement can be placed. The content match type advertising system 1000 includes an associative word group search device 100 corresponding to the website search support device 100 described in the first embodiment.

  The associative word group search apparatus 100 includes means for storing the associative word group in the associative word group DB 600 in the storage means 25 in place of the creation means 90 and the transmission means 95 of the website search support apparatus 100. Yes.

  FIG. 6A is a schematic explanatory diagram of the content match advertising system 1000 according to the second embodiment of the present invention. For comparison, an outline description of a conventional advertising system is also shown in FIG.

  As shown in FIG. 6A, the content match advertising system 1000 of the present embodiment includes an associative word group search device 100, an associative word group DB 600, a matching device 700, and an advertisement DB 800, which will be described below. ing.

  When a keyword is input from the administrator of the content match advertising system 1000, the associative word group search device 100 searches for an associative word group corresponding to the keyword, and is a unique identifier for each associative word group Assigned a unique cluster ID. The cluster ID will be described later.

  The associative word group DB 600 stores an associative word group searched by the associative word group searching apparatus 100 and a cluster ID assigned corresponding to the associative word group. Note that keywords input from the administrator of the content match advertising system 1000 may also be stored in association with the association word group or the like.

  In accordance with an instruction from the administrator of the content matching advertising system 1000, the matching device 700 includes the associative word group stored in the associative word group DB 600 and the text data in the website as the content stored in the DB 300. And the cluster ID that matches the website is determined.

  The advertisement DB 800 stores a pair of an advertisement sent from an advertiser and a cluster ID that matches the advertisement.

  Next, the operation of the content match advertising system 1000 will be described with specific examples. First, the administrator or the like of the content match advertising system 1000 selects a keyword related to the advertisement target and inputs the keyword to the associative word group search device 100.

  For example, if there is a magazine “○ × △” and you want to consult with the advertiser about advertising, “magazine” or “actor name” in charge of the TV commercial A website including keywords such as the “sales company name” of the magazine and the “catchphrase” of the magazine will be selected as the keyword.

  When the keyword selected in this way is input, the associative word group search apparatus 100 searches for an associative word group corresponding to each keyword as described in the first embodiment. When the associative word group search device 100 finishes searching for the associative word group, the associative word group search apparatus 100 assigns a unique cluster ID to the searched associative word group, associates them with each other, and stores them in the associative word group DB 600.

  On the other hand, the administrator of the content-matching advertising system 1000 presents to the advertiser of the magazine “○ × △” and consults the advertisement placement on the basis of the associated word group corresponding to the input keyword, that is, the associated word. Consultation on the purchase of advertising space on a group basis. The advertisement placement based on the association word group means that an advertisement sent from the advertiser is placed on the advertisement space of the website that includes the association words constituting the association word group.

  In fact, if the advertiser of the magazine “○ × △” purchases an advertising space on an associative word group basis in response to the above consultation, the cluster ID corresponding to the associative word group purchased by the advertiser is A pair with the advertisement provided by the advertiser is registered in the advertisement DB 800.

  In this state, for example, it is assumed that a website such as a diary-type blog (weblog) and a bulletin board having an advertising space is created. Further, the website is uploaded to the DB 300, and then the URL of the website is notified to the administrator of the content match advertising system 1000 or the text data of the website is transmitted. Suppose it was made.

  In this case, the administrator of the content match advertising system 1000 inputs the URL and the like to the matching device 700.

  When the URL is input, the matching device 700 executes a process of extracting text data of the web page corresponding to the URL, and the extracted text data and the association word group stored in the association word group DB 600 and / or Perform keyword matching processing. However, when the text data of the website is transmitted to the administrator of the content match advertising system 1000, this extraction process is omitted.

  The matching device 700 determines the cluster ID corresponding to the matching target website based on the matching result obtained by the matching process. Then, the cluster ID matching the website is determined and added to the website stored in the DB 300.

  As a result, when the website stored in the DB 300 is downloaded by some people, the advertisement corresponding to the cluster ID related to the website is read from the advertisement DB 800 as in the known method, Applied to the inventory on the website.

  Here, in the case of an advertisement using a conventional advertisement system, a system administrator or the like has consulted with an advertiser on advertisement placement on a keyword basis related to the advertisement. Specifically, in the above example, for example, “magazine”, “actor name” in charge of the TV commercial, “seller name” of the magazine, “catchphrase” of the magazine, etc. Was consulting about the purchase. The keyword-based advertisement placement means that an advertisement sent from the advertiser is placed on the advertisement space of the website containing the keyword. Therefore, in order to perform an appropriate advertisement, the advertiser has been forced to purchase an advertisement space related to a plurality of keywords.

  Note that the timing for determining the cluster ID corresponding to the website to be matched is not limited to the above example. For example, when an instruction to download a website stored in the DB 300 is received, or a timer is provided, for example, regularly every 10 days, or an access counter is provided to access the website. The number of times may be non-periodic such as every 10 times. In this way, even in the case of a website whose contents are sequentially updated, such as a blog, an advertisement that matches the contents can be posted.

(Embodiment 3)
FIG. 8 is a schematic explanatory diagram of the content match advertising system 1000 according to the third embodiment of the present invention.

  The content match type advertisement system 1000 shown in FIG. 8 is different from the second embodiment in the method of searching for an associative word group in the associative word group searching apparatus 100. Specifically, in the case of the second embodiment, a group of associative words is searched based on one keyword, while in the case of the third embodiment, a plurality of associative words are searched based on one keyword. Search for a group of words.

  When such a search is performed, an associative word group for the keyword is obtained by clustering.

  FIG. 7 is a diagram illustrating an example of data stored in the associative word group DB 600 illustrated in FIG. FIG. 7 shows the association word groups and the cluster IDs corresponding to them.

  Here, for example, it is assumed that there is a magazine with the title “◯ × Δ”, and there is also a juice with a trade name “○ × Δ”. In such a situation, it is assumed that the user inputs a search keyword “OxΔ” to the search engine and searches the website. As a result, a search result page including both the website regarding the juice and the website regarding the magazine is created.

  Therefore, in the case of a general advertisement system, an advertisement of a magazine “OxΔ” may be displayed even when a website related to the juice “OxΔ” is browsed.

  On the other hand, the content match advertising system 1000 according to the present embodiment determines whether the article content of the website relates to the juice “OxΔ” or the magazine “OxΔ”. The advertisement of the magazine “OxΔ” is avoided from being displayed on the website related to “OxΔ”.

  When the administrator of the content-matching advertising system 1000 simply inputs the keyword “OxΔ” to the associative word group search apparatus 100, the associative word groups related to “juice” and “OxΔ” (the cluster in FIG. 7). A group of associative words related to ID = 1) and an associative word group related to “magazine” and “OxΔ” (associative word group related to cluster ID = 2 in FIG. 7) are clustered. Will be obtained.

  As described above, in the content match advertising system 1000 according to the present embodiment, for example, when a keyword “OxΔ”, which is a product name, is input to the associative word group search device 100, an associative word group corresponding to the keyword is input. Are clustered and output. For this reason, the administrator or the like of the content match type advertising system 1000 gives the advertiser of the juice “OxΔ” to the advertiser of the magazine “OxΔ” that is associated with the cluster ID = 1 in FIG. On the other hand, it is only necessary to consult advertisements based on the associative word group related to cluster ID = 2 in FIG. The subsequent operation is as described in the second embodiment.

(Embodiment 4)
FIG. 9 is a schematic explanatory diagram of the content match advertising system 1000 according to the fourth embodiment of the present invention.

  The content match type advertising system 1000 shown in FIG. 9 is different from that of the second embodiment in the advertisement placement target. Specifically, in the case of the present embodiment, the advertisement placement target is the search result page of the website. Hereinafter, a method for realizing this will be described.

  First, an administrator or the like of the content match advertising system 1000 uses a search web page for a web page searcher to input a search keyword, and a search result page for inserting a search result of the web page and an advertisement matching the search result. Prepare.

  In addition, as described above, when a keyword related to an advertisement is input, the associative word group search apparatus 100 searches for an associative word group including the keyword and corresponds to the associative word group and the search result. The cluster ID is stored in the associative word group DB 600. Further, as described above, the advertisement DB 800 stores a pair of advertisements and cluster IDs matching the advertisements.

  In this state, when a search keyword is input to the search web page from several persons and the search keyword is transmitted to the associative word group search device 100 via the Internet or the like, the associative word group search device 100 As in the case described above, search the website containing the received search keyword using a search engine, etc., cluster the search results, create a search result page for each cluster, and search The result is output to matching device 700 together with the cluster ID corresponding to the result page.

  When the matching device 700 inputs a search result page that is the content output from the associative word group search device 100, the matching device 700 reads out an advertisement matching the cluster ID corresponding to the search result page from the advertisement DB 800, and the search result After applying to the advertising space of the page, processing for transmitting to the searcher via the Internet or the like is performed.

  As a result, the searcher can obtain a search result page that is clustered based on the content of the website and matches the content. Therefore, for example, when the searcher inputs the search keyword “○ × △”, at least (1) the URL of the website describing the content related to the juice “○ × Δ” and the juice “○ A search result page on which an advertisement related to “× △” is posted is obtained, and (2) the URL or the like of the website describing the content related to the magazine “OxΔ” and the magazine “OxΔ” A search result page with advertisements will be obtained.

It is a block diagram which shows the typical structure of a website search support system provided with the website search support apparatus 100 which is the associative word group search apparatus of Embodiment 1 of this invention. FIG. 2 is a schematic explanatory diagram of an operation of the website search support apparatus 100 shown in FIG. 1. It is a flowchart which shows operation | movement of the website search assistance apparatus 100 shown in FIG. It is explanatory drawing of the classification method of the associative word group by the classification means 85 of FIG. It is explanatory drawing of the classification method of the associative word group by the classification means 85 of FIG. It is a general | schematic explanatory drawing of the content match type | mold advertisement system 1000 of Embodiment 2 of this invention. It is a figure which shows the example of data stored in the associative word group DB600 shown in FIG. It is a schematic explanatory drawing of the content match type | mold advertisement system 1000 of Embodiment 3 of this invention. It is outline | summary explanatory drawing of the content match type | mold advertisement system 1000 of Embodiment 4 of this invention.

Explanation of symbols

  Receiving means 10, searching means 20, storing means 25, extracting means 30, reading means 40, executing means 50, selecting means 60, obtaining means 70, filtering means 80, classification means 85, creating means 90, transmitting means 95, website Search support apparatus 100, PC 200, DB 300, engine 400, network 500

Claims (8)

A search means for searching a group of websites including the input keyword;
Execution means for executing a process of dividing text data included in some websites of the website group searched by the search means;
An associative word group search device comprising: an output unit that outputs words having a frequency of appearance on each of the websites greater than or equal to a predetermined number of words obtained by the dividing process of the execution unit as an associated word group of the keyword.   The associative word group search device according to claim 1, further comprising a filter unit configured to filter the word based on the word having the predetermined number of appearances and the keyword. The filter means includes
The number of hits of websites that include both the word and the keyword whose appearance count is greater than or equal to a predetermined number,
The number of hits on the website containing the keyword,
The associative word group search device according to claim 2, wherein the words are filtered based on a statistical process using the number of hits of a website including words having a predetermined number of occurrences.   The associative word group search device according to claim 3, wherein the statistical processing includes processing based on a mutual information amount or a t-score.   The associative word group search device according to claim 1, further comprising: an exclusion unit configured to exclude a word when the number of hits of a website including the word whose appearance count is equal to or greater than a predetermined value is equal to or less than a predetermined threshold.   The associative word group search apparatus according to claim 1, further comprising a classifying unit that classifies the associative word group. The associative word group search device according to any one of claims 1 to 6,
An associative word group search system comprising: a storage medium in which an associative word group output from the associative word group search device is stored. The associative word group search device according to any one of claims 1 to 6,
An associative word group storage medium in which an associative word group output from the associative word group search device is stored together with a unique identifier;
An advertisement storage medium in which an advertisement that can be placed in content having an advertisement space and the unique identifier corresponding to the advertisement are stored;
The unique identifier corresponding to the content is determined by matching the associated word group stored in the associative word group storage medium with the content, and any one of the advertisements to which the unique identifier is assigned is stored in the advertisement. A content-matching advertisement system comprising a matching device that reads from a medium.