JP2010186368A - Method and apparatus for determining advertisement - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To present an appropriate advertisement according to a situation by combining a plurality of elements. <P>SOLUTION: The method includes: a step which detects an access to a particular site by a particular user terminal; a step which reads a response value of an instance of a plurality of element classifications corresponding to the access, from a storage section for storing the response value which indicates a tendency for each element classification to form a potential reason for advertisement selection in each instance of the plurality of element classifications showing potential reasons for advertisement selection, and calculates a combined response value of each element classification with respect to the access; a step which divides the number of advertisements capable of being output in response to the access and being contained in a web page of a specified site by the combined response value of each element classification and decides the number of advertisement for each element classification; and a step which calculates an evaluation value of a combination with an advertisement to each element classification on a keyword basis and specifies high rank advertisements in evaluation value by the number of advertisements of each element classification. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a technique for determining an advertisement to be distributed.

  Various sites are opened on the Internet, and various data are made public. In order to search for desired information from the enormous amount of data disclosed on the Internet in this way, it is usual to search with a search engine using a keyword at present. At this time, information provision such as an advertisement may be provided in conjunction with the keyword. However, there are cases where the user's desired result cannot be obtained with only the input keyword. Further, there is a concept of collecting characteristic data of a user by various methods and using it for providing information to the user. For example, Japanese Patent Laid-Open No. 2002-251408 discloses information browsing by an appropriate guide. An information browsing support apparatus capable of supporting the above is disclosed. Specifically, a time-series list of document information requested by the user from the server unit is managed by the access history management unit, and the time-series list is detailed from general information while referring to keywords assigned to the document information. The analysis is divided into the normal behavior that the user refers to the information and the change behavior that refers to other information, and the browsing behavior of the user is supported based on the analysis result.

JP 2002-251408 A

  In the prior art, there are those that select an advertisement according to the characteristics of the user as described above, or select an advertisement according to the characteristics of the site where the advertisement is posted. However, the user does not always select an advertisement according to the characteristics of the user, and does not select an advertisement according to the characteristics of the advertisement posting site. That is, a necessary advertisement is selected according to the situation, but an advertisement appropriate for the situation cannot be presented by only one element such as a user or a site.

  Accordingly, an object of the present invention is to provide a technique for combining a plurality of elements so that an appropriate advertisement corresponding to a situation can be presented.

  The advertisement determination method according to the present invention includes a step of detecting access to a specific site by a terminal of a specific user, and a plurality of element types (for example, user, site, time, etc.) that serve as a potential basis for advertisement selection. For each instance (for example, user A, site B, time division C, etc.), a response data storage unit that stores a reaction value that represents a tendency that each element type is a potential basis for advertisement selection is specified by a specific user terminal. Read the reaction value for each element type instance that corresponds to the access to the site, calculate the total reaction value for each element type for the access, and the specific output that is output according to the access The number of ads for each element type is divided by the total reaction value of each element type, and the number of advertisements that can be included in the web page of The keyword and score for each instance of the element type to which the above-mentioned access corresponds are read out from the element data storage unit that stores the relevant keyword and score for each instance of the plurality of element types, and the read keyword The evaluation value for the combination with the advertisement for each element type is calculated according to a predetermined evaluation formula from the keyword and the score for each advertisement in the advertisement data storage unit that stores the advertisement data, the keyword, and the score for each advertisement. Then, for each element type, a step of specifying an advertisement having a higher evaluation value for the number of advertisements, and an output step of reading out and outputting the advertisement data of the specified advertisement from the advertisement data storage unit are included.

  In this way, for each of the plurality of element types, by using the total reaction value as defined above and selecting the advertisement to be output by distributing the number of advertisements that can be displayed, a more appropriate advertisement can be obtained. The possibility of being presented by the access source user increases.

  The plurality of element types described above may include users and sites. At that time, the reaction data storage unit described above includes a reaction rate indicating a tendency for the user to select an advertisement according to his / her preference and a reaction rate indicating a tendency to select an advertisement according to the contents of the access destination site. A user response rate vector including at least a user instance is stored for each user instance, and further, a response rate indicating a tendency that an advertisement placed on the site is selected according to the user's preference and the content of the site is selected. A site response rate vector including at least a response rate indicating a tendency to be stored may be stored for each site instance.

  Thus, if both the user characteristics and the site characteristics are taken into consideration, an appropriate advertisement can be output.

  Further, the plurality of element types described above may further include time. In this case, the user response rate vector further includes a response rate indicating a tendency to select an advertisement according to time, and the site response rate vector further indicates a response rate indicating a tendency to select an advertisement according to time. May be further included. The reaction data storage unit further represents, for each time segment, a reaction rate indicating a tendency for the user to select an advertisement according to his / her preference and a reaction rate indicating a tendency to select an advertisement according to the contents of the access destination site. And a time response rate vector including a response rate indicating a tendency to select an advertisement according to time may be stored. In this way, it is possible to present advertisements more effectively by taking into account the temporal characteristics in addition to the user characteristics and site characteristics.

  Further, the overall response value calculating step described above reads a user response rate vector corresponding to a specific user and a site response rate vector corresponding to a specific site from the response data storage unit, and a user response rate A step of calculating an average vector of the vector and the site reaction rate vector may be included. In that case, you may use each component value of an average vector as a total reaction value of each element classification. In this way, an appropriate distribution ratio can be calculated.

  Further, in the output step described above, an identifier representing the corresponding element type may be added to the advertisement data of the specified advertisement. In this case, in the present invention, the identifier of the specific user and the identification information of the specific site transmitted by selecting the specific advertisement among the advertisements included in the web page of the specific site displayed on the terminal of the specific user And the element type data added to the advertisement data of the specific advertisement from the terminal of the specific user, a user response rate vector for the specific user, a site response rate vector for the specific site, May further include a step of updating according to the received element type data. In this way, the reaction rate vector can be appropriately updated in accordance with the advertisement selection, and a more appropriate advertisement can be extracted.

  Furthermore, in the present invention, the keyword and score related to each user stored in the element data storage unit are updated based on the keyword specified in at least one of each user's web page browsing and web page search. A step may be further included. By doing so, more appropriate advertisements can be extracted.

  The method according to the present invention may be implemented by a combination of computer hardware and a program. The program according to the present invention is stored in, for example, a flexible disk, a CD-ROM, a magneto-optical disk, a semiconductor memory, a hard disk, or the like. Stored in a medium or storage device. Moreover, it may be distributed as a digital signal via a network or the like. The intermediate processing result is temporarily stored in a storage device such as a main memory.

  ADVANTAGE OF THE INVENTION According to this invention, a suitable advertisement according to a condition can be shown now combining several elements.

FIG. 1 is a system outline diagram in the embodiment of the present invention. FIG. 2 is a diagram showing a processing flow of site information generation processing. FIG. 3 is a diagram illustrating an example of the site profile table. FIG. 4 is a diagram illustrating a process flow of the user information generation process. FIG. 5 is a diagram illustrating an example of a user profile table. FIG. 6 is a diagram illustrating an example of user response rate vector data. FIG. 7 is a diagram illustrating an example of site reaction rate vector data. FIG. 8 is a diagram showing a processing flow of information table generation processing. FIG. 9 is a diagram illustrating an example of the information table. FIG. 10 is a diagram illustrating a processing flow of advertisement display processing. FIG. 11 is a diagram illustrating a processing flow of advertisement display processing. FIG. 12 is a diagram illustrating an example of an advertisement display screen displayed on the user terminal. FIG. 13 is a diagram illustrating a processing flow of a response rate vector learning process. FIG. 14 is a diagram illustrating an example of a season profile table. FIG. 15 is a functional block diagram of a computer.

  In the embodiment of the present invention, when a “specific user” visits a “specific site” (possibly at a “specific time”), a matching method to be adopted (for example, according to the characteristics of the user) A user match UM that performs matching, a content match CM that performs matching according to the characteristics of the content, a season match SM that performs matching according to a time zone, and the like, and numerical values related to the degree thereof are specified. And the effect of an advertisement is improved by extracting a concrete advertisement based on the numerical value regarding the said matching method and its degree, and showing it to a user.

  For example, if a user is obsessed with his / her own interests, he / she distributes only what has a high response rate for the user regardless of what site and what time. In addition, in a diary site such as a blog, information based on content matching is not distributed from the hair color of the site, but advertisements are automatically extracted according to user characteristics or timed and distributed. In the present embodiment, this can be realized.

  An outline of a computer system for this purpose is shown in FIG. For example, in a network 1 such as the Internet, a plurality of user terminals 3 on which a web (Web) browser is executed, a plurality of Web servers 9, a plurality of advertiser terminals 7, and main processing in the present embodiment are performed. The advertisement distribution system 5 is connected.

  The advertisement distribution system 5 has a search function and various information providing functions for the user terminal 3 via the network 1 such as a portal site, for example, and a web page or a user request for displaying a search result, for example. And an advertisement processing function for extracting an advertisement to be output to a Web page for displaying information provided in response to the request. Although the advertisement distribution system 5 can be composed of one or a plurality of servers, it will be described here as being implemented by a single computer.

  The advertisement distribution system 5 includes, for example, a Web server unit 55 serving as an interface with the user terminal 3, a search processing unit 58 that provides a search function to the user terminal 3, and content storage that stores various types of information provided to the user terminal 3. Information element generation that generates data called information elements representing characteristics of each user, each site, etc. based on information on the access to the web server unit 55 from the unit 56 and the user server 3 and text data of various sites, etc. Unit 53, an information element DB 54 that stores information element data that is a processing result of the information element generation unit 53, an advertisement information registration unit 60 that receives registration of advertisement data from the advertiser terminal 7 and generates necessary additional data An information table storage unit 61 for storing an information table for registering data generated by the advertisement information registration unit 60 , Using the data stored in the information element DB 54 and the information table storage unit 61, an advertisement determination processing unit 59 that performs a process of determining an advertisement to be output, and transmitted according to the advertisement selection in the user terminal 3 Based on the data, the response rate vector learning unit 57 that updates the response rate vector that is stored in the information element DB 54 and represents the expected trend of advertisement distribution, and collects text data of various Web pages from the Web server 9 and the like And a Web page DB 52 that stores text data of various Web pages collected by the crawler 51.

  A user who uses the advertisement distribution system 5 is notified of the user ID to the advertisement distribution system 5 by a cookie. For example, when accessing the Web server 9 or the like, the access destination URL may be notified by a cookie when the advertisement distribution system 5 is accessed, for example. In such a case, it is assumed that the user of the user terminal 3 has agreed to provide information to the advertisement distribution system 5.

  Next, processing for accumulating data in the information element DB 54 as preprocessing and data to be accumulated will be described.

  First, generation of a site profile table will be described. The crawler 51 acquires the text of the site by crawling various sites on the Web server 9 connected to the network 1, and stores it in the Web page DB 52 together with the URL (FIG. 2: step S1). Then, the information element generation unit 53 extracts one or more sets of keywords and scores for each site from the text data of each site stored in the Web page DB 52 and the content storage unit 56 by a known TFIDF method or the like. The content ID, URL (Uniform Resource Locator) and additional data are registered in the site profile table of the information element DB 54 (step S3).

  The site profile table is, for example, a table as shown in FIG. In the example of FIG. 3, a content ID, a keyword, a score, a site title, and a site URL are registered. In the site profile table, data used when generating a user profile described below and site profile data of a site on which an advertisement is posted are registered together.

  Next, generation of a user profile table will be described. For example, the Web server unit 55 receives access from the user terminal 3 together with a cookie including the user ID (step S11). If the access is related to search, the Web server unit 55 outputs the user ID and the search keyword to the information element generation unit 53. Further, the Web server unit 55 outputs a search instruction including a search keyword to the search processing unit 58. Since the contents of the search process and the output of the search results are the same as before, the description thereof is omitted here.

  Upon receiving the user ID and the search keyword, the information element generation unit 53 assigns a score “1” to the search keyword (step S13). Then, the user profile table in the information element DB 54 is updated with the pair of the keyword and the assigned score for the user ID (step S19). If no keyword is registered corresponding to the user ID, the keyword and assigned score are registered in the user profile table corresponding to the user ID. On the other hand, when a keyword has already been registered corresponding to the user ID, for example, an assigned score is added to the already registered score to calculate a new score, and update registration is performed. And this process is complete | finished.

  For example, a user profile table as shown in FIG. 5 is generated. In the example of FIG. 5, a user ID, a keyword, and a score are registered. Usually, a plurality of keywords are registered for one user.

  When the access is not related to the search, the Web server unit 55 includes the user ID, the access destination URL (for example, the access URL for the content in the content storage unit 56) and the cookie. The access history (URL list) is output to the information element generation unit 53. When the information element generation unit 53 receives the user ID and the URL, the information element generation unit 53 extracts a keyword and a score registered corresponding to the URL from the site profile table (step S15). Then, a weight c (for example, 0.2) × score is assigned to each keyword (step S17). Then, control goes to a step S19. Since the search can obtain the keyword that the user has intentionally specified, a score of “1” is assigned. However, since it is not possible to know which one of the keywords is intended by browsing alone, it is multiplied by a lower weight. To use.

  Next, user response rate vector data and site response rate vector data registered in the information element DB 54 will be described.

  The user response rate vector is a UM (User Match) response rate indicating the tendency of the user to select an advertisement according to his / her preference, and a CM (Contents Match) indicating a tendency of selecting an advertisement according to the contents of the access destination site. Reaction rate. For example, data as shown in FIG. 6 is registered in the information element DB 54. The sum of the vector component values is “1”, and is normalized.

  The site response rate vector includes a UM response rate that represents a tendency for advertisements to be placed on the site to be selected according to user preferences and a CM response rate that represents a tendency to be selected according to the contents of the site. . For example, data as shown in FIG. 7 is registered in the information element DB 54. The sum of the vector component values is “1”, and is normalized.

  In addition, when the information element generation unit 53 receives the new user ID from the Web server unit 55, the information element generation unit 53 converts the data of the user reaction rate vector having the same component values (0.5 in the case of UM and CM) to the information Register in the element DB 54. Further, when a URL is newly set as an advertising destination, site reaction rate vector data having the same component values (0.5 in the case of UM and CM) are registered in the information element DB 54. .

  As will be described below, when the user selects an advertisement, depending on whether the selected advertisement is an advertisement selected based on the user match UM or an advertisement selected based on the content match CM The user response rate vector and the site response rate vector are updated.

  Furthermore, the process which produces | generates the information table which is a data table about an advertisement is demonstrated using FIG. The advertisement information registration unit 60 receives advertisement data including the advertisement site URL from the advertiser terminal 7 and stores it in a storage device such as a main memory (step S41). The advertisement data may include data such as unit price, for example. Then, the advertisement information registration unit 60 acquires the text of the corresponding site from the site URL (step S43), and extracts keywords and scores by the well-known TFIDF method or the like (step S45). Then, the advertisement data and one or more pairs of keywords and scores are registered in the information table storage unit 61 (step S47).

  For example, an information table as shown in FIG. 9 is registered in the information table storage unit 61. In the example of FIG. 9, an advertisement ID, a keyword, a score, and advertisement data (for example, an advertisement site title and an advertisement site URL) are registered. Other advertisement data may be registered.

  Next, processing when displaying an advertisement will be described with reference to FIGS. The Web server unit 55 receives access to a specific URL together with a cookie including the user ID from the user terminal 3 (step S51). If it does so, it will be judged whether specific URL is URL of the advertisement insertion place of the advertisement delivery system 5 (step S53). If the advertisement distribution destination of the advertisement distribution system 5 is not an access destination, data to be distributed is extracted from the content storage unit 56, for example, as usual, to generate Web page data, and returns it to the user terminal 3 (step) S55). Then, the process ends.

  On the other hand, when the specific URL is the URL of the advertisement posting destination of the advertisement distribution system 5, the Web server unit 55 outputs the received user ID and the specific URL that is the access destination URL to the advertisement determination processing unit 59. To do. For example, at this time, the Web server unit 55 notifies the advertisement determination processing unit 59 of data on the number of advertisements that can be posted on a Web page with a specific URL as an advertisement placement destination. Note that the number of advertisements that can be posted may be determined in advance for each Web page or each site type, for example. In that case, the number of advertisements that can be posted is read from, for example, a table.

  When receiving the user ID and URL from the Web server unit 55, the advertisement determination processing unit 59 reads the user response rate vector corresponding to the user ID and the site response rate vector corresponding to the access destination URL from the information element DB 54. (Step S57). Then, the user reaction rate vector and the site reaction rate vector are added and divided by 2, thereby synthesizing the reaction rate vector (step S59). That is, the average of the vector component values is calculated to calculate the average vector. For example, if the user reaction rate vector Ru (u) = (0.6, 0.4) and the site reaction rate vector Rc (c) = (0.4, 0.6), the combined reaction rate vector Becomes (0.5, 0.5) (= ((0.6 + 0.4) / 2, (0.4 + 0.6) / 2).

  Then, the advertisement determination processing unit 59 determines the number of advertisement placements by the user match UM and the number of advertisement placements by the content match CM based on the combined reaction rate vector (step S63). Specifically, the first component (UM component) × total number of advertisements in the combined reaction rate vector and the second component (CM component) × total number of advertisements are calculated. In the case where the total number of advertisements is “6” and the response rate vector after synthesis described above, the number of advertisements posted by user match UM is 0.5 × 6 = 3, and the number of advertisements posted by content match CM is 0.5 × 6 = 3. For example, when the number does not become an integer, one may be prioritized and rounded up to an integer, and the rest may be set to the other number. The processing shifts to the processing in FIG.

  In this way, the result of comprehensive consideration of the characteristics indicating the advertisement selection tendency of the specific user and the characteristics indicating the advertisement selection tendency of the specific site for the access to the specific site by the specific user. Since it can be obtained from the later response rate vector, if the total number of advertisements is divided by the component value of the response rate vector, it is possible to determine the appropriate advertisement distribution for this access.

  The advertisement determination processing unit 59 selects advertisements whose keywords matching the upper keyword corresponding to the user ID in the user profile table are registered in the information table by the number of advertisements placed by the user match UM (step S65). For example, in the user profile table, the evaluation value is calculated by the product sum of the corresponding score of the upper keyword corresponding to the user ID and the corresponding score of the same keyword of the advertisement registered in the information table, and the evaluation value is higher The advertisement is selected by the number of advertisements placed by the user match UM. As such an advertisement selection method, various conventionally used methods can be employed.

  Further, the advertisement determination processing unit 59 selects, as many advertisements as the number of advertisements placed by the content match CM, in the site profile table in which keywords that match the upper keyword corresponding to the access destination URL are registered in the information table (step S1). S67). For example, in the site profile table, an evaluation value (a score vector between the score vectors for the same keyword) is calculated by multiplying the corresponding score of the upper keyword corresponding to the access destination URL and the corresponding score of the same keyword of the advertisement registered in the information table. The inner product (the value for the keyword is 0 when there is no corresponding keyword) is calculated, and the advertisement with the highest evaluation value is selected by the number of advertisements placed by the site match CM. As such an advertisement selection method, various conventionally used methods can be employed. For example, a cosine value obtained by dividing the inner product of score vectors as described above by the length of both score vectors may be adopted as the evaluation value.

  The advertisement determination processing unit 59 reads out the advertisement data (eg, title, URL, matching type (user match UM identifier or content match CM identifier)) selected in steps S65 and S67 from the information table, and the Web server unit 55 Output to.

  For example, the Web server unit 55 reads out the data about the access destination URL from the content storage unit 56 and uses the advertisement data received from the advertisement determination processing unit 59 to display the Web page data including the data about the access destination URL and the advertisement data. Is transmitted to the user terminal 3 (step S69). In this Web page, for example, when an advertisement title is clicked and the advertisement title is clicked, an access destination URL or the like (which may be an ID of the access destination site) and a matching type (user match) are displayed. UM identifier or content match CM identifier) and a link destination URL that redirects to the actual advertisement site URL.

  By performing such processing, in the example described above, a Web page as shown in FIG. 12 is displayed on the user terminal 3. In the example of FIG. 12, for example, three advertisements 1001 based on the user match UM and three advertisements 1002 based on the content match CM are included in the lower part of the normal content display portion of the site. When the user selects an advertisement based on the user match UM, the URL or ID of the site accessed this time and the identifier of the user match UM are transmitted from the web browser of the user terminal 3 to the advertisement distribution system 5. When the user selects an advertisement based on the content match CM, the URL or ID of the site accessed this time and the identifier of the user match UM are transmitted from the Web browser of the user 3 to the advertisement distribution system 5.

  As described above, the advertisement based on the user match UM and the advertisement based on the content match CM are selected and distributed at an optimal distribution ratio for the current access. The allocation ratio is determined by the user response rate vector and the site response rate vector as described above. Therefore, an optimal distribution ratio cannot be obtained unless the user response rate vector and the site response rate vector are set appropriately.

  Next, the learning process of the user response rate vector and the site response rate vector will be described with reference to FIG. As shown in step S69 of FIG. 11, the Web server unit 55 of the advertisement distribution system 5 reads out the data about the access destination URL from the content storage unit 56 and uses the advertisement data received from the advertisement determination processing unit 59, for example. Web page data including data on the access destination URL and advertisement data (advertisement, advertisement URL, corresponding matching type, etc.) is generated and transmitted to the user terminal 3 (step S21). The browser of the user terminal 3 receives the Web page data including the data about the access destination URL and the advertisement data (advertisement, advertisement URL, corresponding matching type, etc.) and displays it on the display device (step S23). The display as shown in FIG. 12 is made.

  Here, it is assumed that the user selects and clicks one of the advertisements. On the other hand, the browser of the user terminal 3 accepts the click of the advertisement by the user (step S25), and the cookie including the user ID, the access destination URL or ID (the URL of the site displayed in step S23), and the corresponding matching The type is transmitted to the advertisement distribution system 5 (step S27).

  The web server unit 55 of the advertisement distribution system 5 receives the cookie including the user ID, the access destination URL or ID, and the corresponding matching type from the user terminal 3 (step S29), and outputs them to the reaction rate vector learning unit 57. To do. In addition, the user terminal 3 returns a predetermined response to be redirected to the selected advertisement site. Since the subsequent steps are the same as usual, the description of the operation of the user terminal 3 is omitted.

  When the response rate vector learning unit 57 receives the data described above from the Web server unit 55, the response rate vector learning unit 57 registers the received data in a log data storage unit (not shown) and generates a reaction vector r by the corresponding matching type (step S31). ). The response vector r is (1, 0) if the corresponding matching type is a user match UM, and (0, 1) if the corresponding match type is a content match CM.

  Then, the reaction rate vector learning unit 57 updates the user reaction rate vector specified by the user ID with the reaction vector r and registers it in the information element DB 54 (step S33). Update with Ru (u) = (Ru (u) + L · r) / | Ru (u) + L · r |. L is a predetermined learning coefficient.

  further. The reaction rate vector learning unit 57 updates the site reaction rate vector specified by the access destination URL with the reaction vector r and registers it in the information element DB 54 (step S35). Update with Rc (c) = (Rc (c) + L · r) / | Rc (c) + L · r |.

  In this way, each component is enhanced according to the current advertisement selection result.

  As described above, learning progresses so as to obtain a more appropriate distribution ratio by updating the reaction rate vector in response to feedback of data indicating what kind of advertisement the user has selected.

  In the above-described embodiment, two cases of the user and the site have been described as the element types that serve as a potential basis for the advertisement selection. However, there are other element types that can be adopted. It's time as I mentioned above.

  When considering time, for each time segment, a reaction rate indicating the tendency of the user to select an advertisement according to his / her preference, and a reaction rate indicating a tendency to select an advertisement according to the contents of the access destination site, Introduce a season reaction rate vector that includes, as component values, a reaction rate indicating a tendency to select an advertisement according to time. As the time division, for example, a predetermined time, such as year, month, day, day of the week, hour, minute, is employed.

  Note that there are usually a plurality of season response rate vectors corresponding to the time when access from the user terminal 3 is detected. That is, if a season response rate vector is defined for each time segment of year, month, day, day of the week, hour, and minute, the corresponding season response rate vector is obtained for all. Therefore, an average vector is calculated before combining with other reaction rate vectors.

  For other reaction rate vectors, it is necessary to further set a reaction rate representing a tendency to select an advertisement according to time as a component value.

  Furthermore, it is necessary to prepare a season profile table in the information element DB 54. Specifically, for example, data as shown in FIG. 14 is prepared. In the example of FIG. 14, a season ID, a keyword, a score, a title, and a period are registered. There are various methods for generating such a season profile table. For example, keywords and scores are extracted by TFIDF or the like from TV program guide pages, event information pages, and other pages that can be associated with time and duration, and corresponding times and durations are associated with each other.

  If such data is prepared, processing using two reaction rate vectors may be performed using two reaction rate vectors. For example, the apportioned number of advertisements is also divided into two parts. Thereafter, more preferable advertisements for each of the user match UM, the content match CM, and the season match SM are extracted. The reaction rate vector learning process may be performed in the same manner.

  If the season match SM is introduced, for example, products that have become a hot topic in a television broadcast program can be automatically extracted and distributed intensively after broadcasting.

  Note that the three element types do not necessarily have to be used simultaneously. Two or more element types may be used in any combination. Further, it is possible to set so as not to use any reaction rate vector for a special site or user.

  There may be a case where another element type is employed. For example, when the user's position can be acquired, a position response rate vector and a position profile table may be prepared.

  Although the embodiment of the present invention has been described above, the present invention is not limited to this. As described above, the functional block diagram of FIG. 1 is an example, and in particular, the advertisement distribution system 5 is realized by a plurality of computers, and the data stored in the information element DB 54 is separately processed and prepared. May be. Furthermore, the advertisement distribution destination is described in the case of its own site, but can be applied to the case of other sites.

  As for the processing flow, as long as the processing result does not change, the order may be changed or executed in parallel.

  The advertisement distribution system 5, the web server 9, the user terminal 3, and the advertiser terminal 7 are computer devices, and as shown in FIG. 15, a memory 2501, a CPU 2503, a hard disk drive (HDD) 2505, and a display device 2509. A display control unit 2507 connected to the computer, a drive device 2513 for a removable disk 2511, an input device 2515, and a communication control unit 2517 for connecting to a network are connected by a bus 2519. An operating system (OS) and an application program for executing the processing in this embodiment are stored in the HDD 2505, and are read from the HDD 2505 to the memory 2501 when executed by the CPU 2503. If necessary, the CPU 2503 controls the display control unit 2507, the communication control unit 2517, and the drive device 2513 to perform necessary operations. Further, data in the middle of processing is stored in the memory 2501 and stored in the HDD 2505 if necessary. In an embodiment of the present technology, an application program for performing the above-described processing is stored in a computer-readable removable disk 2511 and distributed, and installed from the drive device 2513 to the HDD 2505. In some cases, the HDD 2505 may be installed via a network such as the Internet and the communication control unit 2517. Such a computer device realizes various functions as described above by organically cooperating the hardware such as the CPU 2503 and the memory 2501 described above with the OS and necessary application programs.

1 Network 3 User Terminal 5 Advertisement Distribution System 7 Advertiser Terminal 9 Web Server 51 Crawler 52 Web Page DB
53 Information Element Generation Unit 54 Information Element DB
55 Web server unit 56 Content storage unit 57 Reaction rate vector learning unit 58 Search processing unit 59 Advertisement determination processing unit 60 Advertisement information registration unit 61 Information table storage unit

Claims (8)

Detecting access to a specific site by a specific user's device;
Identification by the specific user terminal from a reaction data storage unit that stores a reaction value indicating a tendency that each element type is a potential basis for advertisement selection for each instance of a plurality of element types that are a potential basis for advertisement selection An overall reaction value calculating step of reading out a reaction value for each instance of the element type corresponding to an access to the site, and calculating an overall reaction value of the element type for the access;
The number of advertisements that can be included in the web page of the specific site that is output in response to the access is divided by the total reaction value of each element type, and the number of advertisements for each element type is determined. Steps,
Read the keyword and score for each instance of the element type to which the access corresponds, from the element data storage unit that stores the relevant keyword and score for each instance of the plurality of element types, and the read keyword and score Calculating an evaluation value for a combination with the advertisement for each element type according to a predetermined evaluation formula from the keyword and score for each advertisement in the advertisement data storage unit that stores the advertisement data, keyword, and score for each advertisement Identifying an advertisement having a higher evaluation value by the number of advertisements for each of the element types;
An output step of reading out and outputting the advertisement data of the identified advertisement from the advertisement data storage unit;
An advertisement determination method for causing a computer to execute. The plurality of element types include users and sites,
The reaction data storage unit
A user response rate vector including at least a response rate indicating a tendency of a user to select an advertisement according to his / her preference and a response rate indicating a tendency of selecting an advertisement according to the contents of the access destination site is stored for each user instance. And
Further, a site response rate vector including at least a reaction rate indicating a tendency that an advertisement placed on the site is selected according to the user's preference and a reaction rate indicating a tendency to be selected according to the content of the site. The advertisement determination method according to claim 1, wherein the advertisement determination method is stored for each instance. The plurality of element types further includes time,
The user response rate vector further includes a response rate representing a tendency to select advertisements according to time,
The site response rate vector further includes a response rate representing a tendency for ads to be selected according to time,
The reaction data storage unit further includes:
For each time segment,
A reaction rate indicating a tendency for a user to select an advertisement according to his / her preference, a reaction rate indicating a tendency to select an advertisement according to contents of an access destination site, and a reaction rate indicating a tendency to select an advertisement according to time The advertisement determination method according to claim 2, further comprising: storing a time response rate vector including: The total reaction value calculating step includes:
Reading a user response rate vector corresponding to the specific user and a site response rate vector corresponding to the specific site from the response data storage unit;
Calculating an average vector of the user response rate vector and the site response rate vector;
Including
The advertisement determination method according to claim 2, wherein each component value of the average vector is a total reaction value of each of the element types. In the output step,
An identifier representing the corresponding element type is added to the advertisement data of the specified advertisement,
The identifier of the specific user, the identification information of the specific site, and the identification information transmitted by selecting a specific advertisement among the advertisements included in the web page of the specific site displayed on the terminal of the specific user, Receiving the data of the element type added to the advertisement data of the specific advertisement from the terminal of the specific user;
Updating a user response rate vector for the specific user and a site response rate vector for the specific site according to the data of the element type;
The advertisement determination method according to claim 2, further comprising: The method further comprises: updating a keyword and a score related to each user stored in the element data storage unit based on a keyword specified in at least one of a web page browsing and a web page search of each user. 1. The advertisement determination method according to 1.   A program for causing a computer to execute the advertisement determination method according to any one of claims 1 to 6. A reaction data storage unit that stores a reaction value indicating a tendency that each element type is a potential basis for selecting an advertisement for each instance of a plurality of element types that are a potential basis for selecting an advertisement;
An element data storage for storing keywords and scores related to each instance of the plurality of element types;
An advertisement data storage unit that stores advertisement data, keywords, and scores for each advertisement; means for detecting access to a specific site by a specific user terminal;
The reaction value for each instance of the element type corresponding to access to a specific site by the specific user terminal is read from the reaction data storage unit, and the total reaction value of each element type is calculated for the access. Means,
The number of advertisements that can be included in the web page of the specific site that is output in response to the access is divided by the total reaction value of each element type, and the number of advertisements for each element type is determined. Means,
From the element data storage unit, the keyword and score for each instance of the element type to which the access corresponds are read out, and from the read keyword and score and the keyword and score for each advertisement in the advertisement data storage unit Means for calculating an evaluation value for a combination with an advertisement for each of the element types according to a predetermined evaluation formula, and specifying an advertisement having a higher evaluation value for the number of advertisements for each of the element types;
Output means for reading out and outputting advertisement data of the specified advertisement from the advertisement data storage unit;
An advertisement determining device.

Images