JP5577385B2 - Content distribution device - Google Patents

Description

  The present invention relates to a technique for distributing advertising content via a network such as the Internet.

  In Internet homepages and the like, there are often advertisement columns for display advertisements typified by banner advertisements and search-linked advertisements.

  The search-linked advertisement is a method in which an advertiser registers keywords in advance in association with advertisements and distributes corresponding advertisements when the search keyword input by the user matches the registered keywords.

  In such a search-linked advertisement, the keyword registered in advance greatly affects the advertising effect, so the advertiser has a hard time selecting the keyword.

  For this reason, there is an interface in which an advertiser inputs / sets a keyword is provided with a function of presenting (suggesting) a keyword that serves as a reference (see Patent Document 1).

Japanese Patent No. 4820888

  As described above, keywords that are useful for entering and setting advertiser keywords have been presented in the past, but a linguistic approach such as presenting synonyms for keywords entered by advertisers. It was only taken and was not customized according to the characteristics of the advertiser. For this reason, the keywords presented by the system are not necessarily meaningful in terms of enhancing the advertising effect.

  The present invention has been proposed in view of the above-mentioned conventional problems, and the object of the present invention is to present effective keywords for enhancing the advertising effectiveness to advertisers of search-linked advertising. is there.

  In order to solve the above problems, in the present invention, a means for selecting a target advertisement to be a target of a search-linked advertisement, and a user who has viewed an advertisement of an advertiser of the target advertisement from log data are specified, Means for extracting a search keyword input by the user, means for calculating a score for each search keyword based on an input frequency of the extracted search keyword, and receiving an input of a search-linked advertisement keyword for the target advertisement; Means for presenting a search keyword having a high score as a candidate.

  In the content distribution apparatus of the present invention, it is possible to present an effective keyword for enhancing the advertising effect to the advertiser of the search-linked advertisement.

It is a figure which shows the structural example of the system concerning one Embodiment of this invention. It is a flowchart which shows the process example of embodiment. It is a figure which shows the data structure example of an advertisement database. It is a figure which shows the example of log data. It is a figure which shows the example of extraction of click user information and non-click user information. It is a figure (the 1) which shows the example of calculation of a keyword score, and extraction of a presentation candidate keyword. It is FIG. (2) which shows the example of calculation of a keyword score, and extraction of a presentation candidate keyword. It is FIG. (3) which shows the example of calculation of a keyword score, and extraction of a presentation candidate keyword. It is a figure which shows the example of calculation of AUC.

  Hereinafter, preferred embodiments of the present invention will be described.

<Configuration>
FIG. 1 is a diagram showing a configuration example of a system according to an embodiment of the present invention.

  In FIG. 1, the system provides a plurality of terminal devices 2 operated by a general user, a plurality of terminal devices 3 operated by an advertiser user, and an information providing service for a general user using the terminal device 2. A plurality of service server devices 4 and an advertisement management device 5 that receives advertisements and accepts keywords for search-linked advertisements from an advertiser user who operates the terminal device 3, and distributes advertisements. Or the like. Note that the advertisement management device 5 may have a path connected directly or indirectly (including offline data exchange) with the service server device 4.

  The terminal devices 2 and 3, the service server device 4, and the advertisement management device 5 have a hardware configuration of a general computer device.

  The advertisement management device 5 is a log acquisition unit 501, an advertisement selection unit 503, a click user information extraction unit 505, a non-click user information extraction unit 507, as functional units mainly realized by software (computer program). A keyword score calculation unit 509, a presentation candidate keyword extraction unit 511, an advertisement submission reception unit 514, a keyword reception unit 515, a keyword presentation unit 516, and an advertisement distribution unit 517 are provided.

  The log acquisition unit 501 has a function of acquiring log data 502 from the service server device 4 online or offline.

  The advertisement selection unit 503 has a function of selecting an advertisement that is a target of a search-linked advertisement that is a target to which a presentation candidate keyword is pre-assigned from the advertisement database 513 and outputting an advertiser ID 504 that has submitted the advertisement. Yes.

  The click user information extraction unit 505 analyzes the log data 502, thereby information on the user who clicked the advertisement submitted by the advertiser indicated by the advertiser ID 504 (clicked on the presented (impression) advertisement) ( The search keyword) is extracted from the log data 502 and output as click user information 506.

  The non-click user information extraction unit 507 randomly identifies a user who has not clicked on the advertisement submitted by the advertiser indicated by the advertiser ID 504 (which has not been determined to have been clicked), and the information (mainly, The search keyword is extracted from the log data 502 and output as non-click user information 508. In addition, since the non-click user information 508 is not used in the first processing pattern described later, in this case, the non-click user information extraction unit 507 can be omitted.

  The keyword score calculation unit 509 has a function of calculating a keyword score 510 indicating a search keyword input frequency from the click user information 506 and / or the non-click user information 508.

  The presentation candidate keyword extraction unit 511 extracts, from the keyword score 510, a keyword that is presumed to be frequently input by a user who clicks the advertisement of the advertiser of the target advertisement, and registers it in the advertisement database 513 as the presentation candidate keyword 512. have.

  The advertisement submission receiving unit 514 has a function of receiving an advertisement submission from an advertiser user who operates the terminal device 3 and registering it in the advertisement database 513.

  The keyword accepting unit 515 has a function of accepting a keyword used for a search-linked advertisement with respect to an advertisement already submitted in the advertisement database 513.

  The keyword presenting unit 516 has a function of presenting (suggesting) the presentation candidate keywords 512 set in the advertisement database 513 when the keyword accepting unit 515 accepts keywords from the advertiser.

  When the general user accesses the service server device 4 from the terminal device 2 and performs a search, the advertisement distribution unit 517 searches for the search keyword based on the advertisement distribution request from the service server device 4 with the input search keyword. Has a function of acquiring and distributing advertisements associated with the advertisement database 513.

<Operation>
FIG. 2 is a flowchart showing a processing example of the above embodiment.

  First, in FIG. 2A, the advertisement submission receiving unit 514 of the advertisement management apparatus 5 receives an advertisement submission from the advertiser user who operates the terminal device 3, and registers it in the advertisement database 513 (step S1). S1).

  FIG. 3 is a diagram showing an example of the data structure of the advertisement database 513, and includes items such as “advertisement ID”, “advertiser ID”, “advertisement content (link or entity)”, “registered keyword”, and “presentation candidate keyword”. .

  “Advertisement ID” is information for specifying an advertisement. “Advertiser ID” is information that identifies the advertiser who submitted the advertisement. The “advertisement content (link or entity)” is the content of the advertisement, and may be a link (reference) such as a URL where the actual data is stored, or the actual data itself. “Registered keyword” is a keyword used for a search-linked advertisement set by an advertiser. The “presentation candidate keyword” is a keyword that the system side presents as an effective keyword when a keyword of a search-linked advertisement is received by an advertiser.

  Next, in FIG. 2B, the log acquisition unit 501 of the advertisement management apparatus 5 acquires log data 502 from the service server apparatus 4 online or offline regularly or according to an operator's instruction (step S2).

  FIG. 4 is a diagram illustrating an example of the log data 502, and includes items such as “cookie information”, “user ID”, “date”, and “behavior content (search keyword, browse advertisement ID,...)”.

  “Cookie information” is information that is written in the browser when accessing the service server device 4 from the browser of the terminal device 2, and it is possible to specify that access has been made from the same browser unless the browser cookie is reset. . The “user ID” is information for specifying the user of the terminal device 2 and is specified by logging in when the service server device 4 provides a service. “Date” is a date when the terminal device 2 accesses the service server device 4. “Behavior content (search keyword, browsing advertisement ID,...)” Is a record of behavior content performed when the terminal device 2 accesses the service server device 4.

  Next, in FIG.2 (c), the advertisement selection part 503 of the advertisement management apparatus 5 selects the advertisement used as the object of the search interlocking type | system | group advertisement used as the object to which a presentation candidate keyword is previously given from the advertisement database 513, The said advertisement Is output (step S3). When the processing for one advertisement is completed, the same processing is repeated until there is no processing target.

  Next, the click user information extraction unit 505 of the advertisement management apparatus 5 analyzes the log data 502 to obtain information on the user who clicked the advertisement submitted by the advertiser indicated by the advertiser ID 504 from the log data 502. The extracted click user information 506 is output (step S4).

  Next, the non-click user information extraction unit 507 of the advertisement management apparatus 5 randomly identifies a user who has not clicked on the advertisement submitted by the advertiser indicated by the advertiser ID 504, and stores the information from the log data 502. The non-click user information 508 is extracted and output (step S5).

  FIG. 5 is a diagram illustrating an example of extraction of click user information 506 and non-click user information 508.

  In FIG. 5, the click user information extraction unit 505 identifies a user (click user) that includes the advertisement ID associated with the advertiser ID 504 in the advertisement database 513 in the browsing advertisement ID of the action content from the log data 502, A search keyword input by the user before the predetermined period from the present time is extracted from the action content, and the user ID and / or cookie information, the search keyword, and the number of searches are associated with each other and output as click user information 506 (step S41). ).

  The non-click user information extraction unit 507 extracts users other than the click user (non-click user) randomly (randomly), extracts the search keyword input by the user from the current time before the predetermined period from the action content, The user ID and / or cookie information, the search keyword, and the number of searches are associated with each other and output as non-click user information 508 (step S51).

  Returning to FIG. 2C, the keyword score calculation unit 509 of the advertisement management apparatus 5 calculates the keyword score 510 indicating the input frequency of the search keyword from the click user information 506 and / or the non-click user information 508 (step S6). ).

  Next, the presentation candidate keyword extraction unit 511 of the advertisement management apparatus 5 extracts keywords that are estimated to be frequently input by the user who clicks the advertisement of the target advertiser from the keyword score 510, and uses the advertisement database as the presentation candidate keywords 512. It registers in 513 (step S7).

  6 to 8 are diagrams illustrating an example of calculating the keyword score 510 and extracting the presentation candidate keyword 512. FIG. 6 illustrates a case where the keyword score 510 is calculated based only on the click user information 506. Indicates a case where the keyword score 510 is calculated from the click user information 506 and the non-click user information 508.

  In FIG. 6, the keyword score calculation unit 509 normalizes (averages) each keyword included in the click user information 506 for each user and the entire user, and calculates a keyword score 510 (step S61).

  Next, the presentation candidate keyword extraction unit 511 outputs the top predetermined number of scores in the keyword score 510 as the presentation candidate keyword 512 (step S71).

  In FIG. 7, the keyword score calculation unit 509 normalizes (averages) each keyword included in the click user information 506 for each user and the entire user, and calculates a temporary score (step S62).

  In addition, the keyword score calculation unit 509 normalizes (averages) each keyword included in the non-click user information 508 for each user and the entire user to calculate a temporary score (step S63).

  Next, for each keyword, the keyword score calculation unit 509 subtracts the temporary score of the non-click user from the temporary score of the click user or divides the temporary score of the click user by the temporary score of the non-click user. Is calculated (step S64).

  Next, the presentation candidate keyword extraction unit 511 outputs the upper predetermined number of scores in the keyword score 510 as the presentation candidate keyword 512 (step S72).

  In FIG. 8, the keyword score calculation unit 509 normalizes (averages) each keyword included in the click user information 506 and calculates a temporary score (step S65).

  Further, the keyword score calculation unit 509 calculates a temporary score by normalizing (averaging) each keyword included in the non-click user information 508 (step S66).

  Next, the keyword score calculation unit 509 attaches a label “+1” to the data corresponding to each user of the temporary score obtained from the click user information 506 for each keyword, and obtains the temporary score obtained from the non-click user information 508. A label “−1” is assigned to the data corresponding to each individual user, the data is sorted in descending order of score, and the upper predetermined number is left as a list (step S67).

  Next, the keyword score calculation unit 509 calculates an AUC (Area Under Curve) from the list (step S68).

  FIG. 9 is a diagram illustrating an example of calculating AUC.

Focus on the list sorted in descending order from the highest score,
FPR (False positive rate)
= (The number of labels “−1” included in the score above the target score)
/ (Total number of labels "-1")
When,
TPR (True positive rate)
= (Number of labels “+1” included in scores above the target score)
/ (Total number of labels "+1")
And the area surrounded by the graph plotted for each focus with FPR as the horizontal axis and TPR as the vertical axis and the horizontal axis is obtained as AUC.

  AUC is an index for evaluating the accuracy of the binary classification. AUC is 0.5 for a random arrangement, and the one with a high score of “+1” is divided into two in the middle. It becomes. The fact that the label “+1” is biased toward the higher score means that users who searched with that keyword are more likely to click on the target ad, and users who do not click will be less likely to search with that keyword, It is suitable as a keyword for search-linked advertising.

  Returning to FIG. 8, the presentation candidate keyword extraction unit 511 excludes keywords whose AUC does not exceed a predetermined value, and outputs a keyword exceeding the predetermined value as the presentation candidate keyword 512 (step S <b> 73). Note that the upper predetermined number of AUCs may be output as the presentation candidate keywords 512 without determining whether or not the AUC exceeds a predetermined value.

  Next, in FIG. 2D, the keyword presenting unit 516 of the advertisement management apparatus 5 presents the keyword set in the advertisement database 513 when the keyword accepting unit 515 accepts the keyword used for the search-linked advertisement from the advertiser. Candidate keywords 512 are presented (suggested) (step S8).

  When the general user accesses the service server device 4 from the terminal device 2 and performs a search, the advertisement distribution unit 517 searches for the search keyword based on the advertisement distribution request from the service server device 4 with the input search keyword. The advertisement associated with is acquired from the advertisement database 513 and distributed.

<Summary>
As described above, according to the present embodiment, it is possible to present an effective keyword for enhancing the advertising effect to the advertiser of the search-linked advertisement.
The present invention has been described above by the preferred embodiments of the present invention. While the invention has been described with reference to specific embodiments, various modifications and changes may be made to the embodiments without departing from the broad spirit and scope of the invention as defined in the claims. Obviously you can. In other words, the present invention should not be construed as being limited by the details of the specific examples and the accompanying drawings.

DESCRIPTION OF SYMBOLS 1 Network 2, 3 Terminal device 4 Service server apparatus 5 Advertisement management apparatus 501 Log acquisition part 502 Log data 503 Advertisement selection part 504 Advertiser ID
505 Click user information extraction unit 506 Click user information 507 Non-click user information extraction unit 508 Non-click user information 509 Keyword score calculation unit 510 Keyword score 511 Presentation candidate keyword extraction unit 512 Presentation candidate keyword 513 Advertisement database 514 Advertisement submission reception unit 515 Keyword reception unit 516 Keyword presentation unit 517 Advertisement distribution unit

Claims (5)

A means of selecting a target ad for search-based advertising,
A means for identifying a user who has viewed an advertisement of an advertiser of the target advertisement from log data, and extracting a search keyword input by the user;
Means for calculating a score for each search keyword based on the input frequency of the extracted search keyword;
A content distribution apparatus comprising: means for presenting a search keyword having a high score as a candidate when receiving an input of a keyword of a search-linked advertisement for the target advertisement. The content delivery apparatus according to claim 1,
Further comprising means for identifying a user who has not viewed the advertiser's advertisement from the log data, and extracting a search keyword input by the user;
The means for calculating a score for each search keyword subtracts the input frequency of the user who has not viewed the advertiser's advertisement from the input frequency of the user who has viewed the advertiser's advertisement as the input frequency of the search keyword. A content distribution apparatus, characterized in that a value or a value obtained by dividing an input frequency of a user who has viewed the advertiser's advertisement by an input frequency of a user who has not viewed the advertiser's advertisement. The content delivery apparatus according to claim 1,
Further comprising means for identifying a user who has not viewed the advertiser's advertisement from the log data, and extracting a search keyword input by the user;
The means for calculating a score for each search keyword calculates a score for each user, and the labels are different for the score of the user who has viewed the advertiser's advertisement and the score of the user who has not viewed the advertiser's advertisement. The content distribution apparatus is characterized in that, after sorting in descending order of the scores, one having a high evaluation value of the binary classification of the labels is extracted. The content distribution apparatus according to claim 3,
The evaluation value of the binary classification of the label is
Pay attention to the top score,
FPR = (number of labels indicating that no click was included in a score equal to or higher than the target score) / (total number of labels indicating that no click was performed)
When,
TPR = (number of labels indicating clicks included in a score equal to or higher than the target score) / (total number of labels indicating clicks)
The content distribution apparatus is characterized in that the FPR is the horizontal axis and the TPR is the vertical axis and the area is surrounded by a graph plotted for each focus and the horizontal axis. A step in which the content distribution device selects a target advertisement that is a target of a search-linked advertisement;
The content distribution device identifying a user who has viewed an advertisement of an advertiser of the target advertisement from log data, and extracting a search keyword input by the user;
The content distribution device calculating a score for each search keyword based on the input frequency of the extracted search keyword;
And a step of presenting a search keyword having a high score as a candidate when the content distribution device accepts an input of a keyword of a search-linked advertisement for the target advertisement.