JP6590880B2 - Extraction apparatus, extraction method, and extraction program - Google Patents

Description

  The present invention relates to an extraction apparatus, an extraction method, and an extraction program.

  Conventionally, techniques for utilizing search queries have been proposed in order to set appropriate keywords for search-linked advertisements and content-linked advertisements. For example, a technique has been proposed in which a plurality of co-occurrence query groups are acquired based on a search query, and a co-occurrence query common to the groups is extracted as a common query from the plurality of acquired co-occurrence query groups.

JP 2012-113486 A

  However, with the above-described conventional technology, it is not always possible to extract the characteristics of a specific user group with high accuracy. Specifically, in the above-described conventional technology, a plurality of co-occurrence query groups are acquired based on the search query, and a common co-occurrence query between the groups is acquired as a common query from the plurality of acquired co-occurrence query groups. However, the feature of a specific user group cannot always be extracted with high accuracy.

  The present application has been made in view of the above, and an object thereof is to provide an extraction device, an extraction method, and an extraction program that can extract the characteristics of a specific user group with high accuracy.

  The extraction device according to the present application includes a first feature amount of a first user of a first group and a feature amount of a second user of a second group, the second feature amount being different from the first feature amount; The third feature amount of the first user of the first group, which is a feature amount obtained based on the first feature amount acquired by the acquisition unit and the second feature amount acquired by the acquisition unit, is extracted. And an extraction unit.

  According to one aspect of the embodiment, there is an effect that the characteristics of a specific user group can be extracted with high accuracy.

FIG. 1 is a diagram illustrating an example of an extraction process according to the embodiment. FIG. 2 is a diagram illustrating a configuration example of the extraction device according to the embodiment. FIG. 3 is a diagram illustrating an example of a search query storage unit according to the embodiment. FIG. 4 is a diagram illustrating an example of a determination process according to the modification. FIG. 5 is a flowchart illustrating an extraction processing procedure according to the embodiment. FIG. 6 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the extraction device.

  Hereinafter, a mode for carrying out an extraction apparatus, an extraction method, and an extraction program according to the present application (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings. Note that the extraction apparatus, the extraction method, and the extraction program according to the present application are not limited to the embodiment. In the following embodiments, the same portions are denoted by the same reference numerals, and redundant description is omitted.

[1. Example of extraction process)
First, an example of the extraction process according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of an extraction process according to the embodiment. In FIG. 1, the extraction device 100 is referred to as a search query (hereinafter referred to as “first search query” as appropriate) input by the first user of the first group as the first feature amount of the first user of the first group. The example of acquiring the weight of the appearance probability in the first group of) is shown. In addition, the extraction device 100 may use a search query (hereinafter referred to as appropriate) that is input by the second user of the second group as a second feature amount that is a feature amount of the second user of the second group and is different from the first feature amount. An example of acquiring the weight of the appearance probability in the second group of “second search query” is shown. Then, the extraction device 100 obtains the difference between the weight of the appearance probability in the first group of the first search query and the weight of the appearance probability in the second group of the second search query, thereby obtaining the first user of the first group. As an example of the third feature quantity, an example of extracting the difference in the weights of the appearance probabilities of the respective search queries is shown.

  As shown in FIG. 1, the extraction system 1 includes a search server 20, an advertisement distribution server 30, and an extraction device 100. These various apparatuses are communicably connected via a network N (for example, the Internet) in a wired or wireless manner. The extraction system 1 shown in FIG. 1 may include a plurality of search servers 20 and a plurality of advertisement distribution servers 30. Moreover, in this embodiment, the search server 20, the advertisement delivery server 30, and the extraction apparatus 100 shall be managed by the provider T1.

  The search server 20 is an information processing apparatus that receives a search query input from a user and returns a search result of content such as a web page. In addition, the search server 20 stores information related to the search query and the search date and time input by the user in the search history storage unit. And the search server 20 transmits the information regarding the search query of the user stored in the search history memory | storage part to the extraction apparatus 100 according to the request | requirement from the extraction apparatus 100 mentioned later.

  The advertisement distribution server 30 is an information processing apparatus that distributes advertisement content to be distributed determined by the extraction device 100 to a distribution destination of advertisement content determined by the extraction device 100 described later. Further, the advertisement distribution server 30 stores information regarding the advertisement content and the keyword set for the advertisement content in the advertisement information storage unit. And the advertisement delivery server 30 transmits the information regarding the keyword set to the advertisement content stored in the advertisement information storage unit to the extraction device 100 in response to a request from the extraction device 100 described later.

  The extraction apparatus 100 takes a difference between the weight of the appearance probability in the specific user group of the search query of the specific user group and the weight of the appearance probability in the other user group of the search query of the other user group. It is an information processing apparatus that extracts a difference in weights of appearance probabilities of search queries for a user group. The extraction device 100 sends a search query transmission request for a user belonging to a specific user group to the search server 20 and acquires a search query for a user belonging to the specific user group from the search server 20. Further, the extraction apparatus 100 sends a request for transmitting information related to the keyword set in the advertisement content to the advertisement distribution server 30 and acquires information related to the keyword set in the advertisement content from the advertisement distribution server 30.

  Hereinafter, an example of the extraction process will be described with reference to FIG. In the example illustrated in FIG. 1, the extraction device 100 acquires a first search query from the search server 20. The extraction device 100 acquires a search query Q1, a search query Q2, a search query Q3, a search query Q4, a search query Q5,... As a first search query (step S1).

  Subsequently, the extraction apparatus 100 calculates the weight of the appearance probability in the first group of each first search query. The extraction device 100 inputs each first search query to the naive Bayes classifier, and calculates a weight related to whether or not there is a high probability of being searched only by the first user of the first group (step S2). And the extraction apparatus 100 acquires the weight of the appearance probability in the 1st group of each 1st search query as the 1st feature amount of the 1st user of the 1st group. For example, if there are 100 first users in the first group, and 90 out of 100 users input the search query Q1, the appearance probability of the search query Q1 in the first group is 90%. Then, the extraction apparatus 100 acquires 0.9 as the weight of the appearance probability in the first group of the search query Q1.

  Subsequently, the extraction device 100 acquires a second search query from the search server 20. The extraction device 100 acquires a search query Q1, a search query Q2, a search query Q3, a search query Q4, a search query Q5,... As a second search query (step S3).

  Subsequently, the extraction apparatus 100 calculates the weight of the appearance probability in the second group of each second search query. The extraction device 100 inputs each second search query to the naive Bayes classifier, and calculates a weight related to whether or not there is a high probability of being searched only by the second user of the second group (step S4). And the extraction apparatus 100 acquires the weight of the appearance probability in the 2nd group of each 2nd search query as a 2nd feature-value of the 2nd user of a 2nd group. For example, if there are 1000 second users in the second group and 100 out of 1000 users input the search query Q1, the appearance probability of the search query Q1 in the second group is 10%. Then, the extraction device 100 acquires 0.1 as the weight of the appearance probability in the second group of the search query Q1.

  Subsequently, the extraction apparatus 100 subtracts the weight of the appearance probability in the second group of each search query from the weight of the appearance probability in the first group of each search query (step S5). For example, the extraction apparatus 100 subtracts the weight 0.1 of the appearance probability in the second group of the search query Q1 from the weight 0.9 of the appearance probability in the first group of the search query Q1.

  Subsequently, the extraction apparatus 100 extracts the difference in the weights of the appearance probabilities of the respective search queries as the third feature amount of the first user of the first group (Step S6). For example, the extraction apparatus 100 subtracts the appearance probability weight 0.1 in the second group of the search query Q1 from the appearance probability weight 0.9 in the first group of the search query Q1, thereby reducing the appearance probability of the search query Q1. 0.8 is extracted as the weight difference. Here, a large value of the difference in the weights of the appearance probabilities of the search query means that there is a high probability of being searched only by the first user of the first group. Therefore, since the extraction device 100 is a search query that has a higher probability of being searched only by the first user of the first group, the search query having a larger difference value of the weights of appearance probabilities is unique to the first user of the first group. It is determined that it is a search query. Then, the extraction apparatus 100 extracts a search query unique to the first user of the first group as the third search query.

  Subsequently, the extraction apparatus 100 determines the distribution destination of the advertisement content based on the extracted third search query of the first user of the first group (step S7).

  In addition, the extraction apparatus 100 determines the advertisement content related to the extracted third search query of the first user of the first group as the advertisement content to be distributed. As described above, the extraction device 100 determines the advertisement content to be distributed based on the extracted third search query of the first user of the first group (step S8).

  Subsequently, the extraction device 100 transmits the advertising content distribution destination and the advertising content determined based on the third search query of the first user of the first group to the advertising distribution server 30 (step S9).

  As described above, the extraction apparatus 100 uses the first group of the first query of the first group as the first feature amount of the first user of the first group, and the appearance probability of the first group in the first group. Get the weight of. In addition, the extraction apparatus 100 uses a search query (second search query) input by the second user of the second group as a second feature quantity that is a feature quantity of the second user of the second group and is different from the first feature quantity. ) Of the appearance probability in the second group. Then, the extraction apparatus 100 subtracts the weight of the appearance probability in the second group of each search query from the weight of the appearance probability in the first group of each search query, thereby obtaining the third feature amount of the first user of the first group. Then, the difference between the weights of the appearance probabilities of the respective search queries is extracted. Then, the extraction apparatus 100 extracts a search query having a large difference in weight of appearance probability of each search query as the third search query.

  Thus, the extraction apparatus 100 can extract the third search query having a high probability of being searched only by the first user in the first group by extracting the difference in the weights of the appearance probabilities of the search queries. . Thereby, the extraction apparatus 100 can extract the 3rd search query which is a search query peculiar to the 1st user of a 1st group with high precision. The third search query, which is a search query unique to the first user of the first group, is information reflecting the characteristics of the first user of the first group. Therefore, the extraction apparatus 100 can extract the characteristics of a specific user group with high accuracy.

  The example extraction system 1 shown in FIG. 1 can be applied to the following premise service. For example, it is assumed that there is a business operator X that provides a proxy service for CRM (Customer Relationship Management) business and a business operator Y that desires to acquire excellent customers. In this case, the business operator X applies the extraction system 1 based on the customer list of the superior customer of the business operator Y and the customer list of the existing customer of the business operator Y. Specifically, the weight of the occurrence probability of the existing customer of the operator Y in the search query of the existing customer of the operator Y is subtracted from the weight of the appearance probability of the excellent customer of the operator Y in the search query of the superior customer of the operator Y. Thus, the search query specific to the good customer of the business operator Y is extracted as the third search query. Then, the business operator X determines a user who inputs a search query similar to the third search query of the superior customer of the business operator Y among the existing customers of the business operator Y as the delivery destination of the advertising content. Propose to. Further, the business X distributes advertisement content to a user who has entered a search query similar to the third search query of the good customer of the business Y based on the third search query of the good customer of the business Y. And proposes to the operator Y. In this way, the business operator X can make an effective proposal for acquiring a good customer of the business operator Y by the extraction system 1 shown in FIG. Therefore, the business operator X can promote up-sell and cross-sell to the existing customer of the business operator Y by the extraction system 1 shown in FIG.

  Moreover, the extraction system 1 of the example shown in FIG. 1 is applicable to the following premise services. Specifically, it is assumed that there is a business operator X that provides a CRM service substitution service and a business operator Z that desires to acquire a new customer. In this case, the business operator X applies the extraction system 1 based on the customer list of the existing customer of the business operator Z and the list of general users who are not customers of the business operator Z. Specifically, in the general user who is not the customer of the operator Z in the search query of the general user who is not the customer of the operator Z from the weight of the appearance probability in the existing customer of the operator Z in the search query of the existing customer of the operator Z By subtracting the weight of the appearance probability, a search query unique to the existing customer of the operator Z is extracted as the third search query. Then, the business operator X is a general user who is not a customer of the business operator Z, and a user who inputs a search query similar to the third search query of the existing customer of the business operator Z is a distribution destination of the business content of the business operator Z And propose to the operator Z. Further, the business operator X distributes an advertisement distributed to a general user who has input a search query similar to the third search query of the existing customer of the business operator Z based on the third search query of the existing customer of the business operator Z. The content is determined and proposed to the operator Z. Thus, the business operator X can make an effective proposal for acquiring a new customer of the business operator Z by the extraction system 1 shown in FIG. Accordingly, the business operator X can prompt the business operator Z to acquire a new customer by using the extraction system 1 shown in FIG.

  In the example illustrated in FIG. 1, the extraction apparatus 100 extracts features of a specific user group based on weights of appearance probabilities in the user group to which the user belongs in the search query input by the user as the user feature amount. As an example, the feature of a specific user group is extracted based on the weight of appearance probability in the user group to which the user belongs, such as a user's purchased product, usage service, or search site. Also good. Further, the extraction apparatus 100 may extract features of a specific user group based on weights of appearance probabilities in the user group to which the user belongs, such as a demographic attribute, psychographic attribute, and behavior attribute of the user. In addition, the extraction apparatus 100 does not use the weight of the appearance probability, but the characteristics of a specific user group based on a user search query, a purchased product, a use service, a search site, a demographic attribute, a psychographic attribute, or an action attribute. It may be extracted.

  In the example illustrated in FIG. 1, the extraction apparatus 100 uses the search query itself input by the user as the feature amount of the user. However, a group of search queries input by the user may be used. . For example, the extraction apparatus 100 may treat hiragana “watermelon”, katakana “watermelon”, and kanji “saijo” as groups of search queries, and group search queries as synonymous.

  In the example illustrated in FIG. 1, the extraction device 100 and the advertisement distribution server 30 are separate devices, but the extraction device 100 and the advertisement distribution server 30 may be integrated. For example, the extraction device 100 has the function of the advertisement distribution server 30, and the advertisement distribution of the advertisement distribution server 30, the extraction of a search query for a specific user group, the determination of the distribution destination of the advertisement content, the distribution of the advertisement content to be distributed Both decisions may be made.

[2. (Extractor configuration)
Next, the extraction apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 2 is a diagram illustrating a configuration example of the extraction apparatus 100 according to the embodiment. As illustrated in FIG. 2, the extraction device 100 includes a communication unit 110, a storage unit 120, and a control unit 130.

(Communication unit 110)
The communication unit 110 is realized by, for example, a NIC (Network Interface Card). The communication unit 110 is connected to the network N in a wired or wireless manner, and transmits and receives information to and from the search server 20 and the advertisement distribution server 30, for example.

(Storage unit 120)
The storage unit 120 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. As illustrated in FIG. 2, the storage unit 120 includes a search query storage unit 121.

(Search query storage unit 121)
The search query storage unit 121 stores various types of information related to a user search query acquired from the search server 20. FIG. 3 shows an example of the search query storage unit 121 according to the embodiment. In the example illustrated in FIG. 3, the search query storage unit 121 includes items such as “user ID”, “search query”, and “date and time”.

  In the example shown in FIG. 3, the first record indicates that the user (user U1) identified by the user ID “U1” searched for the search query “Q1” on the date “August 1, 2017 10:00”. Show.

(Control unit 130)
The control unit 130 is a controller, for example, various programs (an example of a generation program) stored in a storage device inside the extraction device 100 by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like. This is realized by executing the RAM as a work area. The control unit 130 is a controller, and is realized by an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

  As shown in FIG. 2, the control unit 130 includes an acquisition unit 131, an extraction unit 132, and a determination unit 133, and realizes or executes the information processing operation described below. The internal configuration of the control unit 130 is not limited to the configuration illustrated in FIG. 2, and may be another configuration as long as the information processing described later is performed.

(Acquisition part 131)
The acquisition unit 131 acquires a second feature amount that is a first feature amount of the first user of the first group and a feature amount of the second user of the second group and is different from the first feature amount. The acquisition unit 131 acquires the degree of association between the first user of the first group and each element related to the first user as the first feature amount, and the second user and the second user of the second group as the second feature amount. Acquire each degree of association with each element related to the user. Specifically, the acquisition unit 131 acquires information on a search query input by the first user of the first group and the second user of the second group as the first feature amount and the second feature amount. For example, the acquisition unit 131 uses the first feature amount based on the number of times the first user of the first group has input the search query as the degree of association between the first user and each search query related to the first user. The weight of the appearance probability in the first group of one search query is acquired. In addition, the acquisition unit 131 acquires the weight of the appearance probability in the second group of the second search query based on the number of times the second user of the second group has input the search query as the second feature amount.

  The acquisition unit 131 may acquire, as each element related to the user, a purchased product purchased by the user, a service used by the user, a site browsed by the user, or the like, in addition to the search query input by the user. For example, the acquisition unit 131 uses, as the first feature amount, the purchased product as the degree of association between the first user and each purchased product related to the first user based on the number of times the first user of the first group has purchased the product. The weight of the purchase probability in the first group may be acquired. For example, the acquisition unit 131 uses, as the first feature amount, the usage service as the degree of association between the first user and each usage service related to the first user based on the number of times the first user of the first group has used the service. The weight of the use probability in the first group may be acquired. For example, the acquisition unit 131 uses, as the first feature amount, the browsing site as the degree of association between the first user and each browsing site related to the first user based on the number of times the first user of the first group browsed the site. You may acquire the weight of the browsing probability in the 1st group. The acquisition unit 131 acquires the second feature value in the same manner as the first feature value.

  Specifically, the acquisition unit 131 includes a first search query input by a first user in the first group and a search query input by a second user in the second group, which is different from the first search query. Is transmitted to the search server 20. Subsequently, the acquisition unit 131 acquires from the search server 20 the first search query input by the first user as each element related to the first user. Moreover, the acquisition part 131 acquires the 2nd search query which the 2nd user input as each element regarding a 2nd user from the search server 20. FIG.

  Subsequently, the acquisition unit 131 calculates the weight of the appearance probability in the first group of the first search query. Specifically, the acquiring unit 131 inputs the first search query to the naive Bayes classifier and calculates a weight related to whether or not there is a high probability of being searched only by the first user of the first group. And the acquisition part 131 acquires the weight of the appearance probability in the 1st group of a 1st search query as a 1st feature-value of the 1st user of a 1st group.

  Subsequently, the acquisition unit 131 calculates the weight of the appearance probability in the second group of the second search query. Specifically, the acquisition unit 131 inputs the second search query to the naive Bayes classifier and calculates a weight related to whether or not there is a high probability of being searched only by the second user of the second group. And the acquisition part 131 acquires the weight of the appearance probability in the 2nd group of a 2nd search query as a 2nd feature-value of the 2nd user of a 2nd group.

  The acquisition unit 131 acquires the feature amount of the user in the group including the first group as the second feature amount. For example, it is assumed that the first group is a good customer and the second group is an existing customer including the good customer. In this case, the acquisition unit 131 acquires, as the first feature amount, the weight of the appearance probability of the excellent customer search query in the excellent customer. And the acquisition part 131 acquires the weight of the appearance probability in the existing customer which includes the quality customer of the search query of the existing customer which includes the quality customer as a 2nd feature-value. Further, for example, it is assumed that the first group is an existing customer and the second group is a general user who is not a customer. In this case, the acquisition unit 131 acquires the weight of the appearance probability of the existing customer search query in the existing customer as the first feature amount. And the acquisition part 131 acquires the weight of the appearance probability in the general user who is not a customer of the search query of the general user who is not a customer as a 2nd feature-value.

  The acquisition unit 131 acquires, as the second feature amount, the feature amount of the first user in the first group at a time different from the time when the first feature amount is acquired. For example, the acquisition unit 131 acquires the weight of the appearance probability in the first group of the first search query of the first user of the first group in September 2017. And the acquisition part 131 acquires the weight of the appearance probability in the 1st group of the 2nd search query of the 1st user of the 1st group in December, 2017.

(Extractor 132)
The extraction unit 132 is a third feature amount of the first user of the first group, which is a feature amount obtained based on the first feature amount acquired by the acquisition unit 131 and the second feature amount acquired by the acquisition unit 131. To extract. The extraction unit 132 extracts the third feature value by taking the difference between the first feature value acquired by the acquisition unit 131 and the second feature value acquired by the acquisition unit 131. Specifically, the extracting unit 132 includes each degree of association between the first user of the first group acquired by the acquiring unit 131 and each element related to the first user and each of the second user and second user of the second group. By taking the difference between each element and the degree of association, the difference between the degrees of association of each element is extracted as the third feature amount of the first user of the first group. And the extraction part 132 extracts the element with the large value of the difference of each relevance degree of each element as a 3rd element. And the extraction apparatus 100 extracts the element peculiar to the 1st user of a 1st group as a 3rd element.

  For example, the extraction unit 132 calculates the difference between the weight of the appearance probability in the first group of the first search query that is the first feature quantity and the weight of the appearance probability in the second group of the second search query that is the second feature quantity. As a result, the difference in the weight of the appearance probability of each search query is extracted as the third feature amount of the first user of the first group. Subsequently, the extraction unit 132 arranges the search queries in descending order of the extracted difference value of the weights of appearance probabilities. The extraction unit 132 may rank the search queries in descending order of the extracted difference value of the weights of appearance probabilities. Then, the extraction unit 132 extracts a search query having a large difference in the weights of appearance probabilities of the search queries as a third search query.

  The extraction unit 132 is a search query that has a higher probability of being searched only by the first user of the first group, as the search query has a positive difference in the weight value of the appearance probability extracted and a larger absolute value. Therefore, it is determined that the search query is unique to the first user of the first group. That is, the extraction unit 132 determines that the search query is more specific to the first user of the first group as the difference between the weights of the appearance probabilities is a positive value and the absolute value is larger. Therefore, the extraction apparatus 100 extracts a search query having a positive difference in the appearance probability weight difference and a large absolute value as the third search query.

  In addition, the extraction part 132 is a search query with a lower probability of being searched only by the first user of the first group for a search query in which the extracted difference value of the weights of appearance probabilities is a negative value and has a larger absolute value. Therefore, it is determined that the search query is not unique to the first user of the first group. That is, the extraction unit 132 determines that the search query specific to the first user of the first group is not as the negative value and the absolute value of the difference in the weights of the appearance probabilities are larger. Therefore, the extraction apparatus 100 does not extract a search query that has a negative difference in appearance probability weight and a large absolute value as a third search query.

  In addition, when the value of the difference between the weights of the extracted appearance probabilities is 0, the extraction unit 132 searches only by the probability that only the first user of the first group searches and the second user of the second group. Since the search queries have the same probability, it is determined that the search query is not unique to the first user of the first group. That is, the extraction unit 132 determines that the search query is not unique to the first user of the first group when the difference value of the appearance probability weight is 0. Therefore, the extraction device 100 does not extract a search query whose appearance probability weight difference value is 0 as the third search query.

  For example, the extraction unit 132 calculates the weight of the appearance probability in the good customer of the search query of the good customer acquired by the acquisition unit 131 and the weight of the appearance probability in the existing customer of the search query of the existing customer acquired by the acquisition unit 131. By taking the difference, the difference in the weight of the appearance probability of the search query of the excellent customer is extracted. And the extraction part 132 extracts a search query with a large difference value of the weight of the appearance probability of each search query as a 3rd search query of a quality customer.

  For example, the extraction unit 132 uses the weight of the appearance probability of the existing customer search query acquired by the acquisition unit 131 and the general user who is not the customer of the search query of the general user who is not the customer acquired by the acquisition unit 131. By taking the difference with the weight of the appearance probability, the difference in the weight of the appearance probability of the search query of the existing customer is extracted. And the extraction part 132 extracts the search query with a large difference value of the weight of the appearance probability of each search query as a 3rd search query of an existing customer.

(Determining unit 133)
The determination unit 133 controls content distribution based on the third feature amount extracted by the extraction unit 132. Specifically, the determination unit 133 controls content distribution based on the third search query extracted based on the third feature amount extracted by the extraction unit 132. For example, a transmission request for information related to the keyword set in the advertisement content is sent to the advertisement distribution server 30. Subsequently, the determination unit 133 acquires information about the keyword set in the advertisement content from the advertisement distribution server 30. Then, the determination unit 133 controls content distribution based on the keyword set for the advertisement content and the third search query extracted by the extraction unit 132.

  The determination unit 133 determines the content to be distributed based on the third feature amount extracted by the extraction unit 132. Specifically, the determination unit 133 determines the content to be distributed based on the third search query extracted based on the third feature amount extracted by the extraction unit 132. Based on the second feature amount acquired by the acquisition unit 131 and the third feature amount extracted by the extraction unit 132, the determination unit 133 determines the second group of users having a feature amount similar to the third feature amount. Determine the content to be distributed. For example, the determination unit 133 inputs the second search query similar to the third search query based on the second search query acquired by the acquisition unit 131 and the third search query extracted by the extraction unit 132. The content to be distributed to the second user is determined. For example, the determination unit 133 determines advertisement content in which a keyword similar to the third search query is set as content to be distributed to the second user of the second group.

  Further, the determination unit 133 determines a content distribution destination based on the third feature amount extracted by the extraction unit 132. Specifically, the determination unit 133 determines a content distribution destination based on the third search query extracted based on the third feature amount extracted by the extraction unit 132. Based on the second feature amount acquired by the acquisition unit 131 and the third feature amount extracted by the extraction unit 132, the determination unit 133 selects a second group of users having a feature amount similar to the third feature amount. Determine as the content distribution destination. For example, the determination unit 133 inputs the second search query similar to the third search query based on the second search query acquired by the acquisition unit 131 and the third search query extracted by the extraction unit 132. Is determined as a content distribution destination.

  Further, the determination unit 133 may determine the advertisement content to be distributed based on a search query in which the order arranged in the weight order is within a predetermined order from among the third search queries. For example, the determination unit 133 may determine an advertisement content in which a keyword similar to the third search query whose rank arranged in the weight order is within a predetermined rank is set as the advertisement content to be distributed. Further, the determination unit 133 may determine the distribution destination of the advertisement content based on the search query in which the order arranged in the weight order is within a predetermined order from among the third search queries.

  In addition, the determination unit 133 may determine the advertisement content to be distributed based on a search query in which the calculated weight is equal to or greater than a predetermined threshold among the third search queries. For example, the determination unit 133 may determine an advertisement content in which a keyword similar to the third search query in which the calculated weight is equal to or greater than a predetermined threshold is set as the advertisement content to be distributed. In addition, the determination unit 133 may determine the distribution destination of the advertisement content based on a search query in which the calculated weight is equal to or greater than a predetermined threshold among the third search queries.

  Further, the determination unit 133 may determine the advertisement content to be distributed based on the third search query of the good customer extracted by the extraction unit 132. For example, the determination unit 133 may determine advertisement content to be distributed to an existing customer who has input a search query similar to the third search query of the good customer based on the third search query of the good customer. Further, the determination unit 133 may determine the distribution destination of the advertisement content based on the third search query of the good customer extracted by the extraction unit 132. For example, the determination unit 133 may determine an existing customer who has input a search query similar to the third search query of a good customer among existing customers as a distribution destination of the advertisement content.

  Further, the determination unit 133 may determine the advertising content to be distributed based on the third search query of the existing customer extracted by the extraction unit 132. For example, the determination unit 133 determines advertisement content to be distributed to a general user who is not a customer who has input a search query similar to the third search query of the existing customer based on the third search query of the existing customer. Also good. The determination unit 133 may determine the distribution destination of the advertisement content based on the third search query of the existing customer extracted by the extraction unit 132. For example, the determination unit 133 may determine a general user who is not a customer who has input a search query similar to the third search query of an existing customer among general users who are not customers as a distribution destination of the advertisement content.

[3. (Extraction process flow)
Next, the procedure of the extraction process according to the embodiment will be described with reference to FIG. FIG. 5 is a flowchart illustrating an extraction processing procedure according to the embodiment.

  As illustrated in FIG. 5, the extraction device 100 acquires a search query for users in the first group (step S <b> 101).

  Subsequently, the extraction apparatus 100 calculates the weight of the appearance probability in the first group of each search query (step S102).

  Subsequently, the extraction apparatus 100 acquires a search query for the second group of users (step S103).

  Subsequently, the extraction device 100 calculates the weight of the appearance probability in the second group of each search query (step S104).

  Subsequently, the extraction apparatus 100 subtracts the weight of the appearance probability in the second group of each search query from the weight of the appearance probability in the first group of each search query, and extracts the difference in the weight of the appearance probability of each search query. (Step S105).

[4. (Modification)
The extraction system 1 according to the above-described embodiment may be implemented in various different forms other than the above-described embodiment. Therefore, in the following, another embodiment of the extraction system 1 will be described.

[4-1. The target user is determined based on the size of the difference vector)
An example of the determination process according to the modification will be described with reference to FIG. FIG. 4 is a diagram illustrating an example of a determination process according to the modification. FIG. 4 illustrates an example in which the extraction apparatus 100 determines a target user based on the magnitude of a difference vector between a feature amount vector of users in a predetermined group and a feature amount vector of users in other groups.

  In the example shown in FIG. 4, the weight of the appearance probability in each group of the search queries of the users of the first group to the third group is represented by a point in the N-dimensional feature vector space. Specifically, the vertical axis and horizontal axis shown in FIG. 4 may be individual search queries, or may be a group of search queries treated as synonymous. The vertical and horizontal axes shown in FIG. 4 may be, for example, a single search query “baseball”, or synonymous with “suika” in hiragana, “watermelon” in katakana, and “saijo” in kanji. It may be a group of search queries handled as a thing.

  The extraction apparatus 100 acquires the weight of the appearance probability of the existing customer of the operator Z in the search query of the existing customer of the operator Z as the feature vector of the first user of the first group. Further, the extraction apparatus 100 uses the weight of the appearance probability of the existing customer of the operator W in the search query of the existing customer of the operator W who is a competitor of the operator Z as the feature vector of the second user of the second group. get. Further, the extraction apparatus 100 acquires the weight of the appearance probability of the existing customer of the business operator V in the search query of the existing customer of the business operator V who is the competitor of the business operator Z as the feature quantity vector of the user of the third group. .

  Subsequently, the extraction device 100 determines the weight of the appearance probability of the existing customer of the operator Z in the search query of the existing customer of the operator Z and the appearance probability of the existing customer of the search query of the operator W of the operator W. By calculating the difference from the weight, a difference vector between the two is calculated. Further, the extraction apparatus 100 weights the appearance probability of the existing customer of the operator Z in the search query of the existing customer of the operator Z and the weight of the appearance probability of the existing customer of the operator V in the search query of the existing customer of the operator V. The difference vector between the two is calculated.

  As shown in FIG. 4, the weight of the appearance probability in the existing customer of the operator Z of the existing customer of the operator Z that is the first group and the operator of the search query of the existing customer of the operator W that is the second group. The third group is the weight of the appearance probability of the existing customer of the operator Z in the search query of the existing customer of the operator Z, which is the first group, than the magnitude of the difference vector from the weight of the appearance probability of the existing customer of W The size of the difference vector with the weight of the appearance probability of the existing customer of the operator V in the search query of the existing customer of the operator V is smaller.

  Therefore, the extraction apparatus 100 determines that the similarity between the existing customer of the business operator Z and the existing customer of the business operator V is higher than the similarity between the existing customer of the business operator Z and the existing customer of the business operator W. To do. Then, in order to acquire the existing customer of the competitor of the business operator Z as a new customer, the extraction apparatus 100 determines the existing customer of the business operator V having a higher similarity with the existing customer of the business operator Z as a target user.

[4-2. (Determined based on the difference of the search query itself)
In the example illustrated in FIG. 1, the extraction apparatus 100 extracts the difference in the weights of appearance probabilities for each search query. However, the extraction apparatus 100 may extract the difference in the search query itself. Specifically, the extraction device 100 acquires a first search query input by a first user of the first group from the search server 20. For example, the extraction device 100 acquires, as the first search query, the search query Q11, the search query Q12,..., The search query Q15, the search query Q16, the search query Q17,.

  Subsequently, the extraction device 100 acquires the second search query input by the second user of the second group from the search server 20. For example, the extraction device 100 acquires, as the second search query, the search query Q16, the search query Q17,..., The search query Q20,.

  Subsequently, the extraction apparatus 100 removes a search query common to the second search query of the second user of the second group from the first search query of the first user of the first group, thereby obtaining the first user of the first group. The third search query is extracted. For example, the extraction device 100 extracts the search query Q11, the search query Q12,..., The search query Q15,... As the third search query of the first user of the first group. In this way, the extraction apparatus 100 obtains the difference between the first search query of the first user of the first group and the second search query of the second user of the second group, thereby obtaining the first user of the first group. A third search query is extracted.

  Subsequently, the extraction apparatus 100 determines the distribution destination of the advertisement content based on the extracted third search query of the first user of the first group.

  In addition, the extraction apparatus 100 determines the advertisement content related to the extracted third search query of the first user of the first group as the advertisement content to be distributed. As described above, the extraction device 100 determines the advertisement content to be distributed based on the extracted third search query of the first user of the first group.

  Subsequently, the extraction device 100 transmits the advertisement content distribution destination and the advertisement content determined based on the third search query of the first user of the first group to the advertisement distribution server 30.

  As described above, the extraction device 100 acquires the search query input by the first user of the first group as the first feature amount of the first user of the first group. Further, the extraction device 100 acquires a search query input by the second user of the second group as a second feature amount that is a feature amount of the second user of the second group and is different from the first feature amount. Then, the extraction device 100 takes the difference between the first search query of the first user of the first group and the second search query of the second user of the second group, thereby obtaining the third of the first user of the first group. Extract search queries.

  As described above, the extraction apparatus 100 removes a search query that is common to the second search query of the second user of the second group from the first search query of the first user of the first group. The third search query that is a search query specific to the user can be extracted with high accuracy. The third search query, which is a search query unique to the first user of the first group, is information reflecting the characteristics of the first user of the first group. Therefore, the extraction apparatus 100 can extract the characteristics of a specific user group with high accuracy.

[5. effect〕
As described above, the extraction device 100 according to the embodiment includes the acquisition unit 131 and the extraction unit 132. The acquisition unit 131 acquires a second feature amount that is a first feature amount of the first user of the first group and a feature amount of the second user of the second group and is different from the first feature amount. The extraction unit 132 is a third feature amount of the first user of the first group, which is a feature amount obtained based on the first feature amount acquired by the acquisition unit 131 and the second feature amount acquired by the acquisition unit 131. To extract. In addition, the extraction unit 132 extracts the third feature amount by taking the difference between the first feature amount acquired by the acquisition unit 131 and the second feature amount acquired by the acquisition unit 131.

  As described above, the extraction apparatus 100 according to the embodiment removes the feature amount that is common to the second feature amount of the second user of the second group from the first feature amount of the first user of the first group. The third feature quantity specific to the first user of the group can be extracted with high accuracy. The third feature amount unique to the first user of the first group is information reflecting the feature of the first user of the first group. Therefore, the extraction apparatus 100 can extract the characteristics of a specific user group with high accuracy.

  In addition, the acquisition unit 131 acquires, as the first feature amount, each degree of association between the first user of the first group and each element related to the first user, and as the second feature amount, the second user of the second group Each degree of association with each element relating to the second user is acquired. In addition, the extraction unit 132 obtains the third feature amount of the first user of the first group by taking the difference between each degree of association in the first group acquired by the acquisition unit 131 and each degree of association in the second group. Extract.

  As described above, the extraction apparatus 100 according to the embodiment extracts the third element having a high degree of relevance only for the first user of the first group by extracting the difference in the relevance of each element between the groups. Can do. Thereby, the extraction apparatus 100 can extract the 3rd element which is an element peculiar to the 1st user of a 1st group with high precision. The third element, which is an element specific to the first user of the first group, is information reflecting the characteristics of the first user of the first group. Therefore, the extraction apparatus 100 can extract the characteristics of a specific user group with high accuracy.

  In addition, the acquisition unit 131 acquires the feature amount of the user in the group including the first group as the second feature amount.

  Thereby, the extraction apparatus 100 can extract the feature of the first user of the first group included in the second group with high accuracy.

  Further, the acquisition unit 131 acquires, as the second feature amount, the feature amount of the first user of the first group at a time different from the time when the first feature amount is acquired.

  Thereby, the extraction apparatus 100 can extract the feature which reflected the trend of the time with respect to the user of the same group with high precision.

  The extraction device 100 according to the embodiment further includes a determination unit 133 that controls content distribution based on the third feature amount extracted by the extraction unit 132. Specifically, the determination unit 133 determines the content to be distributed based on the third feature amount extracted by the extraction unit 132. Further, the determination unit 133 determines a content distribution destination based on the third feature amount extracted by the extraction unit 132. For example, the determination unit 133 determines the second group having a feature amount similar to the third feature amount based on the second feature amount acquired by the acquisition unit 131 and the third feature amount extracted by the extraction unit 132. Content to be distributed to the second user is determined. In addition, the determination unit 133 determines the second group having a feature amount similar to the third feature amount based on the second feature amount acquired by the acquisition unit 131 and the third feature amount extracted by the extraction unit 132. The second user is determined as a content distribution destination.

  Thereby, the extraction apparatus 100 can determine the user who has the characteristic similar to the 1st user of a 1st group among 2nd users of a 2nd group as a target user. Therefore, the extraction apparatus 100 determines a user having characteristics similar to those of the first user of the first group among the second users of the second group as a target user, and thereby new users of the first group from the second group. Can be earned. Then, the extraction apparatus 100 increases the second group to the first group by determining, as a target user, a user having characteristics similar to those of the first user of the first group among the second users of the second group. You can promote cell and cross-sell.

  Moreover, the acquisition part 131 acquires the information regarding the search query which the 1st user of the 1st group and the 2nd user of the 2nd group input as the 1st feature amount and the 2nd feature amount.

  Search queries reflect user trends with high accuracy. Therefore, the extraction apparatus 100 can extract a feature reflecting the trend of a specific user group with high accuracy.

[6. Hardware configuration)
Further, the extraction apparatus 100 according to the embodiment described above is realized by a computer 1000 having a configuration as shown in FIG. 6, for example. FIG. 6 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the extraction device 100. The computer 1000 includes a CPU 1100, RAM 1200, ROM 1300, HDD 1400, communication interface (I / F) 1500, input / output interface (I / F) 1600, and media interface (I / F) 1700.

  The CPU 1100 operates based on a program stored in the ROM 1300 or the HDD 1400 and controls each unit. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started up, a program depending on the hardware of the computer 1000, and the like.

  The HDD 1400 stores programs executed by the CPU 1100, data used by the programs, and the like. The communication interface 1500 receives data from other devices via a predetermined communication network, sends the data to the CPU 1100, and transmits the data generated by the CPU 1100 to other devices via the predetermined communication network.

  The CPU 1100 controls an output device such as a display and a printer and an input device such as a keyboard and a mouse via the input / output interface 1600. The CPU 1100 acquires data from the input device via the input / output interface 1600. In addition, the CPU 1100 outputs the generated data to the output device via the input / output interface 1600.

  The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1200. The CPU 1100 loads the program from the recording medium 1800 onto the RAM 1200 via the media interface 1700, and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  For example, when the computer 1000 functions as the extraction device 100 according to the embodiment, the CPU 1100 of the computer 1000 implements the function of the control unit 130 by executing a program or data loaded on the RAM 1200. The CPU 1100 of the computer 1000 reads and executes these programs or data from the recording medium 1800. However, as another example, these programs or data may be acquired from other devices via a predetermined communication network.

  As described above, some of the embodiments of the present application have been described in detail with reference to the drawings. However, these are merely examples, and various modifications, including the aspects described in the disclosure section of the invention, based on the knowledge of those skilled in the art, It is possible to implement the present invention in other forms with improvements.

[7. Others]
In addition, among the processes described in the above-described embodiments and modifications, all or a part of the processes described as being automatically performed can be manually performed, or are described as being performed manually. All or part of the processing can be automatically performed by a known method. In addition, the processing procedures, specific names, and information including various data and parameters shown in the document and drawings can be arbitrarily changed unless otherwise specified. For example, the various types of information illustrated in each drawing is not limited to the illustrated information.

  Each component of each illustrated device is functionally conceptual and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured.

  In addition, the above-described embodiments and modifications can be combined as appropriate within a range that does not contradict processing contents.

  In addition, the “section (module, unit)” described above can be read as “means” or “circuit”. For example, the extraction unit can be read as extraction means or an extraction circuit.

DESCRIPTION OF SYMBOLS 1 Extraction system 20 Search server 30 Advertisement delivery server 100 Extraction apparatus 121 Search query memory | storage part 131 Acquisition part 132 Extraction part 133 Determination part

Claims (9)

First search information that is information related to a search query history input by a user of the first group and second search information that is information related to a search query history input by a user of a second group different from the first group. , from the search server, and obtains via the communication unit, based on the first search information acquired, and calculates a first characteristic amount of a user of the first group for each search query, the second search acquired An acquisition unit that calculates a second feature amount of the second group of users for each search query based on the information ;
As the third feature amount of the first group of users, which is a feature amount obtained based on the first feature amount and the second feature amount calculated by the acquisition unit, the first feature amount and the second feature amount An extraction unit that extracts a difference from a feature amount for each search query, and extracts a search query indicating a feature of a user of the first group based on a value of a third feature amount extracted for each search query;
Based on the second search information acquired by the acquisition unit, among users of the second group, a user having a history of inputting the search query extracted by the extraction unit is determined as a content distribution destination, A determination unit that determines that the content related to the search query extracted by the extraction unit is distributed as content to be distributed;
An extraction device comprising: The acquisition unit
As the first feature amount, each degree of association between the user of the first group and each search query related to the user is acquired, and as the second feature amount, each search query related to the user of the second group and the user, Get the relevance of each
The extraction unit includes:
Extracting the third feature amount of the user of the first group by taking the difference between each degree of association in the first group acquired by the acquisition unit and each degree of association in the second group. The extraction device according to claim 1. The acquisition unit
The extraction device according to claim 1 or 2, wherein the feature amount of a user of a group including the first group is acquired as the second feature amount. The acquisition unit
The feature quantity of the user of the first group at a time different from the time when the first feature quantity is obtained is obtained as the second feature quantity. Extraction device. The acquisition unit
As the first search information, information on the history of a search query entered by a good customer who is a user posted on a customer list of a good customer at a predetermined business, and as the second search information, in the predetermined business The information regarding the history of the search query input by the existing customer who is a user posted in the customer list of the existing customer is acquired from the search server via the communication unit, and the above-described excellent information is obtained based on the acquired first search information. calculates the first feature quantity of customers in each search query, based on the acquired second search information, calculates a second characteristic amount of the existing customer for each search query,
The extraction unit includes:
As the third feature amount of the excellent customer, which is a feature amount obtained based on the first feature amount and the second feature amount calculated by the acquisition unit , the first feature amount and the second feature amount For each search query, based on the extracted third feature value, extract a search query indicating the characteristics of the excellent customer,
The determination unit
Among the existing customers, an existing customer having a history of inputting the search query extracted by the extraction unit is determined as a content distribution destination, and content related to the search query extracted by the extraction unit is set as distribution target content It determines to distribute. The extraction device as described in any one of Claims 1-4 characterized by the above-mentioned. The acquisition unit
As the first search information, information on the history of a search query entered by an existing customer who is a user posted on a customer list of an existing customer at a predetermined operator, and as the second search information at the predetermined operator Information related to the history of the search query input by the general user who is a user listed in the general user list is acquired from the search server via the communication unit, and the existing search is performed based on the acquired first search information. Calculating a first feature amount of the customer for each search query, and calculating a second feature amount of the general user for each search query based on the acquired second search information;
The extraction unit includes:
As the third feature amount of the existing customer, which is a feature amount obtained based on the first feature amount and the second feature amount calculated by the acquisition unit , the first feature amount and the second feature amount For each search query, and based on the extracted third feature value, extract a search query indicating the characteristics of the existing customer,
The determination unit
Among the general users, a general user having a history of inputting the search query extracted by the extraction unit is determined as a content distribution destination, and content related to the search query extracted by the extraction unit is set as content to be distributed. It determines to deliver. The extraction device according to any one of claims 1 to 5 characterized by things. The acquisition unit
The information regarding the search query which the user of the 1st group and the user of the 2nd group input as the 1st feature amount and the 2nd feature amount is acquired. The extraction device described in one. An extraction method performed by a computer,
First search information that is information related to a search query history input by a user of the first group and second search information that is information related to a search query history input by a user of a second group different from the first group. , from the search server, and obtains via the communication unit, based on the first search information acquired, and calculates a first characteristic amount of a user of the first group for each search query, the second search acquired An obtaining step of calculating a second feature amount of the second group of users for each search query based on the information ;
As the third feature amount of the first group of users, which is a feature amount obtained based on the first feature amount and the second feature amount calculated by the obtaining step, the first feature amount and the second feature amount An extraction step of extracting a difference from the feature amount for each search query, and extracting a search query indicating the feature of the user of the first group based on the extracted value of the third feature amount;
Based on the second search information acquired by the acquisition step, among the users of the second group, a user having a history of inputting the search query extracted by the extraction step is determined as a content distribution destination, A determination step of determining that content related to the search query extracted by the extraction step is distributed as content to be distributed;
The extraction method characterized by including. First search information that is information related to a search query history input by a user of the first group and second search information that is information related to a search query history input by a user of a second group different from the first group. , from the search server, and obtains via the communication unit, based on the first search information acquired, and calculates a first characteristic amount of a user of the first group for each search query, the second search acquired Obtaining means for calculating, for each search query, a second feature amount of the second group of users based on the information ;
As the third feature value of the first group of users, which is a feature value obtained based on the first feature value and the second feature value calculated by the acquisition means, the first feature value and the second feature value Extraction means for extracting a difference from the feature amount for each search query, and extracting a search query indicating a feature of the user of the first group based on the extracted value of the third feature amount;
Based on the second search information acquired by the acquisition unit, among the users of the second group, a user having a history of inputting the search query extracted by the extraction unit is determined as a content distribution destination, Determining means for deciding to distribute the content related to the search query extracted by the extracting means as the content to be distributed;
An extraction program characterized by causing a computer to execute.

Images