JP7018762B2 - Information processing equipment, information processing methods, and programs - Google Patents

Description

The present invention relates to an information processing apparatus, an information processing method, and a program.

Conventionally, a method for increasing the possibility that a user takes an action (for example, a click operation) with respect to a content such as an article or an advertisement posted on a web page has been known. For example, there is known a method of selecting an advertisement so that the number of unique users who have acted in the past is the largest among a predetermined number of combinations of advertisements based on access history information such as a user's click rate. (For example, see Patent Document 1).

Japanese Unexamined Patent Publication No. 2013-89080

It is known that when a large amount of content similar in content is posted on a certain web page, the click rate of the content on the entire web page tends to decrease. Therefore, in order to improve the click rate, it is important to have diversity in the content posted on the web page.

Further, in the method disclosed in Patent Document 1, user access history information is required to determine the content to be distributed. For this reason, new content that has not been accessed by the user may be difficult to be selected as a distribution target, or may not be selected at all.

The present invention has been made in consideration of such circumstances, and one of the objects of the present invention is to provide an information processing device, an information processing method, and a program capable of increasing the variety of contents to be provided. ..

One aspect of the present invention includes a first generation unit that generates vector data corresponding to each of a plurality of candidate contents distributed via a network, and a plurality of vector data generated by the first generation unit. Based on this, it is an information processing device including a determination unit that determines the content to be distributed from the plurality of contents so as to increase the diversity of the contents.

According to one aspect of the present invention, it is possible to increase the variety of contents to be provided.

It is a figure explaining the diversity of contents which concerns on embodiment. It is a functional block diagram which shows the functional structure of the service providing apparatus 20 which concerns on embodiment. It is a figure which conceptually shows the content of the process of the article vector generation part 25 which concerns on embodiment. It is a figure which shows an example of the content of the distribution article information 33 which concerns on embodiment. It is a figure which shows an example of the content of the article vector information 35 which concerns on embodiment. It is a flowchart which shows an example of the processing flow of the service providing apparatus 20 which concerns on embodiment. It is a figure explaining the distance between the position of the 3rd candidate article in the vector space in embodiment, and the subspace composed of the 1st and 2nd article vectors. It is a figure explaining the distance between the position of the 4th candidate article in the vector space in embodiment, and the subspace composed of the 1st to 3rd article vectors.

Hereinafter, embodiments of the information processing apparatus, information processing method, and program of the present invention will be described with reference to the drawings.

[Overview]
The information processing device (service providing device) of the embodiment communicates with a web server device that provides a web page corresponding to a request from a terminal device operated by a user, and a terminal device in which an application is started to obtain various information. It is an application server device or the like that delivers and provides an application page. These web pages and application pages are pages constituting a search site, a news site, a shopping site, an auction site, a weather forecast site, an SNS (Social Networking Service) site, a game site, and the like.

The above-mentioned web page or application page includes a plurality of articles such as news of various genres and a plurality of contents such as advertisements (electronic advertisements). An article is an electronic article provided via a network such as the Internet, and includes text information. The information processing device selects a plurality of contents so that the diversity of the contents included in the contents is maintained, and generates a web page or an application page containing the selected plurality of contents. In the following, an example in which the content is an article will be described.

[Diversity]
"Diversity" refers to the degree to which the content of an article covers a wide range of categories when there are multiple articles. When multiple articles are posted on a web page, it is estimated that the greater the variety of articles, the wider the range of users who are interested in these articles. On the other hand, it is presumed that the less diverse the articles, that is, the more the content of the articles is concentrated in similar categories, the narrower the range of users who are interested in these articles.

Diversity is quantified, for example, based on article vectors. FIG. 1 is a diagram illustrating diversity. Here, in order to represent the characteristics of the content of the article, multidimensional vector data defined from a plurality of viewpoints is used. In FIG. 1, it is assumed that the features of the article are represented by the elements 1 and 2 constituting the two-dimensional vector. Element 1 and element 2 show characteristics of different viewpoints from each other. Element 1 and element 2 are vector representations of, for example, the characteristics of the viewpoint of "immediateness" when the article is news, and the characteristics of the viewpoint of "type" of sports such as "baseball" or "soccer". be.

In FIG. 1, U1 shows a set of users who referred to articles belonging to the first category, and U2 shows a set of users who referred to articles belonging to the second category. The set U1 and the set U2 overlap each other in the portion of the range R1. When selecting two articles as articles to be posted on a web page, select the articles so that the overlap range R1 becomes small. This can increase the variety of articles on the web page as a whole.

Next, consider a case where an article belonging to the first category and an article belonging to the second category are selected as articles to be posted on a web page, and a third article is selected. Articles belonging to the first category and articles belonging to the second category are arranged side by side in the direction of element 1. That is, the variety of articles related to element 1 is ensured between the articles belonging to the first category and the articles belonging to the second category. Here, when the third article is selected, it is different from the article located in the direction of the element 1 (for example, the article belonging to the third category referenced by the user belonging to the set U3). Selecting an article located in the direction of element 2 (for example, an article belonging to the fourth category referenced by a user belonging to the set U4) increases the diversity of the web page as a whole. That is, diversity is evaluated to be higher not only in the range of article overlap, but also in the direction of other elements when overlap in the reference element has already appeared. The information processing device selects a plurality of articles so as to increase the diversity in this way.

[overall structure]
FIG. 2 is a functional block diagram showing a functional configuration of the service providing device 20. The service providing device 20 is connected to the network NW. In addition to the service providing device 20, for example, at least one terminal device 10 is connected to the network NW. The network NW includes, for example, a part or all of the Internet, WAN (Wide Area Network), LAN (Local Area Network), provider equipment, wireless base station, dedicated line, and the like.

The terminal device 10 is operated by a user who uses a service such as a web page provided by the service providing device 20. The terminal device 10 is, for example, a personal computer, a mobile phone such as a smartphone, a tablet terminal, a PDA (Personal Digital Assistant), or the like.

For example, when the terminal device 10 receives a predetermined operation from the user, the terminal device 10 displays a web page provided by the service providing device 20 by a browser installed in advance. This web page contains multiple articles (eg, at least three articles). Further, when the terminal device 10 receives a predetermined operation from the user, the terminal device 10 communicates with the service providing device 20 via the application installed in advance, and acquires the content to be displayed or reproduced on the application. The content is, for example, video data, image data, audio data, text data, or the like, and includes a plurality of articles. As a result, the terminal device 10 is provided with a service similar to the service provided by the various web pages described above via the application. In the following, an example in which the service providing device 20 provides a web page to the terminal device 10 will be described.

The service providing device 20 is, for example, a communication unit 21, a candidate article extraction unit 23, an article vector generation unit 25 (first generation unit), an article determination unit 27 (calculation unit, determination unit), and a page generation unit 29. (Second generation unit) and a storage unit 31 are provided. The communication unit 21 includes, for example, a communication interface such as a NIC. The communication unit 21 communicates with the terminal device 10 and the like via the network NW.

The candidate article extraction unit 23 extracts a predetermined number of candidate articles from the distribution article information 33 stored in the storage unit 31. For example, when the service provided by the service providing device 20 is provided as a customized service for individual users by inputting a user ID and a password and logging in, the candidate article extraction unit 23 may use the candidate article extraction unit 23. Articles that are presumed to be of interest to the user based on the user's past access history information or the like may be extracted as candidate articles. Alternatively, the candidate article extraction unit 23 may randomly extract articles as candidate articles from the distributed article information 33.

The article vector generation unit 25 generates, for example, an article vector (feature information) corresponding to the candidate article extracted by the candidate article extraction unit 23. The article vector is vector data generated by, for example, extracting words contained in an article by morphological analysis or the like, and then inputting a word vector distributed and expressed by a method such as word2vec into a model.

FIG. 3 is a diagram conceptually showing the contents of the processing of the article vector generation unit 25. The content (text) of the candidate article extracted by the candidate article extraction unit 23 is input to the article vector generation unit 25.

The article vector generation unit 25 first performs morphological analysis or the like on the content of the article and divides it into words (morphemes). Next, the article vector generation unit 25 applies a method such as word2vec to each word to acquire a distributed expression word vector. One or both of the morphological analysis and word2vec may be realized by requesting the processing to an external device instead of the article vector generation unit 25 performing the processing by itself.

Next, the article vector generation unit 25 generates an article vector by inputting one or more word vectors acquired from the article into a predefined model. This model is, for example, a model generated by machine learning. This model is generated by finding the parameters of the activation function of the hidden layer so that the relationship between the known word vector and the article vector can be reproduced by, for example, a method such as deep learning. Regardless of the above definition, the first model may include the identification of word2vec, that is, what kind of corpus is given to word2vec. The article vector generation unit 25 stores the generated article vector as the article vector information 35 in the storage unit 31.

The article determination unit 27 determines a plurality of articles (for example, at least three articles) to be distributed based on each article vector of the candidate articles generated by the article vector generation unit 25. The article determination unit 27 determines an article so as to increase the diversity of the article among a predetermined number of combinations of articles. The article determination unit 27 preferentially determines the article having the maximum distance from the subspace composed of the article vectors of the already selected articles as the distribution target.

The page generation unit 29 generates a web page including a plurality of articles determined by the article determination unit 27. That is, the page generation unit 29 generates display information for displaying the content to be distributed determined by the article determination unit 27 on the display device. The page generation unit 29 transmits the generated web page to the terminal device 10 via the communication unit 21.

Each of the above functional units of the service providing device 20 is realized by, for example, a hardware processor such as a CPU (Central Processing Unit) executing a program (software). In addition, some or all of these components are hardware (circuits) such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), GPU (Graphics Processing Unit), etc. It may be realized by the part; including circuitry), or it may be realized by the cooperation of software and hardware.

The storage unit 31 stores, for example, the delivered article information 33 and the article vector information 35. FIG. 4 is a diagram showing an example of the contents of the delivered article information 33. The delivered article information 33 is, for example, information in which article data is associated with an article ID, which is article identification information. The article data is, for example, text data, image data, a combination thereof, or reference information such as a URL (Uniform Resource Locator).

FIG. 5 is a diagram showing an example of the contents of the article vector information 35. The article vector information 35 is, for example, information in which an article vector is associated with an article ID which is an article identification information.

The storage unit 31 is realized by, for example, a RAM (Random Access Memory), a ROM (Read Only Memory), an HDD (Hard Disk Drive), a flash memory, or a hybrid storage device in which a plurality of these are combined. Further, the storage unit 31 may be an external device such as NAS or an external storage server that can be accessed by the service providing device 20.

[Operation of service providing device 20]
Hereinafter, the operation of the service providing device 20 will be described. FIG. 6 is a flowchart showing an example of the processing flow of the service providing device 20. In the following, an example in which the service providing device 20 generates a web page containing three articles and transmits the web page to the terminal device 10 will be described.

First, the service providing device 20 receives a request for a web page transmitted from the terminal device 10 (S101).

Next, the service providing device 20 extracts two candidate articles (hereinafter, referred to as "first candidate article" and "second candidate article") from the distribution article information 33 stored in the storage unit 31. (S103). For example, the service providing device 20 extracts the first candidate article and the second candidate article that are presumed to be of interest to the user based on the user's past access history information and the like. Further, the service providing device 20 determines the extracted first and second candidate articles as the first and second articles to be posted on the web page.

Next, the service providing device 20 generates an article vector corresponding to each of the determined first article and the second article (S105). For example, the service providing device 20 performs morphological analysis or the like on the content of the first article and divides it into words. Next, the service providing device 20 applies a method such as word2vec to each word to acquire a distributed expression word vector. Next, the service providing device 20 generates an article vector (hereinafter referred to as "first article vector") by inputting one or more acquired word vectors into a predefined model. The service providing device 20 performs the same processing on the second article to generate an article vector (hereinafter, referred to as “second article vector”).

When the article vector information 35 of the storage unit 31 stores in advance the article vectors corresponding to each of the articles included in the delivered article information 33, the service providing device 20 has extracted the first article and the second article. The article vector corresponding to each of the articles may be read from the article vector information 35.

Next, the service providing device 20 extracts a third candidate article other than the first article and the second article from the delivered article information 33 stored in the storage unit 31 (S107). Next, the service providing device 20 generates an article vector (hereinafter, referred to as “third article vector”) corresponding to the extracted third candidate article (S109). Next, the service providing device 20 calculates the distance between the position of the third candidate article on the vector space and the subspace composed of the first and second article vectors (S111). For example, the service providing device 20 repeats the above processing for all articles other than the first article and the second article included in the delivered article information 33. Alternatively, the service providing device 20 may perform the above-mentioned processing on articles belonging to a predetermined category presumed to be of interest to the user among the articles included in the delivered article information 33.

FIG. 7 is a diagram illustrating the distance between the position of the third candidate article on the vector space and the subspace composed of the first and second article vectors. In FIG. 7, each article is represented in a two-dimensional vector space with elements 1 and 2. The subspace composed of the first article vector and the second article vector is a one-dimensional subspace P1. The third candidate article is selected from articles located in the space outside this subspace P1. Therefore, the article A0 located in the subspace P1 is not selected as the third candidate article.

The service providing device 20 calculates the shortest distance between the position of the third candidate article in the vector space and the subspace P1 in this two-dimensional vector space. For example, the distance D1 between the third candidate article A1 and the subspace P1, which is one of the third candidate articles, corresponds to the length of the perpendicular line drawn from the third candidate article A1 to the subspace P1. Further, the distance D2 between the third candidate article A2 and the subspace P1, which is one of the other third candidate articles, corresponds to the distance between the third candidate article A2 and the second article. Further, the distance D3 between the third candidate article A3 and the subspace P1, which is still another one of the third candidate articles, corresponds to the length of the perpendicular line drawn from the third candidate article A3 to the subspace P1.

Next, the service providing device 20 determines the third candidate article having the longest distance to the calculated subspace as the third article to be posted on the web page (S113). In the example shown in FIG. 7, the distance D1 is the longest among the calculated distances D1 to D3. In this case, the service providing device 20 determines the third candidate article A1 as the third article.

Next, the service providing device 20 generates a web page including the determined first to third articles, and transmits the generated web page to the terminal device 10 (S115). The terminal device 10 displays a web page including the first to third articles received from the service providing device 20 by the function of the web browser. As a result, the service providing device 20 ends the processing of this flowchart.

In the above-described embodiment, an example of determining three articles has been described, but in addition to these three articles, a fourth article (hereinafter referred to as "fourth candidate article") is determined. The same process can be applied to the above cases. FIG. 8 is a diagram illustrating the distance between the position of the fourth candidate article on the vector space and the subspace composed of the first to third article vectors. The subspace composed of the first to third article vectors is a two-dimensional subspace P2. The fourth candidate article is selected from articles located in the space outside this subspace P2. Therefore, the article A0 located in the subspace P2 is not selected as the fourth candidate article.

The service providing device 20 determines the shortest distance between the position of the fourth candidate article extracted from the delivered article information 33 and the subspace P2 in the two-dimensional vector space composed of the elements 1 and 2. calculate. For example, the distance D4 between the fourth candidate article A4 and the subspace P2, which is one of the fourth candidate articles, corresponds to the length of the perpendicular line drawn from the fourth candidate article A4 to the subspace P2. Further, the distance D5 between the fourth candidate article A5 and the subspace P2, which is one of the other fourth candidate articles, corresponds to the distance between the fourth candidate article A5 and the third article. In this case, the service providing device 20 determines the fourth candidate article A5 having the longest distance to the calculated subspace as the fourth article to be posted on the web page.

Further, when the service providing device 20 determines the n + 1th n + 1 candidate article, the service providing device 20 calculates the distance between the subspace Pn composed of the selected n candidates and the n + 1th candidate article. , The n + 1 candidate article having the longest distance from the subspace Pn is determined as the n + 1 article to be posted on the web page.

In this case, the distance between the n + 1 candidate article and the subspace Pn composed of the first to nth article vectors is expressed by the following equation.

The above equation (1) shows an article vector corresponding to n selected articles. The subspace Pn composed of the article vectors corresponding to the selected n articles is represented by the equation (2) which is the linear sum of these article vectors. α _k indicates a parameter. The distance between the position of the article x, which is the n + 1 candidate article selected from the space outside the subspace Pn, and the subspace Pn is selected from the position vector of the article x by α _k , which minimizes the distance to the article x. It is expressed by the equation (3) for which the norm is obtained by subtracting the position vector in the subspace Pn in the case of. Based on these equations, the service providing device 20 can determine the n + 1 article.

The information processing apparatus according to the embodiment of the present invention has a first generation unit that generates vector data corresponding to each of a plurality of candidate contents distributed via a network, and a plurality of generation units generated by the first generation unit. It is possible to increase the diversity of the content to be provided by providing a decision unit for determining the content to be distributed from the plurality of contents so as to increase the diversity of the content based on the vector data of. can.

Although the embodiments for carrying out the present invention have been described above with reference to the embodiments, the present invention is not limited to these embodiments, and various modifications and substitutions are made without departing from the gist of the present invention. Can be added.

10 ... Terminal device 20 ... Service providing device 21 ... Communication unit 23 ... Candidate article extraction unit 25 ... Article vector generation unit 27 ... Article determination unit 29 ... Page generation unit 31 ... Storage unit

Claims (6)

A first generation unit that generates vector data corresponding to each of a plurality of candidate contents distributed via a network, and a first generation unit.
A determination unit that determines the content to be distributed from the plurality of contents so as to increase the diversity of the contents based on the plurality of vector data generated by the first generation unit.
The distance between the subspace composed of the plurality of vector data corresponding to each of the plurality of contents to be distributed already determined by the determination unit and each of the plurality of candidate contents to be distributed is calculated. Calculation unit and
Equipped with
The determination unit determines the content to be distributed based on the plurality of distances calculated by the calculation unit.
Information processing equipment. The determination unit determines the content having the maximum distance calculated by the calculation unit as the content to be distributed.
The information processing apparatus according to claim 1. The content is an article distributed via the network.
The vector data is information showing the characteristics of the article, and is
The determination unit determines an article having characteristics different from the already determined article as an article to be distributed.
The information processing apparatus according to claim 1 or 2. A second generation unit for generating display information for displaying the content to be distributed determined by the determination unit on the display device is further provided.
The information processing apparatus according to any one of claims 1 to 3. The computer
Generate vector data corresponding to each of multiple candidate contents to be delivered via the network.
Based on the plurality of generated vector data, the content to be distributed is determined from the plurality of contents so as to increase the diversity of the contents.
It is an information processing method
The distance between the subspace composed of the plurality of vector data corresponding to each of the plurality of content to be distributed already determined and the plurality of content to be distributed is calculated.
The content to be delivered is determined based on the calculated distances.
Information processing method. On the computer
Generate vector data corresponding to each of multiple candidate contents to be delivered via the network.
Based on the plurality of generated vector data, the content to be distributed is determined from the plurality of contents so as to increase the diversity of the contents.
It ’s a program,
The distance between the subspace composed of the plurality of vector data corresponding to each of the plurality of content to be distributed that has already been determined and the plurality of content that is candidates for distribution is calculated.
The content to be delivered is determined based on the calculated distances.
program.

Images