JP6341899B2 - Advertisement system and advertisement delivery method - Google Patents

Description

  The present invention relates to an advertisement system and an advertisement distribution method, and more particularly, to a technique for grasping signs of a trend and efficiently distributing advertisements.

  In the technical field of so-called online advertising, hitting an advertisement that matches the preference and interest of the recipient of the advertisement has been conventionally performed (for example, Patent Document 1). Patent Document 1 describes a technique for determining a more appropriate target advertisement using attributes stored in a local storage.

JP 2014-085812 A

  In this specification, a group of users having a predetermined preference or interest is referred to as a “user segment”. The user segment is a target for distributing advertisements. The user segment is defined as, for example, “beer lover”, “professional baseball team A fan”, “30's male”, and the like.

  In the conventional advertisement distribution technology, users are extracted from a large number of unspecified users such as “male in their 30s”, “male in A prefecture”, etc., and advertisements are made to the extracted user group (user segment). To deliver. Actually, personal information for users is obtained through questionnaires, etc., and the interests of each user segment as described above are analyzed, and advertisements are distributed to user segments that are highly likely to be interested in the product to be advertised. Was always.

  However, in the conventional advertisement distribution technology, even if a user segment can be specified, it is difficult to predict what product demand will be generated in the user segment in the future.

  For example, assuming a user segment that can be named “high teen girls who are sensitive to fashion”, it is difficult to predict what product group will be popular in this user segment. Each user in the user segment may be sending information via social networking services (hereinafter referred to as “SNS”) or weblogs. It may be known whether a group is in fashion. However, even if an article that is already in fashion is being followed up, there is a possibility that the fashion will already be delayed.

  The present invention has been made in view of the above-described problems, and aims to detect signs of epidemic at an early stage and efficiently distribute advertisements to user segments that are likely to accept the epidemic. And

  One aspect of the present invention for achieving the above object is an advertising system including an advertising server and a plurality of user terminals, wherein the advertising server includes at least character information acquired from information transmission on the Internet or cloud. A first frequency dictionary indicating the frequency of words appearing in a line; a first analysis means for defining a user segment to which an information sender related to the timeline belongs based on the first frequency dictionary; Second analysis means for acquiring a word with a sign of epidemic as a product definition keyword based on a temporal change in the first frequency dictionary, and advertisement information acquisition means for acquiring advertisement information based on the product definition keyword And the user terminal is a second frequency dictionary generated in response to information consumption of the user of the user terminal using the user terminal. And advertisement selection means for selecting the advertisement acquired by the advertisement information acquisition means for display according to the strength of association between the user segment defined by the first analysis means and the second frequency dictionary; And an advertising system.

  ADVANTAGE OF THE INVENTION According to this invention, it becomes possible to discover the sign of the epidemic early and to deliver an advertisement efficiently to the user segment with high possibility of accepting the epidemic.

It is a conceptual diagram which shows schematic structure of the 1st Embodiment of this invention. It is a block diagram which shows the function structure of the said embodiment. It is a flowchart figure which shows the procedure of the process of the advertisement server 10 of the said embodiment. It is a flowchart figure which shows the procedure of the process of the user terminal 20 of the said embodiment. It is a block diagram which shows the function structure of the 2nd Embodiment of this invention. It is a flowchart figure which shows the procedure of the process of the advertisement server 10 of the said embodiment.

<First Embodiment>
FIG. 1 shows a schematic configuration of the present embodiment. As illustrated, the advertising system 1 according to the present embodiment includes an advertising server 10, a user terminal 20, and an advertising information providing server 30. There are a plurality of user terminals 20.

  An outline of advertisement distribution in this embodiment will be described. In FIG. 1, the advertisement server 10 acquires information (preferably text information) that is publicly transmitted within a public or a specific viewer via SNS or web log on the Internet or cloud as needed, such as by crawling by a search engine. doing.

  The information to be acquired is preferably in a mode in which utterances are accumulated on a time series. In the present specification, such information is hereinafter referred to as a “timeline”. The advertisement server 10 analyzes the timeline and generates a frequency dictionary indicating the frequency of words included in the timeline. The frequency dictionary is used for grasping a user segment, and used to detect what kind of epidemic is generated in the user segment, that is, a word having a sign of epidemic.

  The advertisement server 10 acquires advertisement information of a product item that is likely to be popular from the advertisement information providing server 30 based on a word having a trend sign. The advertisement information is distributed to the user terminal 20 together with the grasped user segment information. Each user terminal 20 performs matching between each user's interest dictionary and user segment information, and an advertisement is displayed when matching is performed.

  FIG. 2 shows a more detailed functional configuration of the present embodiment. As shown in the figure, the advertisement server 10 includes user segment analysis means 11, fashion analysis means 12, word frequency dictionary generation means 13, word frequency dictionary 14, by information processing by a software program using computer resources of the advertisement server 10. An advertisement information acquisition unit 15 is configured. Similarly, the user terminal 20 includes an advertisement selection unit 21, a display unit 22, a user interest dictionary 23, and an interest dictionary generation unit 24 by information processing using a software program using the computer resources of the user terminal 20.

  Details of the operation of each means will be described below with reference to FIGS. FIG. 3 is a flowchart showing a processing procedure of the advertisement server 10, and FIG. 4 is a flowchart showing a processing procedure of the user terminal 20.

  In FIG. 3, the advertisement server 10 first acquires a timeline such as an SNS or web log on the Internet or cloud transmitted by a desired caller (S101). The timeline of the desired caller acquired here is called a “sample timeline”. Selection of a desired sender is performed by an operator of the advertisement server 10 in this embodiment.

  In the present embodiment, it is assumed that a caller belonging to a user group called “early adapter”, which is a product, a service, a lifestyle, and the like that takes a relatively early stage and forms an initial boom or trend, is a desired caller. Yes. In addition, the spread and influence of the epidemic to the user segment that follows the early adapter is also important. Therefore, when selecting a desired sender (sample timeline), the number of links in the timeline (the total number of trackbacks, the number of readers in SNS, etc.) is used as an index. Note that the advertisement server 10 can assist the selection of a desired sender (sample timeline). For example, the SNS and the web log of a large number of callers are crawled or monitored, and the desired caller is selected from the number of links (the total number of trackbacks, the number of readers in the SNS, etc.) as an index. Can be displayed as candidates.

  Next, the word frequency dictionary generating means 13 analyzes the sample timeline and generates a frequency dictionary (word frequency dictionary 14) of words appearing in the sample timeline (S102). Specifically, by performing morphological analysis on the timeline, only nouns are extracted and frequency information of each noun is obtained. Also, the adjectives used with these nouns are analyzed, and whether the nouns have a positive meaning or a negative meaning is analyzed.

  The above analysis result is a word frequency dictionary 14. When selecting a plurality of desired senders who are considered to belong to the same user segment and making a plurality of sample timelines, the plurality of word frequency dictionaries 14 are handled in an integrated manner according to the plurality of sample timelines accordingly. A single word frequency dictionary may be generated.

  Each word as a feature of the word contained in the word frequency dictionary 14 related to the sample timeline by comparing the frequency dictionary of another desired caller that seems to belong to a different user segment from a certain desired caller Get the vector bias. A plurality of early adapter users may be compared with each other, or may be compared with user segments having the most different characteristics.

  In addition, when the target user segment is very special, a frequency dictionary is generated from a timeline of a sender who is not specified, for example, randomly selected, and is used as a contrast frequency dictionary. The contrast frequency dictionary may acquire an existing corpus before processing. By comparing the word frequency dictionary 14 and the contrast frequency dictionary, a method of obtaining the vector bias of each word as a feature of the word included in the word frequency dictionary 14 related to the sample timeline may be used.

  The process for obtaining the vector bias thus obtained is performed by the user segment analysis means 11, and the analysis of the tendency of the frequency dictionary (S103) in FIG. 3 corresponds to this process. The user segment analysis means 11 acquires a word group characterizing this bias as a segment definition keyword group (S104). For example, a word having a significantly higher appearance frequency than the contrast frequency dictionary is set as a segment definition keyword group. Not only words with high frequency but also words with low appearance frequency may be used. In that case, it distinguishes and uses high and low at the time of collation.

  Next, the trend analysis means 12 analyzes the temporal change of the word frequency dictionary 14 (S105). This analysis, for example, compares the frequency information from the timeline of the caller of the user layer that follows and imitates the product, service, and lifestyle sent by a specific “early adapter” called “follower”, and the latest sample time A word whose appearance frequency has increased rapidly on the line is recognized as a word with a trend sign, and is acquired as a product definition keyword (S106). There may be one or more words with signs of epidemic. The follower may be selected by the operator, or an average value of followers on the SNS of the early adapter may be used.

Next, the advertisement information acquisition unit 15 acquires advertisement information that matches the product definition keyword group from the advertisement information providing server 30 (S107). In the advertisement information database that the advertisement information providing server 30 has, the product and its advertisement, and keywords that define the product (for example, words describing the product such as “autumn” or “cute”) are used as the product description. Alternatively, the advertisement information providing server 30 provides advertisement information according to the number of matches with the product definition keyword group and the level of similarity.
Naturally, when learning advertisement information using the product definition keyword group, a similar word dictionary may be used to eliminate non-conformity caused by fluctuation of the notation.

  Next, the advertisement server 10 distributes the segment definition keyword group acquired in S104 and the advertisement information acquired in S107 to each user terminal 20. In this distribution, a pointer to the segment definition keyword group and a pointer to the advertisement information may be transmitted. The advertisement information may include the advertisement itself, but the advertisement ID may be distributed and the reference based on the ID may be received from the user terminal 20. The same can be said for the segment definition keywords.

  Reference is now made to FIG. In each of the plurality of user terminals 20, the advertisement selection unit 21 matches the segment definition keyword group received from the advertisement server 10 with the user interest dictionary 23 of the user of the user terminal 20 (S201). The user interest dictionary 23 is generated by the interest dictionary generation means 24, and preferably an activity that is an operation history of the user terminal 20 by the user (for example, browsing of websites and news articles on the user terminal 20 and information search history). , Information consumption such as playback history of moving images and music). The user interest dictionary 23 created by the operation history of the user terminal 20 by the user reflects the information that the user likes.

  The manner in which the user interest dictionary 23 is generated by the interest dictionary generating means 24 is more specifically described in the contents and news articles of the website describing the natural language consumed by the user of the user terminal 20 or the video reproduced by the user. Words are extracted by performing morphological analysis from information specifying music and music, a keyword history used when searching for information, and the like and collecting them in a dictionary. By repeating this, frequency information for each word is obtained. The user of the user terminal 20 does not need to send anything on the Internet when generating the user interest dictionary 23.

  In the present embodiment, when matching, a word that characterizes the user interest dictionary 23 by comparing the user interest dictionary 23 of the user of the user terminal 20 with an interest dictionary of other users prepared in advance. (Words that appear frequently) are prepared in advance, and this is matched with the segment definition keyword group.

  If there is no matching as a result of S201, the process in the user terminal 20 is terminated (S202). In the case of matching, next, advertisement display is performed (S203). The display means 22 displays the advertisement acquired from the advertisement server 10 in the advertisement display area (S203).

  In the generation of the user interest dictionary 23, the interest dictionary generation means 24 may enrich the user interest dictionary 23 based on the utterances that were made on the timeline such as SNS used by the user of the user terminal 20. .

  According to the above-described embodiment, the signs of the occurrence of the epidemic are grasped by focusing on the timeline transmitted by the epidemic sensitive layer that can be an early adaptor, and grasped toward the user segment that is strongly related to this segment. Deliver ads that are strongly related to signs of epidemic. Everything from trending to distribution is done by automatic information processing. Therefore, it is possible to quickly catch the signs of the occurrence of the epidemic and lead to the distribution of the advertisement, thereby enabling efficient advertisement distribution.

  In addition, according to the present embodiment, the user interest dictionary 23 is automatically generated by consumption of news articles or the like, so that the user does not need to send anything.

<Second Embodiment>
In the present embodiment, in addition to the configuration of the first embodiment, a central user analysis unit 16 is provided (see FIG. 5), and the advertisement server 10 automatically extracts the “sample timeline” described above regardless of the operator. , Which leads to the discovery of words with signs of epidemic. According to the present embodiment, the detection of signs of epidemic is accurately realized earlier than in the first embodiment.

  FIG. 5 shows the functional configuration of this embodiment, and FIG. 6 shows the flow of processing in the advertisement server 10. The description of the same configuration and processing as in the first embodiment, such as processing on the user terminal 20 side, is omitted. The central user analysis unit 16 is a third analysis unit subsequent to the user segment analysis unit 11 and the trend analysis unit 12, and based on the frequency dictionary generated from a large number of timelines and the relationship between the timelines, Identify the central user at

  In FIG. 6, the advertisement server 10 fetches a large number of timelines from web logs, SNSs, short text posting sites, etc. on the Internet (S201). For example, enter all timelines for a domain. Next, a morphological analysis as described in the first embodiment is performed to create a frequency dictionary based on the content (timeline) of each user (S202).

  Next, the user segment analysis unit 11 performs clustering of users based on the timeline acquired in S201 (S203). It is assumed that the user and the timeline are linked one-on-one. In S203, for example, clustering can be performed based on information such as similarity of frequency dictionaries composed of statements within a specific unit time and follow-up relationships on SNS.

  The user segment analysis means 11 further analyzes the follow relationship between users (S204). For example, in SNS, a social network is formed in the service name space, and users are connected with each other through a mutual link or a one-way link, but a place where the relationship between links is dense is extracted. .

  The user segment analysis means 11 extracts a user segment using the user cluster obtained by clustering in S203 and the analysis result in S204. Then, a segment definition keyword group is acquired based on a comparison between the frequency dictionary generated from the timeline of the user belonging to the extracted user segment and the frequency dictionary belonging to another cluster (S205). Like other clusters, high frequency words are weakened and the frequency of unique words is emphasized.

  The segment definition keyword group acquisition in S205 may be performed for each segment. In addition, words that overlap with other segments may be omitted.

  Next, the central user analysis means 16 uses the number of linked users (the total number of trackbacks, the number of readers in the SNS, etc.) from the users belonging to each segment as an index. “Desired sender” in the form is discovered and specified (S206). S206 may be performed for each segment.

  Next, an analysis of signs of epidemic is performed using the content of the message of the central user (desired sender) as a “sample timeline” (S207). The subsequent processing is the same as in the first embodiment described with reference to FIG. 3, but it can be said that the frequency dictionary has already been corrected by comparison with other clusters.

  According to the above-described embodiment, the central user (desired caller) is automatically discovered and specified, and significant clustering is performed. Will be realized accurately at an early stage.

DESCRIPTION OF SYMBOLS 1 Advertising system 10 Advertising server 11 User segment analysis means 12 Trend analysis means 13 Word frequency dictionary production means 14 Word frequency dictionary 15 Advertisement information acquisition means 16 Central user analysis means 20 User terminal 21 Advertisement selection means 22 Display means 23 User interest dictionary 24 Interest dictionary generating means 30 Advertisement information providing server

Claims (6)

An advertisement system including an advertisement server and a plurality of user terminals,
The advertising server
A first frequency dictionary indicating the frequency of words appearing in a sample timeline including at least character information obtained from information transmission on the Internet or cloud;
First analysis means for defining a user segment to which an information sender related to the timeline belongs based on the first frequency dictionary;
Second analysis means for acquiring a word with a sign of epidemic as a product definition keyword based on a temporal change in the first frequency dictionary;
Advertising information acquisition means for acquiring advertising information based on the product definition keyword;
Have
The user terminal is
A second frequency dictionary generated according to the information consumption of the user of the user terminal using the user terminal;
An advertisement selection means for selecting the advertisement acquired by the advertisement information acquisition means for display according to the strength of association between the user segment defined by the first analysis means and the second frequency dictionary;
An advertising system characterized by comprising: The advertising server
A third analysis means for defining a central user for each of the user segments;
The advertising system according to claim 1, wherein the first frequency dictionary targeted by the second analysis unit is limited to the one related to the information transmission history of the central user. The advertising server
The advertisement according to claim 1 or 2, further comprising first frequency dictionary generation means for acquiring the timeline from information transmission on the Internet or cloud and generating the first frequency dictionary. system. The first analyzing means acquires words that frequently appear in the first frequency dictionary as a segment definition keyword group,
The advertisement selection means matches a group of words that appear characteristically in the second frequency dictionary with the segment definition keyword group in a comparison between the second frequency dictionary and another frequency dictionary, and performs matching. The advertisement system according to any one of claims 1 to 3, wherein an advertisement is selected when it is determined. 5. The method according to claim 1, wherein the second analysis unit detects a word that has recently increased in appearance frequency in the first frequency dictionary and acquires it as a product definition keyword. 6. The advertising system described. An advertisement distribution method in an advertisement system including an advertisement server and a plurality of user terminals,
In the advertising server,
Define a user segment to which an information sender related to the timeline belongs based on a first frequency dictionary indicating the frequency of words appearing in a sample timeline including at least character information acquired from information transmission on the Internet or cloud A first analysis step to
A second analysis step of analyzing a trend sign based on a temporal change of the first frequency dictionary and acquiring a word having a trend sign as a product definition keyword;
An advertising information acquisition step of acquiring advertising information based on the product definition keyword;
In the user terminal,
According to the strength of the association between the second frequency dictionary generated according to the information consumption of the user of the user terminal using the user terminal and the user segment defined by the first analysis step, An advertisement selection process for selecting the advertisement acquired in the advertisement information acquisition process for display;
An advertisement distribution method comprising:

Images