TW201935933A - Personalized advertisement allocation method and apparatus using the same - Google Patents

Abstract

A personalized advertisement allocation method for an video streaming system and an apparatus using the same method are proposed. The method includes steps described as follows: retrieving an advertisement from an advertisement shortlist in an advertisement database, and, thereby, retrieving metadata of the advertisement; transforming the metadata into a feature vector and comparing the vector with features explored from the user logs, we can find out a user and his/her personal advertisement list most closely related. Through matching each advertisement with the electronic program guide, we can allocate an advertisement to a user in a possible time slot. We can, therefore, achieve the goal of personal advertising in this way.

Description

Personalized advertising matching method and advertising matching device

The invention relates to a personalized advertisement matching method and an advertisement matching device, and more particularly to a personalized advertisement matching method and an advertisement matching device in an on-demand video and channel system.

In the traditional advertising system, except for the broadcast period specified by the advertiser, the rest is roughly determined by the advertising method and the popularity of the program. These advertisements placed on users are not necessarily what users want to see. Therefore, for users, it is not only a kind of spam, but also may cause visual distress.

In view of this, the present invention proposes a personalized advertisement matching method and an advertisement matching device in an on-demand video and channel system, which can automatically arrange and consider the number of advertisement impressions to configure appropriate advertisements at appropriate time periods to appropriate viewers. To increase viewers ’interest in watching ads. In this way, it can contribute to the development and revenue of related services such as advertising revenue and related product sales.

The invention provides a personalized advertisement matching method in an on-demand video and channel system, which includes: taking out a first advertisement from a plurality of advertisements to be broadcast in an advertisement database, and obtaining interpretation data of the first advertisement accordingly; The foregoing interpretation data is obtained through a conversion to obtain the characteristic data; based on the foregoing characteristic data, a first user and an advertisement list of the first user related to the foregoing characteristic data are found from a user database and added From the first advertisement to the advertisement list, the user's characteristic data can be obtained by analyzing the user's viewing behavior on the video-on-demand and channel system; through the association analysis between the advertisement list and the program schedule, the advertisement can be configured to a certain Aired on a show. That is, through this method, an advertisement can be configured into a certain program that a user may watch to achieve the purpose of personalized advertisement.

The invention provides an advertisement matching device in an on-demand video and channel system, which includes a storage circuit and a processor. The storage circuit stores multiple modules. The processor is connected to the storage circuit and accesses the foregoing module to perform the following steps: fetching a first advertisement from a plurality of advertisements to be broadcast in an advertisement database, and obtaining the interpretation data of the first advertisement accordingly; A conversion obtains the characteristic data; based on the foregoing characteristic data, a first user and an advertisement list of the first user related to the foregoing characteristic data are found from a user database, and a first advertisement is added to the advertisement In the list, the advertisement list includes a plurality of first pending advertisements corresponding to the first user; analyzing the viewing history of the first user in the on-demand video and channel system; and the foregoing first waiting broadcast according to the viewing behavior record The advertisement is distributed to the first user for viewing.

Based on the above, the personalized advertisement matching method and device in the on-demand video system and channel provided by the present invention can analyze the interpretation data of each advertisement to be broadcasted, and then allocate each advertisement to be broadcasted to users related to its interpretation data. Watch. In this way, users can watch ads that better match their own preferences when using video-on-demand and channel systems, thereby improving the effectiveness of ads.

In order to make the above features and advantages of the present invention more comprehensible, embodiments are described below in detail with reference to the accompanying drawings as follows.

FIG. 1 is a schematic diagram of an advertisement matching device according to an embodiment of the present invention. In this embodiment, the advertisement matching device 100 includes a storage circuit 110, a transceiver circuit 120, and a processor 130. The advertisement matching device 100 may be installed in, for example, an on-demand video and channel system maintained by a telecommunications service provider (for example, a video on-demand (VOD) system and an Internet Protocol Television (IPTV) system, etc.) It can also be used to determine the advertising distribution and advertising matching strategies of multiple video streaming terminal devices (not shown), where each video streaming terminal device is, for example, a MOD deployed in the homes of multiple users or in similar places A set-top box, an IPTV device, and / or an OTT (over-the-top) device, but the present invention may not be limited thereto.

The storage circuit 110 is, for example, a memory, a hard disk, or any other component that can be used to store data, and can be used to record multiple codes or modules. The processor 120 is coupled to the storage circuit 110. The processor 120 is, for example, a general-purpose processor, a special-purpose processor, a conventional processor, a digital signal processor, a plurality of microprocessors, one or more microprocessors combined with a digital signal processor core, and a control unit. Controller, microcontroller, Application Specific Integrated Circuit (ASIC), Field Programmable Gate Array (FPGA), any other kind of integrated circuit, state machine, based on advanced reduced instructions Processors of Advanced RISC Machine (ARM) and the like.

FIG. 2 is a flowchart of a personalized advertisement matching method in an on-demand video and channel system according to an embodiment of the present invention. The method in this embodiment may be executed by the advertisement matching device 100 in FIG. 1, and details of each step of the method are described below with reference to the components shown in FIG. 1.

First, in step S210, the processor 120 retrieves a first advertisement from a plurality of advertisements to be broadcast in an advertisement database, and obtains interpretation data of the first advertisement accordingly. In an embodiment, the advertisements to be broadcast in the advertisement database are, for example, advertisements commissioned by service providers of on-demand video and channel systems commissioned by various advertisers, or advertisements hosted by government agencies, but are not limited thereto. In other embodiments, the interpretation data of each advertisement to be broadcast can be stored in an advertisement interpretation database, and each interpretation data can be edited by relevant personnel according to the content of the advertisement to be broadcast, or automatically by other machine learning algorithms. The location is obtained from the content analysis of the advertisement to be broadcast, but it is not limited thereto. In addition, the content of each interpretation can also be presented in different forms. For example, assuming that the content of an advertisement to be broadcast is about marriage, the corresponding interpretation data may include "marriage", "love", and "love", etc. After conversion, characteristic data such as "marriage", " "Marriage", "Marriage", "Love", "Love", "Love", "Love", etc., but it is not limited to this. In other words, each piece of interpretation data can be a complete word or part of a complete word.

Next, in step S220, the processor 120 finds, from a user database, a first user and a list of the first user's advertisements related to the feature data converted from the foregoing interpretation data based on the first advertisement interpretation data. And add the first ad to the list of ads. The foregoing advertisement list includes a plurality of first to-be-advertised advertisements corresponding to the first user, that is, advertisements that are considered to be related to the first user and are intended to be broadcast to the first user later, but are not limited thereto.

In one embodiment, the user database includes multiple users and each user's preferences for watching on-demand video and channel systems. Based on this, the processor 120 may calculate the similarity between the characteristic data of the first advertisement and the multiple preferences of each user. In addition, when the similarity between the preference of a specific user and the characteristic data of the first advertisement is higher than a preset threshold, the processor 120 may define the specific user as the first related to the characteristic data of the first advertisement. Users, and get a list of ads for first users.

The preferences of each user can be characterized as complete words, such as "love", "marriage", "family", etc., but it is not limited thereto. In addition, the advertisement list of the first user may include, for example, a plurality of advertisements to be broadcasted to the first user, and according to the foregoing teachings, the advertisements in the advertisement list are related to the first user, that is, The similarity between the individual characteristic data and the preferences of the first user is higher than a preset threshold.

In one embodiment, the preferences of each user can be learned from the past viewing behavior of each user. For example, in an initial stage (ie, a stage in which the user's preferences are unknown), the processor 120 may first obtain the interpretation data of each pending advertisement in the current advertisement database. Then, the processor 120 may query the program table and obtain interpretation data of each program. After that, the processor 120 may compare the characteristic data of each program with the characteristic data of each advertisement to be broadcast. If the similarity between the characteristic data of a certain program and the characteristic data of an advertisement to be broadcast is high enough, the processor 120 may schedule the advertisement to be broadcasted to be played in the advertisement period of the program. After the advertisements to be broadcast are inserted into the advertisement period of each program, the processor 120 may observe whether a certain user has actually viewed certain advertisements and programs. If so, the processor 120 may record the interpretation data of these advertisements and programs as the user's preference in the aforementioned user database.

It should be understood that, because the form of interpretation data is generally fragmented and trivial, it is not easy to use for comparison. Based on this, in another embodiment, when executing step S220, the processor 120 may first classify the interpretation data of the first advertisement into multiple advertisement features, and then calculate the relationship between these advertisement features and multiple preferences of each user. Similarity.

Taking the previous example, assuming that the content of an advertisement to be broadcast is about marriage, its corresponding interpretation data may include "marriage", "marriage", "marriage", "love", "love", "love" "," Love ", etc., if they are compared only at the text level, they may be too detailed. In order to achieve a higher level of conceptual comparison, the processor 120 may classify these interpretation data as" marriage "accordingly. "And" love. " Then, when the similarity between the preference of a specific user and the category of the first advertisement is higher than a preset threshold, the processor 120 may define the specific user as the first use related to the characteristic data of the first advertisement. Or get a list of ads for the first user, that is, treating categories as part of the characteristics of the ad. In this way, the foregoing similarity calculation can be made more accurate and reasonable, thereby improving the suitability of subsequent advertising matches.

In addition, during the aforementioned initial stage, the processor 120 may also first classify the interpretation data of each advertisement to be broadcast and each program into a plurality of advertisement features and program features, and then perform similarity calculation to allow each to be broadcast The advertisement can be distributed to the more relevant program period for playing, the details thereof will not be repeated here.

After finding the first user and his advertisement list, the processor 120 may proceed to step S230 to query the first user ’s viewing behavior record in the video-on-demand and channel system, and in step S240, the aforementioned The first pending advertisement is allocated to the first user for viewing. The viewing behavior record includes, for example, multiple programs, multiple viewing periods, viewing cycles, and viewing frequencies that the first user has watched in the on-demand video and channel system, but it is not limited thereto.

In an embodiment, the processor 120 may find a program to be broadcast corresponding to the first user from the program table according to the viewing behavior record, wherein the period of the program to be broadcast includes a plurality of advertisement periods. The aforementioned programs to be broadcast corresponding to the first user are, for example, sequels of programs that the first user has watched, concurrent programs of programs that have been watched, and / or programs of certain specific periods, or similar programs, but It is not limited to this. After that, the processor 120 may configure the first advertisement to be broadcast in the advertisement period of the program to be broadcast.

In this way, when the first user watches a program on the on-demand video and channel system according to his habit, he can see an advertisement that better matches his own preference during the advertising period of the program. In this way, the first user's interest in watching advertisements can be increased, thereby helping to improve advertising revenue and related product sales.

In an embodiment, since each first to-be-advertised advertisement may not exactly fill the advertisement period of the to-be-showed program, so that idle periods may appear in each advertisement period, the processor 120 may configure one or more static periods during the idle period. Ads are viewed by first users to make fuller use of the advertising time.

In other embodiments, in order to increase the matching between the first to-be-advertised advertisement and the to-be-broadcast program, so as to allow the first user to see advertisements of similar nature in the advertisement period of each to-be-broadcast program, the processor 120 may first analyze Multiple program interpretation data of the program to be broadcast. Based on the interpretation data of the program to be broadcast, find out the order of the first and second advertisements to be broadcast. The interpretation data of the second advertisement to be broadcast is similar to the program interpretation data of the program to be broadcast. Degree is higher than a preset threshold. Then, the processor 120 may configure the second advertisement in the advertisement period of the program to be broadcast.

In this way, when the first user is watching a program of a ball game, he can see related advertisements, such as sports goods and sports drinks, during the advertising period of the program. Alternatively, when the first user is watching a food program, he can see related advertisements for restaurants, travel, etc. during the advertising period of the program, but it is not limited to this.

In this way, the willingness of the first user to watch the advertisement can be further increased, thereby improving the effectiveness of the advertisement.

It should be understood that even if the advertisement played matches the preferences of the first user, if the same advertisement is played too frequently, it may lead to resentment by the first user. Therefore, in an embodiment, the processor 120 may obtain the average advertisement view rate of the first user within a preset time interval (for example, a week, a month, or a time length set by other designers), and then based on the average advertisement The view rate determines the number of times the second ad is played in the advertising period, where this number of plays is less than the maximum number of ads played for the program to be broadcast.

In addition, after the first user views an advertisement in the first advertisement to be played, the processor 120 may update the preferences of the first user according to the interpretation data of the advertisement. Specifically, when there is some interpretation data in the interpretation data of this advertisement that does not correspond to the preferences of the first user, the processor 120 may correspondingly add these interpretation data to the preferences of the user and record them in the aforementioned use. In the database; for previous preferences, you can also update their weights and weight them appropriately according to the length of time. With this kind of learning behavior, the advertising matching strategy for the first user can be improved in the future.

To make the above description easier to understand, a specific example is given below to supplement the description. Assume that there are currently three users, namely Xiaoying, Xiaoming and Xiaomei. Some of their viewing behaviors on video-on-demand systems are as follows: Table 1

Based on the characteristics of the programs that each user has watched, as shown in Table 1, the preferences of each user can be obtained. Please refer to FIG. 3, which is, for example, a histogram drawn according to Xiaomei ’s preferences, and which is, for example, a chart obtained by integrating the features of each program. The height of each histogram may correspond to the user ’s preferences. As can be seen from Figure 3, Xiaomei prefers to watch programs featuring "ecology" as the program feature, and "beauty" is the second.

In an embodiment, the histogram of FIG. 3 may also be obtained by integrating the interpretation data of the program. The following description is supplemented by FIG. 4A, FIG. 4B, and FIG. 4C. Among them, FIG. 4B is a histogram of characteristic data. FIG. 4B is a histogram of converting the interpretation data into characteristic data in another part of the program that Xiaoming has watched, and FIG. 4C is a histogram of the program characteristics obtained by classifying FIG. 4A and FIG. As can be seen from FIG. 4A and FIG. 4B, the characteristic data presented by the interpretation data is fragmented and incomplete (such as "big", "three", "big three", "three yuan", etc.). The embodiment teaches the operation of classifying the interpretation data into the characteristics of the program, and FIG. 4A and FIG. 4B can be reorganized as shown in FIG. 4C. For example, "MLB", "NBA", "big three", "professional baseball", "league" and other interpretation data can all be classified as "sports" program features.

As can be seen from FIG. 4C, Xiaoming's preferences roughly include "sports", "cuisine", "landscape", "humanities", "ecology" and "others". Similarly, Xiaoying and Xiaomei's preferences can also be learned and transformed into similar histograms based on the above teachings, and details thereof will not be repeated here.

Based on this, the advertisement matching device 100 of the present invention can distribute advertisements to the above users for viewing according to the aforementioned teachings.

In an embodiment, the processor 120 may use the following formula (1) to calculate the similarity between the first advertisement's interpreted and converted feature data and multiple preferences of a particular user. (1) of which Is a vector formed by the advertising characteristics of the first ad, and Is a vector formed by the preferences of the aforementioned specific user, and Is the aforementioned similarity. when When it is greater than a preset threshold (which can be fine-tuned by the operator according to demand), the processor 120 may determine that the first advertisement is related to the specific user, and may add the first advertisement to the advertisement of this specific user accordingly. List.

In addition, the calculation methods of various similarities (such as the similarity between advertisements and programs) mentioned in the present invention can also be calculated by referring to the principle of formula (1), which will not be repeated here.

In other embodiments, the processor 120 may know the average viewing time of each user in a preset time interval (for example, one week) from the viewing behavior record. If an advertisement must be played N times, it is related to the advertisement. A collection of users is , Played to these users It can be calculated by the following formula (2). (2)

In another embodiment, given that social communication tools are quite developed today, the transmission and sharing of information is quite convenient and fast. Therefore, users may watch some programs that they may not be interested in because of sharing or recommendation by friends. Therefore, the types of programs they watch may change, and their viewing preferences may change accordingly. In response to this change, and taking into account the length of time users watch the program, after regularly analyzing the user's program browsing history, the user's preference weight can be updated using the following formulas (3) and (4): (3) (4) of which It is the j-th update of the i-th keyword in user preferences, and its calculation method is based on the length of time from the day of calculation and is calculated on a daily basis. Is the total number of updates for the ith keyword, Is the weight of the i-th keyword, It is the normalized weight of the i-th keyword. Therefore, user preferences can be changed in response to changes in user watching programs.

In an embodiment, it is assumed that an existing advertisement is intended for Xiaoying, Xiaoming and Xiaomei, and the advertisement features of the advertisement include, for example, "Taiwan", "Landscape", "Humanities", "Historic Sites", and "Sky Lantern" , "Bee Cannon", "Hiking" and "Humanistic". Based on the similarity calculation method taught in the foregoing embodiment, it can be concluded that the similarities between the advertising features of this advertisement and the preferences of Xiaoying, Xiaoming, and Xiaomei are, for example, 0.1, 0.3, and 0.6.

In this case, the processor 120 may determine that these similarities are higher than a preset threshold (for example, 0.3), so Xiaoming and Xiaomei can both be identified as users related to the aforementioned advertisement, and the aforementioned advertisement may be updated. Added to Xiaoming and Xiaomei's advertising list, with a view to playing it for Xiaoming and Xiaomei.

In one embodiment, the processor 120 may adjust the number of times the aforementioned advertisement is played to Xiaoming and Xiaomei based on the similarity calculated above. For example, the processor 120 can allow Xiaoying to watch the advertisement once (below the threshold, but can be set to allow the user to watch the number of times as one time), Xiaoming sees the advertisement twice, and Beauty watched 4 ads.

After the advertisement is played, you can analyze the user ’s browsing data and find that Xiaoying has not watched it, Xiaoming has watched it twice, and Xiaomei has watched it three times. Seconds) or more). Thereby, the processor 120 can update the preferences of each user according to the advertisement characteristics accordingly. Taking Xiaoming as an example, the processor 120 may, for example, adjust the height of the histogram of preferences related to the advertisement feature of this advertisement in FIG. 4C, but the present invention may not be limited thereto.

In summary, the personalized advertisement matching method and its device in the on-demand video and channel system provided by the present invention can be used for advertising on this program when users watch programs on the on-demand video and channel system according to their habits. During the time period, I saw more ads that match my preferences. This can increase users ’interest in watching ads, which can help improve advertising revenue and related merchandise sales. In addition, the present invention can also configure advertisements to be broadcasted during advertisement periods of programs of similar nature, thereby further increasing the user's willingness to watch the advertisements and improving the effectiveness of the advertisements. In addition, the present invention can learn the user's preferences according to the behavior of the user watching the advertisement, so that the advertisement matching strategy for the user can be improved in the future.

Although the present invention has been disclosed as above with the examples, it is not intended to limit the present invention. Any person with ordinary knowledge in the technical field can make some modifications and retouching without departing from the spirit and scope of the present invention. The protection scope of the present invention shall be determined by the scope of the attached patent application.

100‧‧‧ Advertising Matching Device

110‧‧‧Storage Circuit

120‧‧‧ processor

S210 ~ S240‧‧‧step

FIG. 1 is a schematic diagram of an advertisement matching device according to an embodiment of the present invention. FIG. 2 is a flowchart of a personalized advertisement matching method in an on-demand video and channel system according to an embodiment of the present invention. FIG. 3 is a histogram plotted according to the user's preference. FIG. 4A is a histogram of part of the explanatory data of a program that the user has watched. FIG. 4B is a histogram of the interpretation data of another part of the program that the user has watched. FIG. 4C is a histogram of the program characteristics classified from FIG. 4A and FIG. 4B.

Claims (9)

A personalized advertisement matching method in an on-demand video and channel system includes: taking a first advertisement from a plurality of advertisements to be broadcast in an advertisement database, and obtaining a plurality of interpretation data of the first advertisement accordingly; Finding a first user and an advertisement list of the first user related to the interpretation data from a user database based on the interpretation data, and adding the first advertisement to the advertisement list, The advertisement list includes a plurality of first pending advertisements corresponding to the first user; analyzing multiple viewing behavior records of the first user in the on-demand video and channel system; and The first pending advertisement is allocated to the first user for viewing. The method according to item 1 of the scope of patent application, wherein the user database includes a plurality of users and each user's preferences for viewing the video-on-demand and channel system, and based on the interpretation data, the user database is used. The steps of finding out the first user and the advertisement list of the first user in the user database related to the interpretation data include: calculating the interpretation data and the preferences of one of the users. A similarity between them; and when the similarity is higher than a preset threshold, defining the users as the first user related to the interpretation data, and obtaining the advertisement list of the first user . The method according to item 1 of the scope of patent application, wherein the user database includes a plurality of users and each user's preferences for viewing the video-on-demand and channel system, and based on the interpretation data, the user database is used. The steps of finding the first user and the advertisement list of the first user in the database of the user related to the interpretation data include: classifying the interpretation data into a plurality of advertisement characteristics; calculating the advertisement characteristics and A similarity between the preferences of one of the users; and when the similarity is higher than a preset threshold, defining the users as the first user related to the interpretation data And obtain the list of advertisements for the first user. The method according to item 1 of the patent application scope, wherein the viewing behavior record includes multiple programs, multiple viewing periods, a viewing period, and a viewing frequency that the first user has watched in the on-demand video and channel system And assigning the first to-be-advertised advertisements to the first user for viewing according to the viewing behavior record includes: locating a to-be-showed program corresponding to the first user from a program table according to the viewing behavior record , Wherein the period of the to-be-broadcast program includes a plurality of advertisement periods; and configuring the first to-be-advertised advertisements in the advertisement periods. The method according to item 4 of the scope of patent application, wherein when there is an idle period in each of the advertising periods, it further comprises configuring a static advertisement in the idle period. The method according to item 4 of the scope of patent application, wherein the steps of configuring the first pending advertisements in the advertising periods include: analyzing program interpretation data of the pending advertisements; The second to-be-broadcast advertisement is found from the first to-be-broadcast advertisements, and a similarity between the second-to-be-broadcast advertisements and the program interpretation data is higher than one. Preset a threshold value; and configure the second advertisement in at least one of the advertisement periods. The method of claim 6, wherein the step of configuring the second advertisement in at least one of the advertisement periods includes: obtaining an average of the first user within a preset time interval; Ad view rate; based on the average ad view rate, determine the number of times the second ad is played during the advertising periods, where the number of times of play is less than the maximum number of times the ad is played. According to the method described in item 1 of the scope of the patent application, the user database includes multiple users and each user's multiple preferences for viewing the on-demand video and channel system, and the method further includes: After a user has viewed the first pending advertisements, the preferences of the first user are updated according to the interpretation data of the advertisement. An advertisement matching device in a video-on-demand system includes: a storage circuit storing a plurality of modules; a processor connected to the storage circuit and accessing the modules to perform the following steps: from an advertisement database Take out a first advertisement from a plurality of advertisements to be broadcast, and accordingly obtain a plurality of interpretation data of the first advertisement; based on the interpretation data, find a first related to the interpretation data from a user database An advertisement list of the user and the first user, and adding the first advertisement to the advertisement list, where the advertisement list includes a plurality of first pending advertisements corresponding to the first user; query the first A viewing record of a user in the on-demand video and channel system; and allocating the first pending advertisements to the first user for viewing according to the viewing behavior record.