Prediction device and prediction method

Abstract

A prediction device and a prediction method are provided that are capable of predicting a popularity index value for a second content created based on a first content and ensuring the reliability of the prediction. [Solution] A prediction device according to one embodiment of the present invention includes a processor, which obtains first information about the content of the first content from the name of the first content, and inputs the first information into a prediction model that corresponds to the correspondence between information about the content and an index value related to the popularity of the content, thereby predicting an index value for a second content created based on the first content.

Description

[Technical Field]

[0001] The present invention relates to a prediction device and a prediction method for predicting an index value relating to the popularity of target content. [Background technology]

[0002] In businesses that handle content, it is important to predict index values ​​related to the popularity of the content when making decisions about investments in content production, planning projects, etc. For example, when the content is a television program, the audience rating of the television program is predicted before it is broadcast, and the predicted audience rating is sometimes used as an index value representing the popularity (expectations) of the program (see, for example, Patent Document 1).

[0003] According to the technology described in Patent Document 1, it is possible to predict the viewership rating of a target program based on the broadcasting station, broadcast date and time, program title, and program content of programs related to the target program that have been broadcast in the past. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2024-10756 Summary of the Invention [Problem to be solved by the invention]

[0005] Incidentally, when creating content, there are cases where new secondary content is created based on existing content. Hereinafter, the original content will be referred to as the "primary content," and the secondary content created based on it will be referred to as the "secondary content."

[0006] When predicting the popularity index value of a second content, typically, information about the second content itself is obtained and the index value is predicted based on that information, but when the second content has not yet been made available (released) to the public, it may be difficult to obtain sufficient information about the second content itself. In such cases, it is difficult to predict the popularity index value of the second content with high reliability using conventional prediction techniques, including the technique described in Patent Document 1.

[0007] One embodiment of the present invention has been made in consideration of the above circumstances, and aims to provide a prediction device and a prediction method that can predict a popularity index value for a second content created based on a first content, and ensure the reliability of the prediction. [Means for solving the problem]

[0008] In order to achieve the above object, a prediction device according to one embodiment of the present invention includes a processor, which obtains first information relating to the content of the first content from the name of the first content, and inputs the first information into a prediction model according to the correspondence between information relating to the content and an index value relating to the popularity of the content, thereby predicting an index value for a second content created based on the first content. According to the prediction device of one embodiment of the present invention configured as described above, it is possible to predict the popularity index value of a second content created based on a first content based on first information obtained from the name of the first content. Furthermore, since the prediction is made using a prediction model according to the correspondence between information about the content and the index value related to the popularity of the content, it is possible to ensure the reliability of the prediction.

[0009] In the prediction device of the present invention, the processor may calculate a predicted value of the degree of contact with the second content as the index value. According to the above configuration, since the degree of contact with content is predicted as the index value to be predicted, it is possible to obtain more realistic prediction results that are useful in business.

[0010] In the prediction device of the present invention, the processor may acquire the first information by inputting the name of the first content into a generative model that generates information about the content details from the name of the content. According to the above configuration, by using the above generation model, it is possible to automatically and easily (without effort) acquire the first information from the name of the first content.

[0011] The first content may be content that has been provided in the past. In this case, the first information may include at least one of information regarding the time when the first content was provided, the medium through which the first content was provided, and information regarding the attributes of users who came into contact with the first content. According to the above configuration, since the first content is content that has already been provided (i.e., existing content), it is possible to more easily acquire the first information about the details of the first content. Furthermore, by using information about the time of providing the first content, the medium of providing it, and the attributes of the people who came into contact with it, it is possible to more accurately predict the popularity index value of the second content.

[0012] Furthermore, in the prediction device of the present invention, it is preferable that the processor acquires second information regarding at least one of the time and medium of provision of the second content, and predicts an index value for the second content by inputting the acquired first information and second information into a prediction model. According to the above configuration, the popularity index value of the second content is predicted taking into consideration the time and medium in which the second content is provided, thereby further improving the prediction accuracy.

[0013] In the prediction device of the present invention, the processor may predict the index value for the second content before the second content is provided. According to the above configuration, the popularity index value of the second content can be known before the second content is produced, at the release planning stage, etc. As a result, for example, it is possible to avoid risks and optimize costs in business related to the second content.

[0014] In addition, in the prediction device of the present invention, the prediction model may be a model constructed by machine learning using information about the content that has already been provided and index values ​​identified for the content that has already been provided as learning data. According to the above configuration, a prediction model is constructed by machine learning using information (actual data) obtained about content that has already been provided, and by using this prediction model, more reliable predictions can be made.

[0015] In addition, in order to solve the above-mentioned problems, the prediction method of the present invention is characterized in that a processor obtains first information regarding the content of the first content from the name of the first content, and the processor inputs the first information into a prediction model that corresponds to the correspondence between information regarding the content and an index value regarding the popularity of the content, thereby predicting an index value for a second content created based on the first content. According to the above method, it is possible to predict a popularity index value for a second content created based on a first content, and also ensure the reliability of the prediction. [Effects of the Invention]

[0016] According to the present invention, it is possible to obtain information (first information) related to the content of a first content from limited information such as the name of the first content, and to predict the popularity index value of another content (second content) to be generated in the future using the first information. Furthermore, since the prediction is made using a prediction model according to the correspondence between information about the content and the index value related to the popularity of the content, it is possible to ensure the reliability of the prediction. [Brief explanation of the drawings]

[0017] [Figure 1] FIG. 1 is a conceptual diagram of a prediction model used in one embodiment of the present invention. [Figure 2] FIG. 1 is a diagram illustrating an example of the configuration of a prediction device according to an embodiment of the present invention. [Figure 3] FIG. 1 is a diagram showing an information processing flow by a prediction device according to an embodiment of the present invention. [Figure 4] FIG. 10 is a diagram showing an example of a procedure for predicting an index value related to the popularity of a second content. [Figure 5] FIG. 10 is a diagram illustrating an example of a procedure for acquiring first information. DETAILED DESCRIPTION OF THE INVENTION

[0018] One embodiment of the present invention (hereinafter referred to as the present embodiment) will be described in detail below with reference to the accompanying drawings. However, the embodiment described below is merely an example given to facilitate understanding of the present invention and is not intended to limit the present invention. In other words, the present invention may be modified or improved from the embodiment described below without departing from the spirit of the present invention. Furthermore, the present invention includes equivalents thereof.

[0019] Furthermore, the basic data processing technologies (communication / transmission technology, data acquisition technology, data recording technology, data processing / analysis technology, machine learning technology, image processing technology, image display technology, visualization technology, etc.) required to realize the present invention are well-known technologies, and therefore, explanations thereof will be omitted. The devices and apparatuses used to apply the above-mentioned known techniques may be appropriately selected from those available at the time of implementing the present invention.

[0020] In addition, in describing the present embodiment, the definitions, meanings, and interpretations of some of the terms used in this specification will be explained below.

[0021] [Device] In this specification, the concept of "device" includes a single device that performs a specific function, as well as a combination of multiple devices that exist independently and in a distributed manner but cooperate (link) to perform a specific function.

[0022] [User] In this specification, a "user" refers to a person who uses the prediction device and prediction method of the present invention for the purpose of enjoying the benefits of the present invention. Specifically, a user of the present invention is a person who uses the prediction device and prediction method of the present invention to predict index values ​​related to the popularity of content (e.g., audience ratings, etc.) and uses the prediction results to operate or support a content-related business.

[0023] [content] In this invention, "content" refers to a creative work (intellectual property, IP) that is provided by a provider through a medium and is viewed, read, listened to, appreciated, or used at a recipient. The form of content provision is not particularly limited, but "provision" in this invention includes, for example, transmission such as broadcasting, distribution, transfer, publication, sale, loan, distribution, performance, screening, musical performance, dictation, live performance, transmission, and reproduction.

[0024] The term "content" in this invention is to be interpreted broadly, and examples include video content, content published in publications such as books, paintings and photographs, playable content such as video games, music and audio content, electronic content such as websites and apps (software), content for sale such as character goods, and content based on performances such as concerts and plays.

[0025] There are no particular limitations on the type of video content, and "video content" includes, for example, television programs (including television commercials), videos distributed over networks such as the Internet, and movies, regardless of whether they contain audio. With regard to videos distributed over networks such as the Internet, there are no particular limitations on the form of distribution, and these include videos that can be viewed by accessing video sites, videos that can be viewed through specific application programs, and videos that can be viewed through VOD (video-on-demand) services, regardless of whether they are paid or free. Furthermore, "publications" include magazines, comics, books, newspapers, and other publications, and also include e-books.

[0026] In this specification, the term "medium" refers to a content providing medium, specifically, a medium that transmits content for the purpose of providing it. Specific examples of "medium" include broadcast media such as television broadcasts and radio broadcasts (including cable broadcasts), publishing media such as newspapers and magazines, and network media such as the Internet. The term "medium" may also include two-way media that allows two-way information transmission between a content provider and a content recipient. Furthermore, the medium itself may constitute the content, such as character goods or products with character illustrations printed on them.

[0027] <<Outline of this embodiment>> This embodiment relates to an information processing technology for predicting an index value related to the popularity of content, and in particular to an information processing technology for quantitatively predicting the popularity of second content created based on first content. Specifically, this embodiment predicts an index value related to the popularity of the second content based on information related to the first content. The second content is content created based on the first content and is secondary content.

[0028] An index value relating to the popularity of content is a value that quantitatively indicates the degree of exposure to the content, the degree of public attention to the content, the degree of hit, sales, evaluation, reputation, etc., and can be used as a KPI (Key Performance Indicator). Specifically, when the content is a television program or a video distributed via VOD, the popularity index value corresponds to the viewer rating (VOD viewer rating), which is an example of the degree of exposure. Here, the viewer rating may be the total viewer rating (Gross Rating Point: GRP). Furthermore, for television or VOD-distributed videos, index values ​​other than the viewer rating, specifically, the number of viewers and the cost required per 1,000 viewers (Cost Per Mill: CPM), etc., may be used as values ​​indicating the degree of exposure. In addition, "viewing" can include not only viewing in real time, but also playing back and viewing recorded or audio recordings within a certain period after broadcast, or viewing web-based distribution (time-shifted viewing).

[0029] For content other than television programs, the following values ​​can be used as examples of popularity index values: However, the values ​​below are merely examples, and values ​​other than the following can also be used as popularity index values. Index values ​​for radio programs (including radio programs distributed online): listener ratings, etc. Index values ​​for publications such as books: subscription rates and sales volumes, etc. Movie index values: box office revenue, audience numbers, etc. -Indicators for online games and apps: number of downloads, etc. Index values ​​of network content such as websites: number of unique users, etc. Index values ​​for merchandise and other sales content: number of sales and sales amount (sales amount), etc.

[0030] To explain the relationship between the first content and the second content once again, for example, the original work corresponds to the first content, and its derivative work corresponds to the second content. In other words, the second content is created by adapting, rewriting, translating, or dramatizing the first content, or is created based on the first content (with content that can evoke the first content).

[0031] The first content and the second content may be of the same or different types (in other words, the types of media on which the content is provided), and for example, the first content may be a publication and the second content may be a video version of the first content. Alternatively, both the first content and the second content may be video content, such as a television anime and its live-action film adaptation. Furthermore, when content derived from a certain content is provided through various media via cross-media, the original content corresponds to the first content, and the derived content provided by each media corresponds to the second content.

[0032] The first content is existing content, i.e., content that has been provided in the past (hereinafter also referred to as provided content). The provided content may include content that is currently being provided and content that has ceased to be provided at present. In this embodiment, it is assumed that the first content has a right (copyright) holder.

[0033] In contrast, the second content is content that has not yet been provided, and specifically may be content that is scheduled to be provided in the future, content for which a provision date has been determined, content that is currently being conceived or planned, or fictitious content that is not in the planning stage but is hypothetical for hypothetical provision. However, the second content is not limited to these, and may also be existing content, i.e., content that has already been provided.

[0034] In this embodiment, before the second content is provided, a popularity index value for the second content is predicted based on information about the first content (hereinafter referred to as first information). The first information includes text information indicating characteristic words, phrases, and sentences related to the content of the first content, so-called tag information. To give a specific example, if the first content is a story such as a manga, the genre, an outline of the story, etc. may be included in the first information.

[0035] The first information may also include information regarding the medium through which the first content is provided and the time of provision. For example, if the first content is published in a publication, the first information may include the name of the publication and the publication date (corresponding to the time of provision), etc.

[0036] Furthermore, the first information may include information about the attributes of those who come into contact with the first content, specifically, the age group, gender, etc. of the target group set as recipients of the first content. Here, "coming into contact with content" may include not only viewing, browsing, listening, purchasing, appreciating, or using the content, but also recognizing the content.

[0037] In this embodiment, the first information is acquired from the name of the first content (content name). Specifically, articles related to the first content (more specifically, the contents of websites) are searched for on the web based on the title of the first content, the contents of the first content are summarized from the searched articles, and characteristic words, phrases, sentences, etc. are extracted from the summaries, thereby acquiring tag information as the first information related to the contents of the first content. The number of pieces of tag information to be extracted (acquired) is not particularly limited, and may be set to a number appropriate for predicting the popularity index value for the second content.

[0038] The method and procedure for acquiring the first information are not particularly limited, but the search for related articles may be performed using the functions of a publicly known cloud service or API (Application Programming Interface). Furthermore, the summary of the first content based on the related articles and the extraction of tag information from the summary may be performed using the functions of a publicly known text generation AI. Furthermore, a generative model (more specifically, a tag generation model described below) that integrates these functions may be constructed, and the name of the first content (specifically, the title) may be input into this generative model to acquire tag information as the first information related to the content of the first content.

[0039] The source of the first information is not particularly limited, and the first information may be obtained from a producer or provider of the first content (for example, a television station or publisher), or from a person who has come into contact with the first content (for example, a viewer or purchaser). The first information may also be obtained using a search engine or a social networking service (SNS). The first information may also be obtained via a research organization such as an audience rating research company.

[0040] In this embodiment, the first information is input into a prediction model (hereinafter, prediction model M) to predict a popularity index value for the second content. The prediction model M is a mathematical model according to the correspondence between information about the provided content and the popularity index value of the provided content, and more specifically, the correspondence is expressed as a mathematical formula or model.

[0041] In this embodiment, a prediction model M is constructed by performing machine learning using information about the provided content and a popularity index value identified for the provided content as training data. The provided content related to the training data (hereinafter referred to as training content) may be the same type of content as the second content, and if the second content is a television program, it is preferable that the training content is also a television program. Furthermore, in this embodiment, the training content, like the second content, is created based on original content (hereinafter referred to as original content). However, the present invention is not limited to this, and the training content may be content for which no original content exists.

[0042] The information about the study content may include information about the content of the study content, specifically, text information indicating characteristic words, phrases, and sentences related to the content of the study content, i.e., tag information. The tag information about the content of the study content can be obtained using the same procedure as the tag information about the content of the first content, for example, by searching the Internet for articles related to the study content, summarizing the content of the study content from the found articles, and extracting characteristic words, phrases, sentences, etc. from the summary.

[0043] The information about the study content may also include information about the media and timing of providing the content. For example, if the study content is a television program, meta information about the program, specifically, the broadcast date and time, broadcast station, broadcast time slot, broadcast day, program title, broadcast start year, number of broadcasts, broadcast area, and broadcast period, may be used as study data. Furthermore, the information about the learning content may include information about the attributes of those who come into contact with the learning content, specifically, the age group and gender of the target audience to whom the learning content is to be provided.

[0044] Furthermore, in this embodiment, as described above, the study content is created based on the original content, and information about the original content may be included in the study data. For example, if the original content is published in a publication, the name of the publication, the publication period, the number of times it was published, the year it was first published, the number of copies of the publication, etc. may be used as study data.

[0045] The method, procedure, and source of information for obtaining information about the learning content are not particularly limited, and information may be obtained from the creator or provider of the learning content (e.g., a television station or publisher), or from a person who comes into contact with the learning content (e.g., a viewer or purchaser). Information about the learning content may also be obtained using a search engine or social networking service (SNS), or through a research organization such as an audience rating research company.

[0046] Regarding the popularity index value of the study content, an index value of a type corresponding to the type of the study content is used as the study data. For example, among the index values ​​described above, a value corresponding to the type of study content is used. Specifically, if the study content is a television program, the program's audience rating is used as the study data. Furthermore, the index value used as the study data is an actual measured value determined by surveys, measurements, etc.

[0047] The popularity index of the learning content may be determined, for example, by a research company or a content provider (hereinafter, "research company"). Data indicating the determined index (actual measurement value) may be available from the research company. For example, if the learning content is a television program, a ratings research company may determine the viewership rating as the popularity index of the television program, and the determined viewership rating data may be available from the ratings research company. Here, the viewership rating may be calculated, for example, by surveying the television viewing habits of a research panel using a known mechanical method and aggregating the survey results. Alternatively, the viewership rating may be calculated by aggregating manufacturer logs / device logs generated by the devices used to watch the television or viewing logs held by the television station. Alternatively, the viewership rating may be calculated using published statistical data. Alternatively, the viewership rating may be calculated by conducting a questionnaire survey of the research panel, aggregating the survey results, and analyzing the aggregated survey results.

[0048] Regarding the machine learning for constructing the predictive model M, the learning method is not particularly limited, and any known learning method can be used, as long as a predictive model is constructed in which information about the content is used as the explanatory variable (input variable) and an index value of the content's popularity is used as the objective variable. Furthermore, the machine learning algorithm is also not particularly limited, and any known learning algorithm can be used, such as a neural network, a convolutional neural network, a recurrent neural network, a Bayesian neural network, attention, a transformer, a variational autoencoder, a generative adversarial network, a deep learning neural network, a Boltzmann machine, a matrix factorization, a factorization machine, an M-way factorization machine, a field-aware factorization machine, a field-aware neural factorization machine, a support vector machine, a Bayesian network, a decision tree, a random forest, and other machine learning algorithms.

[0049] In addition to the first information described above, the input information to the prediction model M can also include information related to the second content (hereinafter, "second information"). The second information can be the time and medium for providing the second content. For example, if the second content is a television program, broadcast slot information (i.e., information about the broadcast date and time and the broadcast station) can be used. Furthermore, as described above, the second content is content that has not yet been provided, and therefore the time and medium for providing the second content may not yet be determined at the stage of predicting the index value. In such a case, the time and medium for providing the second content can be arbitrarily set, and the set content can be used as the second information. Alternatively, if content production is being considered or planned, candidates for the time and medium for providing the second content can be listed, and information about the candidates can be used as the second information.

[0050] <<Configuration example of a prediction device according to this embodiment>> A specific configuration example of the prediction device of this embodiment (hereinafter, the prediction device 10) will be described. The prediction device 10 is made up of a computer, and more specifically, may be made up of a single computer or a plurality of computers distributed in parallel.

[0051] The computer constituting the prediction device 10 may be a PC (Personal Computer), a workstation, or a server. When the prediction device 10 is constituted by a server, the server may be a server for an ASP (Application Service Provider), SaaS (Software as a Service), PaaS (Platform as a Service), or IaaS (Infrastructure as a Service). In this case, when necessary information is input into a client terminal, the server performs various processes and calculations based on the input information, and the calculation results are output on the client terminal side. In other words, the functions of the server that is the prediction device 10 can be used on the client terminal side. The computer constituting the prediction device 10 may also be a quantum computer.

[0052] The prediction device 10 may also be equipped with artificial intelligence (AI). Here, artificial intelligence is a technology that realizes intelligent functions such as inference, prediction, and judgment using hardware and software resources.

[0053] Regarding the hardware configuration of the prediction device 10, as shown in FIG. 2, the prediction device 10 has a processor 11, a memory 12, a communication interface 13, a storage 14, an input device 15, and an output device 16, and these devices are electrically connected via a bus 17.

[0054] The processor 11 may be composed of a CPU (Central Processing Unit), MPU (Micro-Processing Unit), MCU (Micro Controller Unit), GPU (Graphics Processing Unit), DSP (Digital Signal Processor), TPU (Tensor Processing Unit), NPU (Neural network Processing Unit), DRP (Dynamically Reconfigurable Processor), ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Arrays), SoC (System on Chip), etc.

[0055] The memory 12 may be configured by semiconductor memories such as a ROM (Read Only Memory) and a RAM (Random Access Memory).

[0056] The communication interface 13 may be configured, for example, by a network interface card or a communication interface board. The standard of data communication via the communication interface 13 is not particularly limited, and examples include communication via a wireless LAN based on Wi-fi (registered trademark), communication via a 3G, 4G, or 5G mobile communication system, or communication based on LTE (Long Term Evolution). The computer constituting the prediction device 10 can communicate with external communication devices or terminals connected to the Internet or a mobile communication line, etc., via the communication interface 13.

[0057] The storage 14 may be configured by a flash memory, a hard disc drive (HDD), a solid state drive (SSD), a flexible disc (FD), a magneto-optical disc (MO disc), a compact disc (CD), a digital versatile disc (DVD), a secure digital card (SD card), a universal serial bus memory (USB memory), or the like. The storage 14 may be built into the computer main body constituting the prediction device 10, or may be attached in an external form. Furthermore, the storage 14 may be configured by a computer (e.g., a database server) communicably connected to the computer main body constituting the prediction device 10. The input device 15 may be configured by, for example, a keyboard, a mouse, a touch panel, etc. The output device 16 may be configured by, for example, a display, a speaker, a printer, etc.

[0058] Furthermore, a program for an operating system (OS) and a predetermined application program (hereinafter, a prediction program) are installed as software in the computer constituting the prediction device 10. When the processor 11 operates in accordance with these programs, the computer functions as the prediction device of the present invention, and specifically, executes a series of processes for predicting an index value related to the popularity of the second content. The above-mentioned prediction program includes a module for constructing the aforementioned prediction model M (hereinafter referred to as the model construction module) and a module for predicting an index value related to the popularity of the second content using the prediction model M (hereinafter referred to as the index value prediction module).

[0059] 2, a database 18 of information related to provided content is constructed, and the prediction device 10 can read information necessary for predicting an index value related to the popularity of the second content from the database 18. Specifically, the database 18 stores, for each piece of provided content, information related to the name (title), content, medium of provision, time of provision, attributes of the target demographic, etc., of each piece of provided content, whether or not there is original content and information related to the original content, and an index value of popularity actually identified for each piece of provided content (specifically, actual values ​​such as viewership rating, number of unique users, and box office revenue). The prediction device 10 then performs machine learning to construct a prediction model M using the information stored in the database 18 as learning data. Note that the database 18 may store the prediction results by the prediction device 10, i.e., the predicted values ​​of the index values ​​related to the popularity of the second content. Also, the database 18 does not necessarily have to be provided, and for example, the database 18 may be unnecessary if data is input every time the prediction device 10 predicts an index value.

[0060] <<Prediction flow according to this embodiment>> Next, we will explain a prediction flow, which is a data processing flow using the above-mentioned prediction device 10. In the prediction flow, the processor 11 of the computer constituting the prediction device 10 executes a prediction program and performs a series of data processing to predict the popularity index value of the second content.

[0061] In the following, we will assume that the first content is manga A, the second content is television program B which is an animated version of manga A, manga A is content that has already been provided, and television program B is content that has not yet been provided, and for ease of understanding, we will explain the case where the first content is manga A, the second content is television program B which is content that has not yet been broadcast.

[0062] In the prediction flow, a prediction model construction phase S001 and an index value prediction phase S002 are carried out in this order, as shown in Fig. 3. Each phase will be described below.

[0063] (Predictive model building phase) The prediction model construction phase S001 is a process of constructing a prediction model M by performing machine learning, and is performed by the processor 11 executing a model construction module of the prediction program. Specifically, in the prediction model construction phase S001, the processor 11 reads information about the learning content and the popularity index value actually identified for the learning content from the information stored in the database 18, and performs machine learning using this as learning data. In this embodiment, secondary content created based on the original content among the provided content is considered to be the learning content. However, this is not limited to this, and content that does not rely on the original content may also be included in the learning content.

[0064] It is preferable that the learning content be the same type of content as the second content, television program B, i.e., a television program, and even more preferable that, like television program B, it be content that is an animated or live-action version of an original work.

[0065] Furthermore, which items of information regarding the learning content are to be used as learning data, i.e., the granularity of the learning data, can be set appropriately depending on the target predicted by the prediction model M obtained as a result of machine learning (more specifically, the expected audience rating of TV program B). After machine learning, it is advisable to verify the prediction accuracy of the prediction model M using a known verification method such as the holdout method or cross-validation method.

[0066] In the present embodiment, the prediction device 10 performs the above-described machine learning to construct the prediction model M, but the present invention is not limited to this, and a device different from the prediction device 10 (hereinafter, a learning device) may perform machine learning to construct the prediction model M. In this case, the prediction device 10 may use the prediction model M by downloading it from the learning device, or may make the prediction model M stored in the learning device available via a cloud service or API.

[0067] (Index value prediction phase) The index value prediction phase S002 is a process of predicting the index value of the popularity of TV program B, specifically, the viewership rating (degree of contact) of TV program B, and is implemented by processor 11 executing the index value prediction module of the prediction program.

[0068] Furthermore, the index value prediction phase S002 is performed before the broadcast (before provision) of television program B, in other words, in the index value prediction phase S002, a predicted audience rating for television program B is calculated before the broadcast of television program B. Here, "before the broadcast of television program B" refers to a stage where the broadcast schedule of television program B (i.e., the broadcast station and broadcast date and time) has been decided, but is not limited to this, and may also be a stage where the broadcast schedule of television program B has not been decided, for example, before television program B is produced or in the planning stage of production.

[0069] The index value prediction phase S002 employs the prediction method of the present invention and proceeds according to the flow shown in Fig. 4. That is, each step shown in Fig. 4 corresponds to each element constituting the prediction method of the present invention. The flow shown in FIG. 4 is merely an example, and unnecessary steps may be deleted, new steps may be added, or the order in which steps are performed may be changed, without departing from the spirit of the present invention.

[0070] In the index value prediction phase S002, first, the user identifies manga A, which is the original work (first content) of television program B, and inputs its name (specifically, title) through the input device 15 of the prediction device 10. The processor 11 accepts the title of manga A input by the user (S011). The user here is, for example, an operator of a computer that constitutes the prediction device 10.

[0071] Next, processor 11 acquires first information about manga A from the input title of manga A (S012). In this step S012, processor 11 acquires tag information about the content of manga A as the first information. Specifically, processor 11 acquires tag information as the first information by inputting the title of manga A into a tag generation model.

[0072] The tag generation model is a generation model that generates information about the content from the name (title) of the content, and is composed of, for example, a Web API and generation AI. Explaining in more detail, as shown in Figure 5, by inputting the title of Manga A into the tag generation model, related articles about Manga A are searched for on the Web, a summary of the content of Manga A is created based on the related articles found, and characteristic words, phrases, and sentences are extracted from the summary as tag information.

[0073] The tag information items extracted by the tag generation model are not particularly limited, but tag information related to predetermined items (for example, genre, publication period, etc.) may always be extracted. For predetermined items, multiple answer contents may be prepared as options, and an answer selected from these may be acquired as tag information. For items other than the predetermined items, tag information indicating the content of a free reply may be used.

[0074] Furthermore, in this embodiment, the first information acquired by the tag generation model includes tag information relating to the time and medium of distribution of manga A, as well as tag information relating to the attributes of those who have come into contact with manga A (more specifically, the target demographic). Specifically, tag information indicating the publication period of manga A, the name and serialization period of the magazine in which manga A is serialized, the publication date and number of copies of the comic book version of manga A, and the gender and age of the target demographic of manga A is acquired as the first information. It is not limited to cases where all information regarding the time of release of Manga A, the media through which it is released, and the attributes of the target demographic is obtained; it is preferable to obtain at least one of the above three pieces of information as the first information.

[0075] Next, processor 11 acquires second information about television program B (S013). In step S013, processor 11 acquires, as the second information, information about the broadcast time (provision time) and broadcast station (provision medium) of television program B. Specifically, processor 11 may acquire the second information by communicating with a server of each television broadcast station or an audience rating research company, for example, or may acquire the second information by a user inputting the information through input device 15 of prediction device 10. It should be noted that the present invention is not limited to the case where information about both the broadcast time and the broadcast station of TV program B is acquired, and it is preferable to acquire information about at least one of the above two pieces of information as the second information.

[0076] Furthermore, in the index value prediction phase S002, step S013 is not a required step, and step S013 may be omitted if, for example, television program B is in the planning stage or an earlier stage and the second information cannot be obtained because the broadcast date and broadcast station have not yet been decided.

[0077] Next, processor 11 inputs the first information about manga A acquired in step S012 into the prediction model M constructed in prediction model construction phase S001 to predict an index value of popularity for television program B, specifically, calculates a predicted value of the audience rating for television program B (S014). Also, if step S013 is performed to acquire second information about television program B, in step S014 processor 11 inputs the first information about manga A and the second information about television program B into the prediction model M to predict the audience rating for television program B.

[0078] Thereafter, processor 11 outputs the audience rating (predicted audience rating value) of television program B predicted in step S014 from output device 16 of prediction device 10 (S015). Then, when the above series of steps are completed, index value prediction phase S002 is completed. The predicted viewership rating for TV program B output in step S015 is reported to, for example, the TV station that provides TV program B, a company that advertises or promotes within the broadcast time slot of TV program B, or a person who requested a survey of the predicted viewership rating.

[0079] <<Effectiveness of this embodiment>> According to this embodiment, the popularity index value of content (second content) that has not yet been provided can be predicted based on information (first information) obtained from the name of the content (first content) that is the source of that content. In this way, information about the content can be obtained from limited information, such as the name of existing content, and the popularity of the second content can be quantitatively predicted based on the obtained information. Furthermore, the popularity prediction result can be used as useful information for investment in, planning for, etc. the second content. Specifically, it can be used as a basis for determining whether to provide the second content, and if so, the specific means and timing of providing it. It can also be used as information for optimizing resources in the content business.

[0080] Furthermore, in this embodiment, the popularity index value of the second content is predicted by inputting the first information into a prediction model M constructed by machine learning. In other words, by modeling the correspondence between information about the content and the popularity index value of the content by machine learning and using a prediction model M according to that correspondence, the popularity index value of the content can be predicted with high reliability.

[0081] Furthermore, the above machine learning is performed using information about the provided content and index values ​​(actual measured values) such as viewership ratings actually determined for the provided content as learning data. As a result, higher reliability (credibility) can be obtained for the prediction of index values ​​using the prediction model M obtained as a result of machine learning.

[0082] In this embodiment, the second content, i.e., the content whose popularity index value is predicted as the prediction target, is not particularly limited as long as it is content created based on the first content, and the popularity of a variety of content can be predicted. In other words, according to this embodiment, the popularity of content can be predicted using a prediction model M that is highly versatile and practical. [Explanation of symbols]

[0083] 10 Prediction Device 11 processors 12 Memory 13 Communication Interface 14. Storage 15 Input Devices 16 Output Devices 17 Bus M Prediction Model

Claims

1. a processor, the processor comprising: acquiring first information about the content of a first content by inputting the name of the first content into a generative model that generates information about the content from the name of the content; A prediction device that obtains, when first information is input into a prediction model that corresponds to the correspondence between information about the content of the content and an index value related to the popularity of the content, the value output by the prediction model as a predicted value of the index value for a second content created based on the first content.

2. The prediction device according to claim 1 , wherein the processor calculates a predicted value of a degree of contact with the second content as the index value.

3. the first content is content that has been provided in the past, The prediction device according to claim 1 , wherein the first information includes at least one of information regarding a time when the first content is provided, a medium for providing the first content, and information regarding attributes of a person who comes into contact with the first content.

4. The predictive model is constructed by performing machine learning using, as training data, information regarding the content, information regarding at least one of the time and medium for providing the content, and an index value regarding the popularity of the content; The processor: acquiring second information relating to at least one of a time when the second content is provided and a medium through which the second content is provided; The prediction device according to claim 1 , wherein a value output by the prediction model when the acquired first information and second information are input to the prediction model is acquired as a predicted value of the index value for the second content.

5. The prediction device according to claim 1 , wherein the processor predicts the index value for the second content before the second content is provided.

6. 2. The prediction device according to claim 1, wherein the prediction model is constructed by machine learning using information about the content of the provided content and the index value identified for the provided content as learning data.

7. a processor inputs a name of a first content into a generative model that generates information about the content from the name of the content, thereby acquiring first information about the content of the first content; A prediction method in which a processor inputs first information into a prediction model corresponding to the correspondence between information about the content of the content and an index value related to the popularity of the content, and obtains the value output by the prediction model as a predicted value of the index value for a second content created based on the first content.

Images