JP7522512B1 - Intent sales support device - Google Patents

Abstract

[Problem] To support intent sales to journalists by automating the process of sending documents such as article proposals to journalists.
[Solution] The intent sales support device 1 of the present invention comprises an article collection unit 111 that collects articles attached with information identifying reporters from the Internet, an intent analysis unit 112 that analyzes the intent, which is a tendency of articles posted by each reporter, by processing using a large-scale language model that includes the collected articles as input, and a document generation unit 117 that generates a document including information that contributes to the creation of articles by reporters by processing using a large-scale language model that includes the results of the intent analysis by the intent analysis unit 112 as input.
[Selected Figure] Figure 1

Description

The present invention relates to an intent sales support device.

Using external newspapers, magazines, etc. for public relations can provide a reliable and objective source of information, which strengthens the company's credibility and expertise. In addition, incorporating a third-party perspective can increase the diversity and persuasiveness of the public relations content. However, it is not easy to get public relations topics featured in articles in external newspapers, magazines, etc.

Due to the above circumstances, there is a demand for a method to support intent sales to journalists, in other words, sales that analyze the interests and concerns of journalists and sell them article topics based on the results of that analysis. The "intent" of journalists referred to here includes the fields and topics that the journalist tends to report on, the reporting style, the target readership, the content of past articles, the reactions and engagement to articles written by the journalist, the frequency of article posting, etc.

Regarding technology for providing information according to the interests of a target, Patent Literature 1 discloses a technology for personalized delivery of content items to users of a social networking system. The technology uses attributes of content posted on a brand page in conjunction with characteristics of the user profile of the social networking system user to select content to be presented to the user when the brand page is accessed.

The technology described in Patent Document 1 can increase the likelihood that various content posted on a brand page will interest users.

JP 2018-129052 A

Incidentally, in intent sales to journalists, it is important to provide the journalist with a proposal. If the information necessary for the journalist to write an article is compiled in the proposal, the likelihood that the journalist will write the desired article can be increased. The technology in Patent Document 1 is limited to being able to select content based on a user profile, and does not show how a journalist who has not necessarily made public a user profile indicating intent can compile in a proposal the information necessary to write an article. Therefore, the technology described in Patent Document 1 has room for further improvement in terms of supporting intent sales to journalists.

The present invention has been made in consideration of the above circumstances. The object of the present invention is to support intent sales to journalists by automating the process of sending documents such as article proposals to journalists.

As a result of intensive research into solving the above problem, the inventors have found that the above object can be achieved by collecting so-called bylined articles from the Internet, analyzing the intent of journalists through processing using a large-scale language model based on the bylined articles, and generating article proposals through processing using a large-scale language model based on the results of the analysis. The inventors have then completed the present invention. Specifically, the present invention provides the following.

The present invention provides an intent sales support device that includes an article collection unit that collects articles attached with information identifying a reporter from the Internet, an intent analysis unit that analyzes intent, which is a tendency of articles posted by the reporter, through processing using a large-scale language model that includes the articles as input, and a document generation unit that generates a document including information that contributes to the creation of an article by the reporter through processing using a large-scale language model that includes as input the results of the intent analysis by the intent analysis unit.

The present invention analyzes intent, which is a tendency of articles posted by journalists, by processing signed articles and the like collected from the Internet by an article collection unit using a large-scale language model. Then, a document generation unit of the present invention generates a document containing information that contributes to the creation of articles by the journalists, by processing using a large-scale language model that includes the results of this analysis as input. This automates the intent analysis, which is a tendency of articles posted by journalists, and the generation of proposals and the like based on this analysis, in the process of sending article proposals to journalists. Thus, the present invention can support intent sales to journalists by automating the process of sending documents such as article proposals to journalists.

The present invention can support intent sales to journalists by automating the process of sending documents such as article proposals to journalists.

FIG. 1 is a block diagram showing an example of the hardware configuration and software configuration of a system S according to the present embodiment. FIG. 2 is an example of the intent analysis database 121 . FIG. 3 is an example of the company database 122 . FIG. 4 is a main flow chart showing an example of a preferable flow of the assistance process executed by the assistance device 1 of this embodiment. FIG. 5 is a continuation of FIG. FIG. 6 is a continuation of FIG.

The following describes in detail an example of an embodiment of the present invention with reference to the drawings.

<System S>
Fig. 1 is a block diagram showing an example of a hardware configuration and a software configuration of a system S of this embodiment. An example of a preferred embodiment of the system S of this embodiment will be described below with reference to Fig. 1. The system S of this embodiment includes a support device 1. The support device 1 is capable of communicating with a terminal T via a network N.

[Support device 1]
The support device 1 realizes each software component shown in Fig. 1 by using hardware components such as a control unit 11, a storage unit 12, and a communication unit 13. The support device 1 executes a support process that supports intent sales to journalists by automating the process of sending article proposals to journalists using each software component. The type of the support device 1 is not particularly limited. The type of the support device 1 may be, for example, various server devices, cloud servers, etc.

[Control unit 11]
The control unit 11 includes a central processing unit (CPU), a random access memory (RAM), a read only memory (ROM), and the like.

The control unit 11 cooperates with the memory unit 12 and/or the communication unit 13 as necessary. The control unit 11 then realizes the software components of the program of this embodiment executed by the support device 1, such as an article collection unit 111, an intent analysis unit 112, an advertising effectiveness measurement unit 113, an analysis result management unit 114, a reporter extraction unit 115, a data acquisition unit 116, a document generation unit 117, a document transmission unit 118, and a machine learning unit 119.

[Memory unit 12]
The memory unit 12 is a device in which data and/or files are stored, and includes a storage unit that stores data non-temporarily using a hard disk, a semiconductor memory, a recording medium, a memory card, and the like.

The storage unit 12 stores programs executed by the microcomputer, articles collected by the article collection unit 111, etc. It is also preferable that the storage unit 12 further stores an intent analysis database 121 to manage the intent analysis results. In addition, it is also preferable that the storage unit 12 further stores a company database 122 to manage data related to topics desired to be covered in articles exemplified by data of companies, etc.

It is preferable that the memory unit 12 further stores data associating the URL of an article with the name or pen name of a reporter, so as to enable identification of the reporter from the URL of the article. It is preferable that the memory unit 12 further stores feedback from recipients of a document generated by the support device 1 of this embodiment, so as to enable collection and utilization of feedback on the generated document. It is preferable that the memory unit 12 further stores associations between the reporter and/or the media in which the reporter publishes an article and a format, so as to enable document generation in a format corresponding to the reporter and/or the media in which the reporter publishes an article. It is preferable that the memory unit 12 stores, for each reporter, information on a channel, which is a means of contacting the reporter.

(Intent Analysis Database 121)
The storage unit 12 preferably stores an intent analysis database 121, which is a database for storing intent analysis results. The intent analysis database 121 stores data associating reporters with analysis results regarding intents, which are tendencies of articles posted by the reporters. The intent analysis database 121 is managed by the analysis result management unit 114.

Intents include topics, keywords, sentiment, subtopics, time series of interests, reporting style, etc. Topics are the main themes and topics of an article. Keywords are important words and phrases in an article. Sentiment is the tone and emotional tendency of an article. Subtopics indicate finer categories within a main topic. Time series of interests is data showing changes in topics, interests, etc. over time. Reporting style is a style related to reporting, such as objective reporting, opinion pieces, investigative reporting, etc.

To facilitate data management, the above-mentioned data is preferably stored in association with an analysis result ID that identifies the analysis result. The analysis result preferably includes one or more of various information related to the article, exemplified by the analysis result of the article's influence, the analysis result of reader engagement with the article, the identification result of the article's trend pattern, the comparison result with related articles, etc. The influence of the article includes the influence of the article, and a so-called reach-related evaluation indicating the number of views of the article. The analysis result of engagement includes the measurement result of engagement, the analysis result of reader responses to the article, etc. The identification result of the article's trend pattern includes the analysis result of the current and past trend patterns, etc. related to the article. The comparison result with related articles includes the comparison analysis result with other articles dealing with similar topics, etc.

Figure 2 is an example of the intent analysis database 121. This example includes data related to the intent analysis results associated with the analysis result IDs "A0001", "A0002", and "A0003".

The data related to analysis result ID "A0001" includes, for reporter "Kagawa reporter," the analysis results of the reporter's intent, such as the topic "lifestyle," keyword "lifestyle, events, cooking," sentiment "calm," subtopic "lifestyle, cooking," time series of interests "...," and reporting style "investigative reporting." In addition, the data related to analysis result ID "A0001" includes, as various information related to the article, the analysis results of the article's influence "Editor-in-Chief of △△△△△ Magazine," the analysis results of reader engagement with the article "△△ million engagements," the identification results of the article's trend patterns "...," and the results of comparison with related articles "...," etc.

The data related to analysis result ID "A0002" includes, for reporter "S Mizu reporter," the results of an analysis of the reporter's intent, such as the topic "crime," keywords "criminal case, real culprit, scandal, false accusation," sentiment "objective, righteous indignation," subtopic "...," time series of interest "...," and reporting style "investigative reporting." In addition, the data related to analysis result ID "A0002" includes, as various information related to the article, the results of an analysis of the article's influence "Awarded the Galaxy Award, etc.", the results of an analysis of reader engagement with the article "△△△ million engagements," the results of identifying the article's trend patterns "...," and the results of comparison with related articles "...," etc.

The data related to analysis result ID "A0003" includes, for reporter "Tazaki reporter," the analysis results of the reporter's intent, such as the topic "society," the keyword "new coronavirus, PCR testing," the sentiment "subjective, inflammatory," the subtopic "new coronavirus," the timeline of interest "...," and the reporting style "announcement reporting." In addition, the data related to analysis result ID "A0003" includes, as various information related to the article, the analysis results of the article's influence "...," the analysis results of reader engagement with the article "△△ million engagements," the identification results of the article's trend patterns "...," and the results of comparison with related articles "...," etc.

By storing the above data in the intent analysis database 121, the support device 1 can extract, for example, a reporter "Kagawa reporter" as an information source for an article related to the topic "Japanese cuisine." The support device 1 can then generate a document for the reporter who is conducting investigative reporting, containing various pieces of information that will be useful when conducting research related to the topic.

(Company database 122)
The storage unit 12 preferably stores a company database 122, which is a database that stores data related to topics desired to be covered in articles. The company database 122 stores topics desired to be covered in articles in association with topic data, which is data related to the topics. The data stored in the company database 122 is acquired by the data acquisition unit 116. The topics are, for example, specific companies, specific products, specific services, etc.

Topic data includes, for example, company information, etc. Company information includes, for example, characteristics of the company, etc. (corporate characteristics), the company, etc.'s response to industry trends, stories related to the company, etc., campaigns and events related to the company, etc., features of products and services provided by the company, etc., insights from the management of the company, etc., social responsibility efforts by the company, technological innovations by the company, etc., success stories by the company, etc., outlook for future plans for the company, etc. To facilitate data management, the above data is preferably stored in association with a topic ID that identifies the topic.

Figure 3 is an example of the company database 122. This example includes data on topics associated with topic IDs such as "T0001".

Data related to topic ID "T0001" includes topic data related to the topic "Japanese restaurant chain △△". The topic data includes company characteristics "mainly operating within Tokyo", response to industry trends "sustainability and local production for local consumption", story "since its founding, the company has maintained the traditions of Japanese cuisine while also challenging itself to create new dishes...", campaigns and events "seasonal special menus...", product and service characteristics "diverse menus and high-quality service", management insights "pursuit of a sustainable business model", social responsibility initiatives "contribution to the local community and support for education", technological innovations "digitalization and efficiency", success stories "successfully expanding store openings not only to urban areas but also to suburban areas...", and future plans outlook "expansion both domestically and overseas...".

By storing the above-mentioned data in the company database 122, the support device 1 can acquire target data, which is data that a reporter can use to write an article about, for example, the topic "Japanese restaurant chain △△." The support device 1 can then provide the reporter with a document generated based on the target data. In this way, the support device 1 can increase the likelihood that the reporter will write an article based on the target data.

[Communication unit 13]
The communication unit 13 is not particularly limited as long as it can connect the support device 1 to the network N and communicate with the terminal T, etc. An example of the communication unit 13 is a network card compatible with the Ethernet standard.

[Network N]
The type of the network N is not particularly limited as long as it enables the support device 1 and the terminal T to communicate with each other. The type of the network N is, for example, the Internet, a mobile phone network, a wireless LAN, and the like.

[Terminal T]
There is no particular limitation on the type of the terminal T. The terminal T may be, for example, a personal computer, a laptop computer, a smartphone, a tablet terminal, or the like.

[Main Flowchart of Support Processing]
Fig. 4 is a main flow chart showing an example of a preferred flow of the support process executed by the support device 1 of this embodiment. Fig. 5 is a diagram continued from Fig. 4. Fig. 6 is a diagram continued from Fig. 5. Below, an example of a preferred flow of the support process executed by the support device 1 of this embodiment will be described with reference to Figs. 4 to 6.

First, as part of the support process, the support device 1 executes a series of processes to collect articles attached with information identifying the reporter from the Internet. Steps S1 to S2 are an example of this process. Because the target of collection in this process is "articles attached with information identifying the reporter," the support device 1 can analyze the interests and concerns of the reporter, i.e., the intent of the reporter, based on the collected articles.

[Step S1: Determine whether to collect articles]
The control unit 11 cooperates with the storage unit 12 and the communication unit 13 to execute the article collection unit 111. The control unit 11 executes a process to determine whether to collect articles with information identifying a reporter from the Internet using the article collection unit 111 (step S1, article collection determination step). If it is determined that the articles should be collected, the control unit 11 moves the process to step S2. If it is determined that the articles should not be collected, the control unit 11 moves the process to step S7.

[Step S2: Collect articles]
The control unit 11 executes a process of collecting articles with information identifying the reporter from the Internet using the article collecting unit 111 (step S2, article collecting step). The control unit 11 moves the process to step S3.

The "information identifying the reporter" in this step is various types of information attached to the article that can be used to identify the reporter, such as the reporter's signature included in a bylined article, the reporter's name or pen name determined based on the article's URL, or the reporter's name or pen name stored in the Author field included in the article's metadata.

[Step S3: Analyze Intent]
The control unit 11 executes a process of analyzing intent, which is a tendency of articles posted by the reporters related to the articles, by processing with the articles collected in the article collection step as input by the article collection unit 111 (step S3, intent analysis step). The control unit 11 shifts the process to step S4.

The process of analyzing the intent in the intent analysis step is not particularly limited. The process may include a procedure of referring to analysis data received from an external source. The process preferably includes a process of analyzing the intent, which is a tendency of articles posted by reporters related to the articles, by a process using a large-scale language model including the articles collected in the article collection step as input. The process is realized by a procedure of inputting, for example, a prompt asking about each element constituting the intent, which is exemplified by topics, keywords, sentiment, subtopics, time series of interests, reporting style, etc., and the article to be analyzed for the intent, into the large-scale language model. At this time, it is preferable that the large-scale language model has been subjected to pre-processing such as fine tuning and advance learning related to the intent analysis. The pre-processing is performed, for example, using learning data that associates articles, elements constituting the intent, and examples of analysis results related to the elements.

This enables the support device 1 to realize topic identification to identify the main theme/topic of an article, keyword identification to identify important words/phrases, sentiment analysis to analyze the tone/emotional tendency of an article, subtopic analysis to identify finer categories within the main topic, time series analysis of interests to track changes in topics/interests over time, and reporting style identification to identify styles such as objective reporting, opinion pieces, investigative reporting, etc.

In addition, the intent analysis step preferably includes a process for analyzing current and past trends and patterns related to the reporter's intent, and a process for performing a comparative analysis with other articles dealing with similar topics. This process is realized using time-series data of intent stored in the intent analysis database 121, data of other articles stored in the memory unit 12, etc. In addition, the intent analysis step preferably includes a process for analyzing the language used by the reporter, and the vocabulary preferred by the reporter.

As part of the support process, the support device 1 preferably executes a series of processes for measuring advertising effectiveness. Steps S4 to S5 are an example of such processes. Through such processes, the support device 1 can measure advertising effectiveness, exemplified by an article's influence, reach, engagement evaluation, etc., and extract reporters based on advertising effectiveness such as influence.

[Step S4: Determine whether to measure advertising effectiveness]
The control unit 11 cooperates with the storage unit 12 to execute the advertising effectiveness measurement unit 113. The control unit 11 then executes a process of determining whether to measure the advertising effectiveness related to the article written by the reporter related to the intent analysis step, using the advertising effectiveness measurement unit 113 (step S4, advertising effectiveness measurement determination step). If it is determined that the advertising effectiveness is to be measured, the control unit 11 moves the process to step S5. If it is determined that the advertising effectiveness is not to be measured, the control unit 11 moves the process to step S6.

[Step S5: Measuring advertising effectiveness]
The control unit 11 executes a process of measuring the advertising effectiveness related to the article written by the reporter related to the intent analysis step (step S5, advertising effectiveness measurement step) by the advertising effectiveness measurement unit 113. The control unit 11 shifts the process to step S6.

The measurement related to the advertising effectiveness measurement step is not particularly limited. The measurement includes, for example, a procedure of collecting various information related to the influence of an article from the Internet and measuring the influence based on the collected information. The various information related to the influence of an article includes, for example, the number of views of the article, the number of comments on the article, the awards the reporter has received, the reporter's popularity, the engagement with the article, etc.

As part of the assistance process, the assistance device 1 preferably executes a process for managing intent analysis results, etc. Step S6 is an example of such a process. This allows the assistance device 1 to assist with intent marketing using intent analysis results, etc.

[Step S6: Storing the analysis results]
The control unit 11 cooperates with the storage unit 12 to execute the analysis result management unit 114. Then, the control unit 11 executes a process of storing the analysis result related to the intent analysis step in the intent analysis database 121 by the analysis result management unit 114 (step S6, analysis result management step). The control unit 11 shifts the process to step S7.

In addition, when the advertising effectiveness measurement step is performed, the analysis result management step executes a process of storing the measurement results related to the advertising effectiveness measurement step in the intent analysis database 121.

As part of the support process, the support device 1 executes a process of generating a document including information that contributes to the creation of an article by a reporter based on the intent analysis results, etc. Steps S7 to S11 are an example of this process.

[Step S7: Determine whether a document generation instruction has been issued]
The control unit 11, in cooperation with the storage unit 12 and the communication unit 13, executes a process of determining whether an instruction to generate a document has been received (step S7, document generation determination step). If it is determined that an instruction has been received, the control unit 11 moves the process to step S8. If it is not determined that an instruction has been received, the control unit 11 moves the process to step S14. This determination includes, for example, a procedure of determining that an instruction to generate a document has been received when an instruction to generate a document is received from the terminal T.

[Step S8: Extract reporters]
The control unit 11 executes the reporter extraction unit 115 in cooperation with the storage unit 12 and the communication unit 13. The control unit 11 then executes a process of extracting a reporter who is to be provided with the document generated in response to the instruction related to the document generation determination step, using the reporter extraction unit 115 (step S8, reporter extraction step). The control unit 11 then moves the process to step S9.

The reporter extraction step includes a procedure for extracting a reporter specified from the instructions related to the document generation determination step, a procedure for extracting a reporter using the intent analysis results, and the like. The procedure for extracting a reporter using the intent analysis results includes, for example, a procedure for extracting a reporter associated with a topic by comparing a topic desired to be covered in an article with the analysis results related to the intent analysis step. In this way, the support device 1 can extract a reporter whose intent is associated with the topic, and provide the reporter with a document including information that will contribute to the creation of the article.

The procedure for extracting reporters based on advertising effectiveness preferably further includes a procedure for extracting reporters based on advertising effectiveness. The procedure for extracting reporters based on advertising effectiveness includes, for example, a procedure for extracting, from a plurality of reporters, reporters related to a topic desired to be covered in an article, reporters whose advertising effectiveness satisfies given conditions. In this way, the support device 1 can extract reporters who satisfy desired conditions, such as reporters with high advertising effectiveness, such as influence and engagement, from reporters whose intent is related to the topic, and provide the reporters with documents including information that will contribute to the creation of the article.

As part of the support process, the support device 1 preferably executes a series of processes to acquire data related to the topic. Steps S9 to S10 are an example of such processes. Through such processes, the support device 1 can acquire target data, which is data related to the topic that the reporter will use to create an article, from a database such as the corporate database 122, and generate a document to be provided to the reporter through processing using a large-scale language model that includes the acquired target data as input. Thus, the support device 1 can provide the reporter with a document that is rich in information that will contribute to the creation of the article.

[Data preparation steps]
When the above-described series of processes are executed, it is preferable that the support process includes a data preparation step of collecting data relating to the topic desired to be covered in the article and storing it in the company database 122 so that data relating to the topic is pre-stored in the company database 122.

[Step S9: Determine whether to acquire data related to the topic]
The control unit 11 cooperates with the storage unit 12 to execute the data acquisition unit 116. The control unit 11 then executes a process of determining whether or not to acquire data related to a topic desired to be covered in an article by the data acquisition unit 116 (step S9, data acquisition determination step). If it is determined that data should be acquired, the control unit 11 shifts the process to step S10. If it is determined that data should not be acquired, the control unit 11 shifts the process to step S11.

The procedure for determining in the data acquisition determination step is not particularly limited. For example, this procedure includes a procedure for determining that data related to a topic desired to be covered in an article is to be acquired when the data is pre-stored in a database such as the company database 122.

[Step S10: Obtain data related to the topic]
The control unit 11 executes a process of acquiring, by the data acquisition unit 116, target data, which is data that the reporters extracted in the reporter extraction step will use to write articles, from a database such as the company database 122 (step S10, data acquisition step), among data related to topics desired to be covered in articles. The control unit 11 then shifts the process to step S11.

The procedure for acquisition in the data acquisition step is not particularly limited. The procedure includes, for example, a procedure for generating a query for acquiring target data based on the intent analysis result, and acquiring the target data using the query. If a large-scale language model has been subjected to pre-learning or the like related to the generation, the query may be generated using such a large-scale language model.

[Step S11: Generate document]
The control unit 11 cooperates with the storage unit 12 to execute the document generation unit 117. The control unit 11 then executes a process of generating a document including information that contributes to the creation of an article by the reporter extracted in the reporter extraction step, by processing using a large-scale language model that includes as input at least the intent analysis result by the intent analysis unit 112, through the document generation unit 117 (step S11, document generation step). The control unit 11 then shifts the process to step S12.

By generating documents through processing using a large-scale language model that includes the results of intent analysis as input, the support device 1 can create documents as proposals based on the reporter's intent, i.e., create target-oriented content. This also enables the support device 1 to generate documents that propose innovative and attractive proposals. In other words, the support device 1 can generate documents that include creative proposals.

The procedure for generating documents in the document generation step is not particularly limited. In order to generate documents that are rich in information that contributes to the creation of articles, it is preferable that the procedure includes a step of generating documents by processing using a large-scale language model that includes as input the target data acquired in the data acquisition step.

The target data includes, for example, information and analysis data such as basic information, history, major products, services, the current situation and trends of the industry to which the company belongs, the competitive situation in the industry related to the company, the position and influence of the company in the market, the latest news and events related to the company, profiles of the management team of the company, i.e., key executives and leaders, financial situation, profitability, and growth analysis results, the social responsibility and sustainability of the company, technological innovation and research and development efforts of the company, and the reputation and brand image of the company in public places. This allows the support device 1 to generate documents that make persuasive proposals using these various types of information and analysis data, that is, data-driven proposals. In addition, since various information related to a topic is included in the target data, the support device 1 can generate documents that explore various themes including proposals from different angles and perspectives.

When the document generation step includes a procedure for generating a document by processing using a large-scale language model that includes target data as input, the procedure preferably includes a procedure for customizing information about a company, etc. Targets of customization include, for example, items such as highlighting company characteristics, responding to industry trends, utilizing storytelling, introducing campaigns and events, highlighting product and service features, insights from management, social responsibility initiatives, highlighting technological innovations, presenting success stories, and outlook for future plans. Examples of customization include customization that highlights items specified by the user for emphasis, and customization that removes items specified by the user not to be featured. This allows the support device 1 to generate a document customized in accordance with the user's wishes.

To generate a document in a language tailored to the reporter, if the intent analysis result includes a language used by the reporter, the procedure preferably includes a step of generating the document by processing with a large-scale language model whose input includes instructions to generate the document in that language.To generate a document in a language tailored to the reporter, if the intent analysis result includes language preferred by the reporter, the procedure preferably includes a step of generating the document by processing with a large-scale language model whose input includes instructions to generate the document in that language.

When the large-scale language model is associated with an image database and/or an image generating neural network to generate a more attractive proposal document with integrated visual elements, the document generating step preferably includes a procedure for generating a document including visual content. This procedure is realized by referencing an image database and/or generating images using an image generating neural network.

In order to generate a document as a proposal in a format customized to the journalist's preferences and the media format, the document generation step preferably includes a procedure for generating a document in a format associated with the journalist and/or the media in which the journalist will publish an article. This procedure is realized, for example, by referring to the association between the journalist and/or the media in which the journalist will publish an article and the format stored in the storage unit 12.

As part of the support process, the support device 1 preferably executes a series of processes to send the document generated in the document generation step to a reporter. Steps S12 to S13 are an example of such processes. Through such processes, the support device 1 can consistently support the process from intent analysis to the generation and transmission of documents that contribute to intent sales, thereby reducing the user's efforts related to the intent sales.

[Step S12: Determine whether to transmit the document]
The control unit 11 cooperates with the storage unit 12 and the communication unit 13 to execute the document transmission unit 118. The control unit 11 then executes a process of determining whether or not to transmit the document generated in the document generation step by the document transmission unit 118 (step S12, document transmission determination step). If it is determined that the document should be transmitted, the control unit 11 shifts the process to step S13. If it is determined that the document should not be transmitted, the control unit 11 shifts the process to step S14.

The procedure for determining whether to send a document in the document sending determination step is not particularly limited. In order to reduce the user's effort in sending, the procedure preferably includes a procedure for determining that a document is to be sent if the user has set automatic sending to the reporter extracted in the reporter extraction step. In addition, in order to prevent an inappropriate document from being sent, the procedure preferably includes a procedure for determining that a document is to be sent if the user has confirmed the document and authorized its sending.

In order to transmit the document in accordance with a distribution plan determined based on various circumstances, it is preferable that the procedure includes a procedure for determining that the document is to be transmitted when the timing is specified in the distribution plan. The distribution plan may be, for example, a distribution plan that matches the journalist's intent, a distribution plan that matches the optimal time, etc.

[Step S13: Send document]
The control unit 11 executes a process of transmitting the document generated in the document generation step to the reporter involved in the reporter extraction step by the document transmission unit 118 (step S13, document transmission step). The control unit 11 moves the process to step S14.

The procedure for sending the document is not particularly limited. In order to send the document by a means appropriate for the recipient, the document sending step preferably includes a procedure for sending the document via a channel associated with the reporter among multiple channels such as email, social media, and the web. In order to give a favorable impression to the document recipient, the document sending step preferably includes a procedure for sending the document accompanied by a message appropriate for the recipient. The message appropriate for the recipient may be a message based on a template stored in the memory unit 12, or may be a message generated by a large-scale language model based on the intent of the recipient, i.e., the reporter.

As part of the support process, the support device 1 preferably executes a series of processes to collect feedback from recipients of the document generated in the document generation step and utilize the feedback for improvements. Steps S14 to S16 are an example of such processes. Through such processes, the support device 1 can consistently support the collection and utilization of feedback from recipients who have received documents that contribute to intent sales.

[Step S14: Determine whether to collect feedback on the transmitted document]
The control unit 11, in cooperation with the storage unit 12 and the communication unit 13, executes a process of determining whether to collect feedback on the transmitted document (step S14, feedback collection determination step). If it is determined that feedback is to be collected, the control unit 11 moves the process to step S15. If it is determined that feedback is not to be collected, the control unit 11 returns the process to step S1 and repeats the processes from step S1 to step S16.

The procedure for determining in the feedback collection determination step is not particularly limited. The procedure includes, for example, a step of determining that feedback is to be collected when feedback is received, when feedback is stored in the memory unit 12, etc. Furthermore, in order to enable periodic feedback collection after sending a document, it is preferable that the procedure includes, for example, a step of determining that feedback is to be collected when a given time has elapsed since the document sending step or the previous feedback collection.

[Step S15: Collect feedback]
The control unit 11 cooperates with the storage unit 12 and the communication unit 13 to execute a process of collecting feedback on the transmitted document (step S15, feedback collection step), and the control unit 11 moves the process to step S16.

[Step S16: Execute machine learning]
The control unit 11 executes the machine learning unit 119 in cooperation with the storage unit 12 and the communication unit 13. The control unit 11 then executes a process of executing machine learning using the learning data including the data related to the document generated in the document generation step and the feedback collected in the feedback collection step, by the machine learning unit 119 (step S16, machine learning step). The control unit 11 returns the process to step S1 and repeats the processes from step S1 to step S16.

To improve the generation of documents according to the reporter's intent by the large-scale language model, the machine learning step preferably includes a procedure of fine-tuning the large-scale language model based on training data that includes the documents corrected based on the feedback as a target variable and the topics, the reporters, and the reporters' intent as explanatory variables. The correction of the documents based on the feedback may be performed manually, but is preferably performed by the large-scale language model to further reduce the effort.

To further improve the generation of documents according to the intent and the data stored in the database, the machine learning step preferably includes a procedure for fine-tuning a large-scale language model based on training data that includes the documents revised based on the feedback as the objective variable, and the topics, reporters, the intents of the reporters, and data related to the topics as the explanatory variables.

[Follow-up steps]
In order to measure the effect of intent sales using documents, the support process preferably includes a follow-up step of automatically following up on an article based on the document transmitted in the document transmission step. This step includes, for example, a procedure in which the control unit 11 cooperates with the storage unit 12 and the communication unit 13 to crawl an article written by a journalist to whom the document was sent on the Internet, and automatically follow up on the article when it is determined that the article is based on the transmitted document by processing using a large-scale language model or the like. This allows the support device 1 to automatically follow up on an article after the initial distribution.

[Engagement Tracking Step]
In order to continuously measure the effectiveness of intent sales through documents, it is preferable that the advertising effectiveness measurement step includes an engagement tracking step of tracking engagement and responses after the delivery of articles based on the documents sent in the document sending step.

[Report generation step]
When an engagement tracking step is executed to visualize the effects of intent selling through documents, it is preferable that the support process includes a report generation step of analyzing the effects of the article distribution by processing using a large-scale language model that includes the tracking results as input, and generating a report summarizing the results of the analysis.

[Content Update Step]
In order to sustain the effect of intent sales through documents, the support process preferably includes a content update step of generating documents that contribute to continuous updating of content (articles) based on various tracking results such as responses tracked in the engagement tracking step, etc., and transmitting the generated documents. The generation is realized in the same manner as the document generation step, except that various tracking results are included in the input.

[Obfuscation step]
In order to comply with privacy protection based on legal regulations, the assistance process preferably includes an obfuscation step for obfuscating data related to privacy protection among various data stored in the storage unit 12. This reduces the risk of insufficient privacy protection due to data leakage, etc.

<Usage example>
The following is an example of how the support device 1 of this embodiment is used.

[Data preparation]
The user works to store data related to a topic that the user wishes to cover in an article in the company database 122. This work is carried out by the user directly registering various data, instructing the support device 1 to collect data related to the topic, and the like.

[Intent Analysis]
The support device 1 processes signed articles and the like collected from the Internet using a large-scale language model, and analyzes intent, which is a tendency of articles posted by journalists. The support device 1 then stores the analysis result in the intent analysis database 121.

[Implementing Intent Sales]
A user specifies a topic and commands the support device 1 to conduct intent sales for the topic. The support device 1 extracts appropriate reporters based on the topic and the intent of each reporter. The support device 1 then generates a document including information that contributes to the creation of the article by the extracted reporter through processing using a large-scale language model based on the intent analysis result.

[Article Publication]
The support device 1 sends the generated document to a reporter. The reporter creates an article based on the document and publishes it on the Internet. The support device 1 follows up on the article and measures the advertising effect of the article. The support device 1 then performs machine learning related to a large-scale language model based on the article, etc., and feeds back the results for the next intent sales.

Note that within the scope of the concept of the present invention, a person skilled in the art may come up with various modifications and alterations. It is therefore understood that these modifications and alterations fall within the scope of the present invention. For example, even if a person skilled in the art appropriately adds, deletes, or modifies the design of the above-mentioned embodiment, or adds, omits, or modifies processes, these modifications are also included in the scope of the present invention as long as they incorporate the gist of the present invention.

S System 1 Support device 11 Control unit 111 Article collection unit 112 Intent analysis unit 113 Advertising effectiveness measurement unit 114 Analysis result management unit 115 Reporter extraction unit 116 Data acquisition unit 117 Document generation unit 118 Document transmission unit 119 Machine learning unit 12 Storage unit 121 Intent analysis database 122 Company database 13 Communication unit N Network T Terminal

Claims (5)

an article collection unit that collects articles from the Internet that are accompanied by information that identifies the reporter;
an intent analysis unit that analyzes intent, which is a tendency of articles posted by the journalists, by processing using a large-scale language model including the articles as input;
a document generation unit that generates a document including information that contributes to the creation of an article by the reporter through processing using a large-scale language model that includes an intent analysis result by the intent analysis unit as an input;
An intent sales support device comprising: an analysis result management unit that manages an analysis result of the intent;
a reporter extraction unit for extracting reporters related to a topic desired to be covered in an article by comparing the topic with the analysis result;
Further equipped with
The analysis result management unit manages analysis results relating to a plurality of reporters;
The reporter extraction unit extracts reporters related to the topic from the plurality of reporters;
The document generation unit generates a document including information that contributes to the creation of an article by the reporter extracted by the reporter extraction unit.
The support device according to claim 1 . An advertising effectiveness measurement unit for measuring advertising effectiveness relating to the article written by the reporter is further provided,
The reporter extraction unit extracts, from among the plurality of reporters, reporters related to the topic whose advertising effect satisfies a given condition;
The support device according to claim 2 . A data acquisition unit for acquiring data on a topic desired to be covered in the article from a database,
The data acquisition unit acquires, from the database, target data that is used by the reporter to write an article, among data related to the topic, based on a result of the intent analysis;
The document generation unit generates the document by processing using a large-scale language model that includes the target data as an input.
The support device according to claim 1 . The support device according to claim 1, further comprising a document transmission unit that transmits the document generated by the document generation unit to a reporter involved in the generation.