JP7824842B2 - News distribution system, news distribution method - Google Patents

Description

The present invention relates to a news distribution system and a news distribution method.

Fake news, intentionally or unintentionally false reports, malicious and unreliable information, and rumors have become a social problem. With the spread of smartphones and social media, these issues have become more apparent and familiar to us. Newspaper articles have also emphasized that excessive government guidance on freedom of speech risks creating a controlled society, and that transparency on the part of businesses is important in distributing fair and accurate information. The government has also stated that "intervention should be extremely cautious in the interest of freedom of expression," and that it places the basic priority on autonomous problem-solving within the private sector.

Generally, media businesses rely on advertising fees from sponsors to survive, and it is important to write articles that attract readers or page views (PVs). This is what is known as the attention economy. News articles are generally consumed instantly, and in an environment overflowing with diverse information, there is a danger that breaking, interesting information will be presented without any certainty in order to capture users' attention, i.e., to increase PVs. For example, there are cases where information posted on social media is republished without verifying its authenticity.

From the perspective of social responsibility, the media strives to write fair and accurate articles, in part to prevent the spread of rumors, and is considering technical methods to support this. For example, Patent Document 1 discloses technology that extracts features from articles and takes into account trends and the number of views to determine whether or not information should be published. Patent Document 2 proposes technology to improve the accuracy of articles by using a system that awards or denies points to users who publish articles based on a determination of their authenticity, and also proposes technology to confirm the authenticity of information within a user network, such as a method in which users of SNSs check each other's authenticity.

Japanese Patent Application Publication No. 2020-024489 Japanese Patent Application Publication No. 2019-185768

The spread of social media has made it possible for information to be provided directly from the source to the user. While traditional media's advantage in terms of speed and the transmission of unvarnished information is being challenged, they are expected to play a role in collecting and organizing information and assembling diverse opinions. However, in a news format where pageviews are important, there is little incentive to spend a long time meticulously writing fair and accurate news articles. Fairness, however, is slightly different from accuracy and factuality and includes subjective concepts. Even a single fact can have many different meanings depending on the means of transmission and the recipient's background and perspective, resulting in many different pieces of information that "seem correct." It has been pointed out that users are at risk of being exposed to fake news in the process of searching for accurate information. In particular, articles that seem reasonable based on their own information literacy are easily mistaken for the truth. In other words, we need to create a news distribution system that eliminates obvious distortions, ensures accuracy, and also ensures fairness. To achieve high fairness, we need transparency in the process, including the selection and rejection of information, that is, transparency in clarifying the perspectives and positions from which one's opinions are based.

The purpose of this invention is to provide technology that ensures transparency of the sender's position and enables the delivery of accurate and fair news.

The news distribution system according to the present invention is a news distribution system in which a computer executes a program to distribute predetermined news, and includes a buzz news detection unit that monitors news sites on a network and detects buzz news in which quotes or comments on news articles published on the news sites satisfy predetermined conditions; a sentiment information collection unit that collects transaction data including comments on the buzz news and cumulative values of the comments that are posted on the network in association with the buzz news; a sentiment analysis unit that analyzes the comments included in the transaction data and, if the number of clusters obtained by clustering opinions included in the comments satisfies predetermined conditions, determines that the news article to which the comments are attached is a news article with high fairness; and a media analysis unit that receives information on the news article from the network. a related article collection unit that collects related articles including primary sources before editing; an accuracy determination unit that determines the accuracy of the news article based on the news article determined by the sentiment analysis unit and the related articles collected by the related article collection unit; a fairness determination unit that calculates a fairness score that quantifies the fairness of the news article determined by the sentiment analysis unit and the accuracy of the news article determined by the accuracy determination unit; a fairness score management unit that manages a fairness management ledger that associates the fairness scores with the news articles and a fairness contribution management ledger that associates users with fairness contributions, which are fairness scores given to users when they perform actions that contribute to an increase or decrease in the fairness score; and a fairness reward distribution unit that uses the fairness management ledger and the fairness contribution management ledger to reward users who perform actions that increase the fairness of the news article.
The news distribution system is configured as having the following features:

This invention ensures transparency of the sender's position and makes it possible to deliver accurate and fair news.

Overall processing Processing flow Functional Block Diagram Interface Screen Accuracy judgment section Fact Check Department Fairness judgment department Fairness management ledger for each news article Expert Database Management ledger of each user's fairness contribution Data Blocks News Database Hardware Configuration Processing by the accuracy determination unit when information is provided by the user

The following describes examples using the drawings. Note that the embodiments described below do not limit the invention as claimed, and not all of the elements and combinations described in the embodiments are necessarily essential to the solution of the invention.

In other words, the following description and drawings are examples for explaining the present invention, and have been omitted or simplified as appropriate for clarity of explanation. The present invention can be implemented in various other forms. Unless otherwise specified, each component may be singular or plural.

The position, size, shape, range, etc. of each component shown in the drawings may not represent the actual position, size, shape, range, etc. in order to facilitate understanding of the invention. Therefore, the present invention is not necessarily limited to the position, size, shape, range, etc. disclosed in the drawings.

In the following explanation, various types of information may be described using terms such as "database," "table," and "list," but the various types of information may also be expressed in other data structures. To indicate that the information is not dependent on the data structure, "XX table," "XX list," etc. may be referred to as "XX information." When describing identification information, terms such as "identification information," "identifier," "name," "ID," and "number" are used, and these terms are interchangeable.

When there are multiple components with the same or similar functions, they may be described using the same reference numeral with different subscripts. However, when there is no need to distinguish between these multiple components, the subscripts may be omitted.

In addition, the following description may describe processing performed by executing a program. However, since a program is executed by a processor (e.g., a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit)) to perform a specified process using storage resources (e.g., memory) and/or interface devices (e.g., communication ports) as appropriate, the processor may be the entity performing the processing. Similarly, the entity performing the processing by executing a program may be a controller, device, system, computer, or node having a processor. The entity performing the processing by executing a program may be any computing unit or computing device, and may include a dedicated circuit that performs specific processing (e.g., an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit)).

A program may be installed on a device such as a computer from a program source. The program source may be, for example, a program distribution server or a computer-readable storage medium. If the program source is a program distribution server, the program distribution server may include a processor and storage resources for storing the program to be distributed, and the processor of the program distribution server may distribute the program to be distributed to other computers. Also, in the following description, two or more programs may be realized as one program, and one program may be realized as two or more programs.

In Example 1, we will explain a configuration that realizes the process leading up to the distribution of highly fair news, as explained in the solution to the problem.

The overall flow of the fair news distribution system is shown in Figure 1. The fair news distribution system 1000 (Figure 3) detects news that has attracted a lot of attention and received a lot of feedback (hereinafter referred to as buzz news) (S101). Next, when the fair news distribution system 1000 detects buzz news, it collects public reactions to the news article included in the buzz news, analyzes comments and reactions posted by users, and performs sentiment analysis to compile and visualize the results as statistical data, thereby visualizing diverse opinions (S102).

Next, the fair news distribution system 1000 evaluates the accuracy and fairness of the news article detected in S101 (S103). Evaluating the fairness of a news article involves assessing whether it incorporates information from diverse perspectives. To perform a fairness evaluation, the fair news distribution system 1000 preferably detects whether the content of the news article includes arguments from multiple perspectives and is comprehensively analyzed. The accuracy of a news article is evaluated by searching various websites online for the original source on which the news article is based, such as a written or audio transcript of the sender's entire speech. The search results, including numbers and original sources, are compared with the news article being evaluated to check whether the intent has been altered. Furthermore, the fair news distribution system 1000 evaluates whether the information on which the news article is based is disclosed, and if so, the reliability of that information. For example, one criterion for evaluating reliability is whether the information on which the news article is based has been peer-reviewed for public publication, such as in a journal or a reputable academic conference. In this case, in the case of academic societies, the higher the number in an indicator called the impact factor, the more reliable the information is considered to be, and in the case of academic papers, the number of citations can also be used as an indicator. The fair news distribution system 1000 may automatically detect these using a program, but it is also possible to send the information that is the basis of the news article in order to inquire with an industry expert, have that expert fact-check the information, and then receive an evaluation and comment from that expert after fact-checking the information.

The fairness score is a comprehensive numerical representation of the evaluation results for both fairness and accuracy. If a discourse only covers a portion of diverse opinions and is not comprehensive, the fair news distribution system 1000 will calculate a low fairness score, assuming that it does not meet certain standards. On the other hand, if the information on which the judgment is based is clear and the factuality is high to a certain extent, the fair news distribution system 1000 will evaluate the accuracy higher and calculate a high fairness score, assuming that it meets certain standards. In other words, even if a news article is biased, it can be made transparent by covering multiple perspectives and clearly stating which of those perspectives the author's opinion is based on. The more factual an article is, and the more logically explained it is based on the information on which the news article is based, the higher the fairness score it will have.

After evaluating the fairness and accuracy of a news article in S103, the fair news distribution system 1000 formats the information contained in the news article, generates a site that serves as a reference information source for users to view, and distributes the news article (S104). It is assumed that the reference information source will be updated on a regular basis. For example, immediately after buzz news is detected, it is assumed that expert fact-checking has not yet been completed and accuracy determination is incomplete. Even in this case, the fair news distribution system 1000 will create a thread for the buzz news being verified on the reference information source site, describing the progress of the fact-check as "under determination." This thread will include objects that display the results of the sentiment analysis performed in S102 and the evaluation results of accuracy and fairness performed in S103. Each of these objects will be described below using Figure 4.

In this manner, the fair news distribution system 1000 scores the fairness contribution (described below) of users who contributed to generating fair and accurate information in the process of generating reference information sources. The fair news distribution system 1000 then provides rewards according to the scores, providing incentives to the framework for providing fair information (S105). The fair news distribution system 1000 continues processing from S102 to S105 until the fairness score reaches or exceeds a predetermined threshold or a predetermined time has elapsed (S106). If the fair news distribution system 1000 determines that the fairness score has not reached or exceeded the predetermined threshold or that the predetermined time has not elapsed (S106; No), it returns to S102 and continues the subsequent processing. On the other hand, if the fair news distribution system 1000 determines that the fairness score has reached or exceeded the predetermined threshold or that the predetermined time has elapsed, it terminates processing.

The more attention the buzz news receives, the more closely the fair news distribution system 1000 may track people's reactions by shortening the interval between checks of the sentiment analysis information 502 in S102. Alternatively, in S103, the fair news distribution system 1000 may increase the number of people requested to fact-check, or may exercise control to encourage the provision of data for verification, in order to improve the accuracy of the accuracy judgment.

The sequence for realizing the above processing and the functional block diagram for realizing it will be explained using Figures 2 and 3. Figure 2 is a sequence diagram specifically illustrating the processing flow shown in Figure 1. Figure 3 is a block diagram showing the functional configuration of the fair news distribution system 1000. As shown in Figure 3, the fair news distribution system 1000 is composed of a back-end unit 1001, a front-end unit 1002, and a fairness management unit 1003.

These may be operated together, but to ensure transparency, it is advisable to configure the parts relating to increasing or decreasing fairness and distributing rewards to actions that contribute to fairness so that they can be disclosed. For example, the front-end unit 1002 and back-end unit 1001 could be separate nodes on the same platform, and the fairness management unit 1003 could be separated as a separate node so that its operation can be entrusted to a third-party organization. It is advisable to configure the communications between each node so that they can be disclosed. This is thought to be compatible with transactions using smart contracts on Blockchain B and data management using ledgers.

As shown in Figure 3, this system is characterized by having a buzz news detection unit 201 that analyzes the original sources on which buzz news articles are based and traces the origin of the original sources; a fact-checking unit 207 that requests fact-checking from experts to confirm that the news does not contain misinformation; a sentiment analysis unit 203 that evaluates the bias in the discourse of the original sources of the buzz news; a fairness determination unit 206 that evaluates the fairness of the buzz news articles and the reporters who wrote them; a display information generation unit 209 that outputs the output results obtained from the processing of these units as a viewable screen or website along with past history; and a fairness reward distribution unit 210 that manages the fairness evaluation results in a ledger on Blockchain B and distributes tokens as rewards based on fairness. Hereinafter, there are two types of fairness scores: a fairness score assigned to a news article and a fairness score assigned to a person. A fairness score is assigned to a person when they perform an action that contributes to an increase or decrease in the fairness score assigned to a news article. Therefore, the fairness score assigned to a person is called the fairness contribution.

This system can be implemented using general-purpose information processing devices such as servers and terminals. An example of the hardware configuration will be described later using Figure 11.

Furthermore, the various data stored in the information processing device or used for processing can be realized by the CPU reading it from memory and using it. Furthermore, each functional unit of the information processing device (e.g., the buzz news detection unit 201, fact check unit 207, sentiment analysis unit 203, fairness determination unit 206, fairness reward distribution unit 210, etc., each functional unit shown in FIG. 3) can be realized by the CPU loading a specific program stored in an external storage device into memory and executing it.

In the following, we will explain an example in which this system is configured using a single computer, but similar functions may also be realized by distributing all or part of these functions across one or more computers, such as a cloud, and communicating with each other via a network.

The blocks that realize the above functions will be explained in detail below. Here, let's assume that a media server 1200 distributes news articles via the Internet, which has become commonplace in recent years, and that the distributed news articles attract a large amount of user interest on a certain portal site 1300. In fact, existing news media also have experts and users comment on news articles that attract a lot of attention (S300).

The buzz news detection unit 201 of this system detects news that has attracted a large amount of interest in S300 as buzz news (S301). The buzz news detection unit 201 monitors news sites and SNS where news articles are posted, and detects whether the number of citations and comments on the news article exceeds a predetermined threshold. Of course, if a news article that meets predetermined conditions is detected, it can be assumed that buzz news has occurred. Hereinafter, when referring to news sites, this is taken to include various network services where news articles are posted, such as SNS.

Buzz news indicates high public interest, and it is assumed that there is a high demand for accurate and unbiased information. There are two actions that can be taken to obtain unbiased information: assessing the accuracy and fairness of the news article itself, and collecting public opinion to reflect diverse opinions and republish the information after filling in any missing perspectives.

The sentiment information collection unit 202 then collects user reactions (S302). The sentiment information collection unit 202 collects transaction data, including comments and cumulative reaction button values on news sites, portal sites, and SNS, as user reactions. Site viewing formats include not only internet browsers but also apps. Buzz news often sees a large number of quote buttons, such as "Like" and retweet, pressed. Quotations often are followed by comments, and it is believed that the sentiment information collection unit 202 can efficiently collect user comments associated with the buzz news by following the links provided by the buzz news.

The sentiment analysis unit 203 analyzes, visualizes, and quantifies the user reaction information collected as described above (S303). The evaluation axis used for visualization depends on the topic of the news article; for example, for a political topic, attributes include right-wing/left-wing, hawkish/dovish, and support/opposition to a policy. For other evaluation axes, sentiment analysis using, for example, sentiment analysis, moral analysis, or surprise analysis as evaluation axes may be used. The sentiment analysis unit 203 can analyze and visualize user reactions from a variety of perspectives in this way. The sentiment analysis unit 203 can classify these using machine learning, but several technologies that contribute to performance improvement are required, such as extracting keywords from news articles, performing name matching, and removing noise from keywords that have different meanings depending on the context.

The sentiment analysis unit 203 may collect user responses to the extent that multiple opinion clusters can be formed, thereby visualizing the diversity of opinions. The sentiment analysis unit 203 clusters the diverse user opinions, analyzes and visualizes which opinion cluster the buzz news reporter belongs to, and the perspective from which the article provides information, and thereby makes it possible to grasp the bias of the news in an aware state. This is considered to be a state of high fairness. For example, the sentiment analysis unit 203 clusters the diverse user opinions regarding a news article, and if the number of clusters obtained by this clustering is equal to or greater than a predetermined threshold, it determines that the news article is a news article with high fairness. For the above clustering, various conventionally known techniques may be used, such as Ward's method, group average method, shortest distance method, and furthest distance method.

Meanwhile, the related article collection unit 204 collects related articles related to the news article in addition to the original news article contained in the buzz news detected in S301 (S304). Here, related articles refer to, for example, the original source used as the news subject, past similar news and analysis articles, and literature and distribution that forms the basis of the news article. Distribution that forms the basis of a news article refers to raw data before media editing, including primary sources such as government and corporate public relations releases. The related article collection unit 204 also targets audio and video data, and it is recommended to incorporate information other than text media by using techniques such as transcription. Transcription technology has made great advances in recent years, and it is thought that audio and video data that has been transcribed into text can also be used in practical applications by passing it through an appropriate noise filter.

This information may be collected manually by the system operator, according to a predetermined algorithm, or by relying on user posts. Relying on user posts can be achieved, for example, by the fairness information user interface (UI) 208 accepting user posts and sending the content to the system, or by the sentiment analysis unit 203 extracting the information in the process of analyzing posted comments. The sentiment analysis unit 203 collects information on these related articles in the article DB 3001. The accuracy determination unit 205 then analyzes the related articles, and the display information generation unit 209 outputs the results in a list for display on the screen of the fairness information UI 208.

The accuracy determination unit 205 diagnoses the accuracy of the news article based on the news article for which reaction information was collected in S302 and the related articles collected in S304 (S305: Accuracy evaluation). When making the diagnosis, the accuracy determination unit 205 may tag and extract summaries from the news articles and literature collected by the related article collection unit 204, and store and manage them in the article DB 3001 so that they can be searched for as similar news in the past. The accuracy determination unit 205 evaluates the accuracy of the news article by performing processing according to the flow shown in Figure 5. In this processing, the accuracy determination unit 205 diagnoses whether the argument of the news article is consistent with the related articles, which are the collected primary sources, and whether the news article is written logically. The purpose is to prevent news articles that manipulate impressions by extracting only parts of the primary source.

Figure 5 is a flowchart showing an example of the accuracy evaluation process in S305 shown in Figure 3. As shown in Figure 5, the accuracy determination unit 205 analyzes a news article by referring to the article DB 3001 (S501), and determines whether the news article contains a description disclosing a primary source and whether the primary source can actually be accessed and viewed (S502). If it is determined that the above-mentioned viewing is possible (S502; Yes), the fairness score of the news article is added (S503), and the primary source is analyzed (S504).

On the other hand, if the accuracy determination unit 205 determines that the above-mentioned browsing is not possible (S502; No), it determines whether the related article collection unit 204 has searched for a primary source (S505). If the accuracy determination unit 205 determines that the related article collection unit 204 has searched for a primary source (S505; Yes), in S504, it analyzes the primary source searched by the related article collection unit 204. When analyzing a primary source, the accuracy determination unit 205 reads each of the primary source and the news article, and compares and analyzes the two to detect missing arguments, different interpretations, and simple errors. These processes may be performed algorithmically or by the operator of the system. To ensure transparency, the accuracy determination unit 205 records the process of how it made its determination (for example, the order and method of the above-mentioned comparison, and the points where the two differ). When using machine learning, it is recommended to adopt an explainable model, such as one that displays the importance of features.

Based on the analysis results of S504, the accuracy determination unit 205 determines whether the content of the primary source has been altered (S506). If the accuracy determination unit 205 determines that the content of the primary source has not been altered (S506; No), it adds to the fairness score (S507). On the other hand, if the accuracy determination unit 205 determines that the content of the primary source has been altered (S506; Yes), it significantly reduces the fairness score by a value equal to or greater than a predetermined threshold, and regards it as malicious reporting (S508).

The accuracy determination unit 205 outputs a fairness score for accuracy that reflects the values calculated in S507 and S508 (S509). Furthermore, it outputs a fairness contribution corresponding to the fairness score for the reporter of the analyzed news article (S510).

Returning to Figure 2, the accuracy determination unit 205 further outputs the primary source and the news article to the fact checking unit 207, which then diagnoses the accuracy of the news article. When diagnosing accuracy, it is important to incorporate the opinions of external experts. Therefore, the fact checking unit 207 sends the primary source and the news article to the expert's website or terminal and issues a fact checking request (S305: Fact Check Request). When selecting who to send the fact checking request to, the fact checking unit 207 may use the criteria for selecting an expert whose profession matches the field of the news article based on a pre-prepared expert DB (database) 3002, or it may select from experts who have a track record of contributing to improving fairness in the past regarding similar news, based on a ledger managed in a distributed manner on the blockchain.

Figure 8B is a diagram showing an example of an expert DB. As shown in Figure 8B, expert DB 3002 associates an expert ID for identifying the expert, the expert's name, the expert's contact email address, the expert's occupation, and past performance indicating whether the expert has conducted fact-checking in the past. In Figure 8B, for example, expert "T.Y." identified by expert ID "S0000001" is a university professor, and the email address registered as his or her contact address is "α," indicating that he or she has received requests for fact-checking in the past.

An expert who receives a fact-checking request from this system will respond with the information that serves as the basis for their argument and the resulting opinion by operating a computer such as a fact-checking terminal or server, and send this to this system as the fact-checking result (S307). Because a variety of opinions is necessary from the perspective of fairness, in S305 the fact-checking unit 207 may send the primary source and news article to the expert's website or terminal and seek the opinions of multiple experts. The information presented by the expert and the source of its acquisition can also be used as a data collection source for the related information collection unit 204, and the fact-checking unit 207 may register this information in the article DB 3001 for future use.

Figure 6 is a flowchart showing an example of the processing related to the fact-check request in S305 shown in Figure 3. As shown in Figure 6, the fact-checking unit 207 analyzes the genre of the primary source and news article to be sent (S601), and selects an expert who specializes in the genre obtained by the analysis from the expert DB 3002 (S602). The fact-checking unit 207 sends the primary source and news article to be sent to the email address or website associated with the selected expert, requesting a fact check (S603).

The fact check unit 207 receives the fact check results from the expert who was requested to perform the fact check, and determines whether the content was evaluated as good or appropriate (S604). Whether the content was evaluated as good or appropriate can be determined, for example, as a negative evaluation or an evaluation of inappropriateness if the fact check results include negative comments such as the news article being highly likely to be false, or as a good evaluation or appropriateness if the news article is evaluated as positive and the like.

If the fact-checking unit 207 determines that the content was rated good or appropriate (S604; Yes), it determines that the news article is accurate, adds a predetermined value to the fairness score of the news article (S605), and proceeds to S606. On the other hand, if the fact-checking unit 207 determines that the content was not rated good (S604; No), it determines that the news article is inaccurate, and proceeds to S606 without doing anything.

When the fact-checking unit 207 adds a predetermined value to the fairness score of the news article, it updates the fairness score (accuracy) of the article to the fairness score to which the predetermined value has been added (S606), and adds a predetermined value to the fairness contribution of the expert who performed the fact-check, updating that value (S607). Through the above processing, it is possible to calculate the accuracy and fairness contribution, which are indicators of the fairness score.

Returning to Figure 2, the fairness determination unit 206 calculates a fairness score by comprehensively quantifying the evaluation results of both fairness and accuracy (S308). The fairness determination unit 206 calculates a low fairness score if the discourse only covers a portion of the diverse opinions and is not comprehensive. On the other hand, if the information on which the judgment is based is clear and highly factual, the accuracy is evaluated higher and a high fairness score is assigned. In other words, even if an article is considered to be biased, if it covers multiple viewpoints, is transparent in clarifying its position as to which viewpoint its opinion is based, and provides clear supporting data and uses the argument correctly, it will be considered to be logical, i.e., highly factual, and will be judged to have a high fairness score.

For example, in S308, the fairness determination unit 206 may calculate a fairness score using a flow such as that shown in FIG. 7. The fairness determination unit 206 obtains various opinion clusters obtained through analysis by the sentiment analysis unit 203 (S701). The fairness determination unit 206 analyzes the intention of each opinion cluster, classifies the viewpoints, and assigns identifiers to users belonging to each opinion cluster. For example, the fairness determination unit 206 associates multiple user identifiers (user IDs 001, 002, etc.) with a certain opinion cluster A. The fairness determination unit 206 also associates multiple user identifiers (user IDs 001, 003, etc.) with another opinion cluster B. The fairness determination unit 206 then analyzes the buzz news article and calculates the similarity between the two to determine which opinion cluster the news article was written from (S703).

Furthermore, the fairness determination unit 206 determines whether the buzz news article reaches a conclusion after referring to opinions belonging to multiple opinion clusters. In other words, the fairness determination unit 206 determines whether the news article has been analyzed from a variety of perspectives (S704).

If the fairness determination unit 206 determines that the news article has been analyzed from a variety of perspectives (S704; Yes), it determines that the news article is unbiased and adds a fairness score (S705). Bias in this system means that the news article is summarized from a single perspective. The fairness determination unit 206 considers a news article to be unbiased and fair if it touches on the arguments of multiple perspectives and then reaches a conclusion. For example, if a news article satisfies the condition that it is an article that describes the views of more than half of the opinions in multiple opinion clusters and then expresses its own opinion, the fairness determination unit 206 determines that the news article is unbiased and fair. Furthermore, when the fairness determination unit 206 satisfies the above condition and adds a fairness score, the greater the number of opinion clusters, the higher the fairness score added.

In summary, a fair news article in this system is one that meets two requirements: it is free of unnecessary alterations from the primary source, and it reduces bias by considering multiple perspectives before reaching a conclusion. Anyone who takes actions that make up for shortcomings in these requirements (or actions that improve these requirements to a certain extent) is awarded a fairness score, or fairness contribution.

As explained above, this system reveals the fairness of an article in the process of calculating its fairness score. In doing so, the system also takes into account the fairness contribution, and rewards users with tokens based on their fairness contribution. This provides an incentive for the system's operators to operate a system that benefits the public.

The diversity of perspectives contained in the opinion clusters obtained through sentiment analysis in S701 contributes significantly to fairness. Therefore, a mechanism may be implemented to redistribute tokens acquired by the system to users who provide information constituting the opinion clusters (312). To achieve this, ID linkage is required between the system, other portal sites or SNS servers 1300 where comments can be posted, and users who provided the information by operating user terminals 1400. That is, in S302, the sentiment information collection unit 202 collects the user IDs of users who commented on the buzz news detected in S301 (i.e., the comments added in S300 shown in FIG. 2) along with their reactions, and registers them in the system. Linking user IDs allows users who provided the information (or who intend to provide the information in the future) to casually respond to their reactions on the SNS, and by receiving tokens redistributed, they recognize that they have participated in an action to improve fairness, thereby automatically receiving incentives. This is expected to encourage user participation.

Furthermore, in S304 shown in FIG. 2, the related article collection unit 204 collects related articles in addition to the original news article included in the buzz news, but the media server 1200 may issue a correction via the information provision interface 508 in FIG. 4 or online or offline (S306a). Normally, when media issues a correction as a follow-up, it not only reduces the space available for transmitting new information but also requires a great deal of effort. Therefore, in this system, the issuance of a correction is evaluated as an action to improve fairness, and in S312, tokens are distributed as a reward to the media as well. This encourages the media to issue corrections and disseminate additional information themselves, enabling a spliced-in type of news distribution that goes beyond the traditional method of transmitting information at the fastest speed.

It is recommended that fairness scores be managed using a blockchain ledger, so that changes made to the system can be tracked and transparency can be ensured. This will create an environment that is less susceptible to information manipulation, while simultaneously ensuring transparency regarding actions that contribute to improving fairness and providing the gamification element of collecting tokens. Controlling fairness contribution is sensitive and high-risk information that influences public opinion, so robustness is more important than scaling, and a private system is considered advantageous. It is desirable for this information to be operated and managed by a non-profit organization.

The fairness reward distribution unit 210 and fairness score management unit 211 manage users' fairness contributions and news article fairness scores, and control transactions related to these increases and decreases on blockchain B. Along with these transactions, they record evaluations of actions that contributed to fairness and the amount of score updates, along with the reasons for them. The data managed in the distributed ledger is shown in Figures 8A and 9.

The fairness reward distribution unit 210 and the fairness score management unit 211 award fairness contribution points to experts who cooperate with fact-checking. Fairness contribution points may be issued, for example, using tokens circulating on the blockchain. It is desirable that the fairness contribution points and the history of token transactions be managed on the blockchain to ensure transparency and resistance to tampering. As mentioned above, the fairness reward distribution unit 210 and the fairness score management unit 211 may be operated by a fairness management unit that is independent from the backend logic unit. If the provider of the backend logic unit and the provider of the fairness management unit are the same, there is no need to separate them.

The fairness score management unit 211 may store a fairness management ledger 801 for each news article, as shown in FIG. 8A. The fairness management ledger 801 is a table for managing the fairness of news articles. As shown in FIG. 8A, the fairness management ledger 801 stores, in association with one another, a news ID for identifying buzz news, the fairness score of a published news article included in the buzz news, a raw score which is the fairness score of the original news article of the buzz news, the last update date of the fairness score, and a transaction ID for identifying the transaction data when reactions to the news identified by the news ID were collected. The raw score is the fairness score before reflecting any increase or decrease in fairness contribution to the fairness score of the original news article of the buzz news.

In Figure 8A, for example, the fairness score of a news article published as buzz news identified by news ID "N0000001" is calculated to be "80," and the raw score of the original news article for the buzz news was "55." The fairness score of "80" was most recently updated on May 20, 2021, and was updated using transaction data identified by transaction IDs "T00000001, T00000003." As described below, by displaying the evaluation of accuracy and fairness (fairness of the news article) as a result of sentiment analysis on the same fairness information UI 208, the fairness score can be viewed from an objective perspective, thereby improving the fairness of the news article. Note that the raw fairness score is assumed to be the value initially calculated by the accuracy determination unit 205.

The accuracy determination unit 205 may write to the fairness management ledger 801 for the news article through a specified API (application programming interface) configured on the blockchain. The fairness determination unit 206 calculates the fairness score for the news article, as described in S704, etc., and updates it through the specified API.

When the fact-checking unit 207 receives the results of the fact-checking from the expert, it may carry out procedures for updating the fairness score of the news article via the specified API. The procedure for updating the fairness score for these news articles involves exchanging information using data block 9001 exchanged on the blockchain, as shown in Figure 10, for example. Here, we have described updating the fairness score for a news article, but the same can be thought of for the fairness contribution, which is the fairness score for a person.

Figure 10 is a diagram showing an example of a data block included in a token exchanged on a blockchain. As shown in Figure 10, data block 9001 is a history ledger of data blocks exchanged for each transaction ID. Data block 9001 is data that associates the above-mentioned transaction ID, the date on which the block data constituting the transaction data identified by the transaction ID was exchanged on the blockchain, the type of increase/decrease in the fairness score, the amount of increase/decrease in the fairness score identified by the type, the evaluation action that contributed to the increase/decrease in the fairness score, the news ID of the news article for which the fairness score was calculated, the action date/time which is the date/time when the above-mentioned evaluation action was performed, and the user name and user ID who performed the above-mentioned evaluation action.

In Figure 10, for example, the block data constituting the transaction data identified by transaction ID "T00000002" indicates that user ID "U0000022" (user name: T.Y.) fact-checked news article "N0000002" as an expert on May 18, 2021, and that a fairness score (here, the sum of the fairness score for the news article and the fairness contribution, which is the fairness score for the person) was calculated on the same day, with the value set to "+15." In this way, this system updates fairness scores in the same format on the blockchain, allowing data regarding fairness evaluations of various news articles to be exchanged with the same level of security. Note that user names may be nicknames.

The fairness score management unit 211 maintains a management ledger that manages the fairness contribution for each user, as shown in Figure 9. When the fairness score management unit 211 updates the fairness data related to a news article, it updates the fairness contribution information managed for each user. The fairness contribution management ledger 9101 is a table for managing the fairness contribution, which is the fairness score for a person.

As shown in Figure 9, the fairness contribution management ledger 9101 associates and stores the above-mentioned user ID and user name, a type indicating the occupation of the user identified by the user ID and user name, the last update date of the fairness contribution, the fairness contribution history of the user, the amount of tokens (token amount) acquired by the user, and the transaction ID of the transaction data that is the basis for acquiring that amount of tokens.

In Figure 9, for example, a user with the username "T.Y." identified by the user ID "U0000022" is an expert in fact-checking, and the fairness contribution was calculated as "1500" with a last update date of May 18, 20201. It also shows that a token amount of "1485" was calculated for this user in the past on May 10, 2021. As will be described in more detail later, by displaying the results of the sentiment analysis, as well as evaluations of accuracy and fairness (fairness toward individuals), on the same fairness information UI 208, the fairness score can be viewed from an objective perspective, thereby improving the fairness of news articles.

The fairness score management unit 211 approves updates to the ledger that manages the fairness score information for each user, based on the evaluation actions taken by each user to improve fairness and the amount of increase or decrease in fairness for the original news article in Buzz News. Processing performed by the system is also managed as a fairness score. When the accuracy determination unit 205 searches and discovers a primary source, or when the fairness determination unit 206 is able to add a new perspective that was not present in the news article, it recognizes that the system has contributed to improving the fairness score and awards a fairness contribution. The fairness reward distribution unit 210 prepares tokens that circulate on the blockchain and are linked to fairness contributions, and distributes them as rewards to blockchain participating users.

The problem with this system is how to ensure economic rationality for the participation of media and users. As this system has a highly public nature, it is required to be run by a non-profit organization and have a structure that is less susceptible to influence from specific sponsors such as political or corporate sponsors. Therefore, the gamification of collecting tokens is used as one incentive. As an example, it is advisable to maintain tables such as those shown in Figures 8A to 10, and manage fairness scores and tokens on the blockchain, so that they can ultimately be exchanged for currency outside the blockchain. For example, the fairness reward distribution unit 210 may manage the tokens in a format that allows them to be exchanged for other currencies, based on the amount of increase or decrease in fairness contribution.

In addition, for example, in the GUI shown in Figure 4, donations obtained through the donation interface 507 or other routes can be pooled and exchanged for tokens collected by each user. This allows for movement between economic activities managed on the blockchain and economic activities in the real world. In this case, the fairness reward distribution unit 210 can provide a series of mechanisms related to the exchange in cooperation with an external system.

Returning to Figure 2, the display information generation unit 209 compiles information in the form of a news aggregation site so that a variety of information can be viewed at a glance as a reference information source (S309). For example, the display information generation unit 209 constructs a screen for a site with a fairness information UI 208 as shown in Figure 4. When constructing the screen, the display information generation unit 209 generates threads that reflect information calculated at the intermediate points in time by each unit, such as the fairness determination unit 206, sentiment analysis unit 203, accuracy determination unit 205, and fact check unit 207, and stores the generated threads in the display DB 3003. In Figure 4, the above screen includes a thread 401 updated on May 15, 2021, which is stored in the display DB 3003. When constructing the screen, the display information generation unit 209 reads the threads stored in the display DB 3003 and generates the above screen. For example, when the input of a keyword "safety of coronavirus vaccine" is received, threads about buzz news articles that are somewhat similar to the keyword are extracted from the display DB 3003 and displayed on a single screen. The above screen may be provided by registering the output results of each unit, such as the news article, sentiment analysis unit, information verification unit, and fairness determination unit, along with past history, on a viewable website. The above screen will be described in detail below.

Figure 4 is a screen showing an example of a fairness information UI. As shown in Figure 4, the fairness information UI 208 may display information on primary sources collected by the sentiment analysis unit 203 and a Verification object 501 with a link to allow reference to similar past news. The Verification object 501 may also display information on the fairness contribution calculated by the accuracy determination unit 205. In Figure 4, the display information generation unit 209 indicates that for a buzz news article titled "Safety of the Coronavirus Vaccine," two references, Reference [1] ○○ and Reference [2] △△, are displayed in the fairness information UI 208 as information collected as primary sources. It also indicates that the fairness contribution for the news article has been calculated as "+5".

The accuracy determination unit 205 also analyzes the accuracy of the original news article of the buzz news article. Therefore, the display information generation unit 209 may display the result in the fairness information UI 208 as the raw score before improving fairness, alongside the text of the original news article of the buzz news, as in object 503. In Figure 4, it can be seen that the display information generation unit 209 displays "65" as the raw score in association with the text of the original news article of the buzz news written by reporter A.

The display information generation unit 209 may also display an object 504 in the fairness information UI 208 that displays the status of the fact-check assessment by experts. The display information generation unit 209 reads the fact-check results output from the fact-check unit 207 and classifies them into three states: currently being assessed, assessed, and not yet assessed. When the result is assessed, the display information generation unit 209 determines that the fairness of the displayed content of the page has increased, adds information about the fairness contribution rate, and updates the display. In Figure 4, it can be seen that the status is "being assessed," indicating that the fact-check is currently being carried out.

The display information generation unit 209 may also display an object 502 that illustrates, as the output result of the sentiment analysis unit 203, opinion clusters, a summary of reactions, and meters showing emotions such as anger and joy. In Figure 4, an emotion meter is output indicating that, of the opinion clusters obtained by processing by the sentiment analysis unit 203, there are more opinion clusters indicating "joy" than opinion clusters indicating "anger." The emotion expressed by an opinion cluster can be determined, for example, by summarizing the viewpoints of each opinion cluster classified by the fairness determination unit 206 through semantic analysis. If, as a result of the sentiment analysis provided by this system, an opinion cluster is formed from a viewpoint not present in the article, it is advisable to display the fairness contribution degree of the users who posted the opinions that make up that opinion cluster.

Furthermore, the display information generation unit 209 may have a mechanism for displaying, in the fairness score transition object 505, how the fairness of the display content of the page has been updated in chronological order when the fairness contributions assigned by each functional block are added, starting from the raw score information of the original news article of the buzz news (Figure 8A). For example, it may be possible to display the fairness contributions added by analyzing the original news article of the buzz news and the current overall fairness score of the display information of the page. The overall fairness score is the fairness score at the starting point plus the added fairness contribution (+15), and in this example is displayed as "80." In this way, the raw score of the original news article of the buzz news and the fairness score after the fairness contributions have been subsequently increased or decreased are displayed as the current overall fairness score, and the transition of these increases or decreases is output on the screen as shown below. Here, we designed a GUI that allows users to trace logs to ensure transparency about what actions contributed to fairness when there is a change in the fairness time series value. Figure 4 displays a graph showing the time progression from the starting fairness score S1 to the current overall fairness score S2.

Furthermore, the display information generation unit 209 may display, in the object 506, the behavioral history of all users who contributed to the change in the fairness score for each buzz news article. Information related to the increase or decrease in the fairness score may be managed on the blockchain (e.g., Figures 9A and 9B), and the display information generation unit 209 may refer to and retrieve this information when creating the object 506. In this example, the display information generation unit 209 reads the fairness contribution management ledger 9001 and displays, in the object 506, a breakdown of the users who contributed to the increase or decrease in the fairness contribution.

The display information generation unit 209 uses the transaction ID recorded in the fairness management ledger 801 for each news article as a key to display information in the object 506, such as the transaction ID recorded in the fairness management ledger 801 for each news article as shown in Figure 8A, and presents which user's behavior was evaluated to increase or decrease the fairness.

The fairness information UI 208 of this system accumulates threads over time, each consisting of a news article for each buzz news item and its analysis results (corresponding to objects 501-506). It is also advisable to provide an interface 507 that allows donations on a thread-by-thread basis. In Figure 4, for example, when the display information generation unit 209 receives an operation (e.g., a click) on the interface 507, it calculates the amount corresponding to the operation (e.g., multiplying the number of clicks, assuming 100 yen per click) and outputs this as the donation amount (S311). The display information generation unit 209 then converts the donation amount into a fairness score at a predetermined rate. There are several possible distribution methods. Since donations themselves do not directly contribute to improving the fairness of news articles, one possible method is to manage the amount of tokens held by the user, as shown in Figure 9, and exchange the donation for tokens in a lump sum upon a user's exchange request. This role is performed by the fairness reward distribution unit 210. On the other hand, there is also a method of distributing the fairness score to information providers by recording it in the fairness increase/decrease amount of the transaction data described above. By managing the fairness score as a token circulating on the blockchain, it becomes possible to provide a reward distribution function that is resistant to tampering. In Figure 10, user "S.N." identified by user ID "U0000003" viewed a BuzzNews article identified by news ID "N0000003" on May 22, 2021, and made a donation amount equivalent to the fairness increase/decrease amount of "+10", which the display information generation unit 209 outputs as transaction data identified by transaction ID "T0000001".

Similar to the donation interface 507, the display information generation unit 209 may attach an interface 508 to each article thread that provides information contributing to improving fairness. The fairness information provision interface 508 can be used to provide primary source information referenced by the accuracy determination unit 205, collect user opinions that serve as material for the sentiment analysis unit 203, and receive information from experts for the fact-checking unit 207. In FIG. 4, for example, when the display information generation unit 209 receives an operation on the interface 508 (e.g., clicking the information provision icon, attaching the provided primary source, and sending), it calculates an amount corresponding to the operation (e.g., an amount corresponding to the number of documents, such as 1,000 yen per government-approved document) and outputs it as the information provision amount. As with the interface 507, the display information generation unit 209 converts the information provision amount into a fairness score at a predetermined rate and distributes it to the information provider by recording it in the fairness increase/decrease amount of the transaction data described above. In this case, as with interface 507, fairness scores are managed as tokens circulating on the blockchain, making it possible to provide a reward distribution function that is resistant to tampering. In this case, the fairness reward distribution unit 210 awards rewards based on fairness scores that reflect the fairness increase/decrease amount that increases/decreases for each token identified by the transaction ID.

Furthermore, for each article thread, it is advisable to provide a backlog interface 509 that calls up historical information, allowing the user to view the backlog when they wish to refer to past information beyond the information displayed. By selecting the backlog interface 509, the user can refer to the display information of the thread at a certain point in the past. For example, when the display information generation unit 209 receives an operation on the backlog interface 509 (for example, clicking the backlog icon and inputting a date to identify the thread they wish to refer to), it may extract threads containing that date from the display DB 3003 and display them in chronological order on the screen. If a certain keyword (for example, "safety of the coronavirus vaccine") is entered on the screen, threads containing that keyword and the date mentioned above will be displayed on the screen.

The display information generation unit 209 reads related news IDs and tag information obtained during the analysis of news articles and comments from the article DB 3001 shown in FIG. 11, and extracts news articles that include the above keywords, for example. The article DB 3001 is stored by the sentiment analysis unit 203.

Figure 11 is a diagram showing an example of article DB 3001. As shown in Figure 11, article DB 3001 includes a news ID for identifying a buzz news article, the headline of the news article identified by the news ID, a related article ID for identifying an article related to the news article, tag information attached to the news article, the date and time the news article was published, the date the buzz about the news article occurred, the name of the reporter who reported the news article, and the institution to which the reporter belongs. For example, Figure 11 shows that the headline of the news article identified by news ID "N0000001" is "After vaccination...", and the related article identified by related article ID "RN000002" is a news article searched by the related article collection unit 204. The tag information attached to the news article is "vaccine, politics, medical care, corona", and the news article was reported by reporter D. of Company A on May 14, 2021. This article was created by U, and indicates that the buzz occurred the following day, May 15, 2021. Furthermore, the website of the original news article for the buzz news in question is "http://www.xxx.co.jp/yyy.html", and the references collected as primary sources are Reference [1] ○○ and Reference [2] △△, which are published at "http://www.zzz.org/20210514.html".

In this way, the data that forms the basis for calculating the fairness score calculated by this system is displayed on the fairness information UI 208. The fairness information UI 208 may be provided as an internet site or a mobile device app (S310).

Just as heated arguments are prominent on social media and news comment sites, when people come across information that they subjectively judge to be convenient or correct, they tend to strongly perceive that information as accurate and to feel a strong resentment towards those who hold different opinions. Since providing unbiased information is difficult in the form of news distribution, one of the requirements is that the information be organized in a format that allows for a clear comparison with different opinions. Therefore, it is desirable to be able to check, on a single screen, the results of a diagnosis of the accuracy of a news article, feedback such as evaluations and comments from trusted experts who fact-checked the news article, and feedback from the public obtained through opinion clusters, etc., as in this system.

Sites widely used as reference sources for words disclose information and information sources from a variety of perspectives, based on the idea that "everyone's opinions are likely to be correct." Information can be considered "probably correct" if it is aggregated, diverse, independent, and decentralized. The news-based reference source provided by this system is similar to the widely used sites mentioned above in that it relies on users to gather evidence, displays reference information (verification), keeps a revision history, and is funded by donations. Differences include the ability to analyze and display user responses in near real time, track changes in public opinion over time, display a news fairness score, and visualize which actions contribute to improved fairness. In this system, for example, as shown in Figure 4, public opinion data (e.g., opinion clusters) are corrected and updated over time, ensuring transparency of the sender's position. The delivery of accurate and fair news is beneficial to users.

The likelihood that reading this article will provide fair information overall is called the overall fairness score. The display information generation unit 209 displays the fairness score of the original news article in the buzz news and the improved fairness in the process of compiling the article as an additional fairness contribution ("+15" in Figure 4). The display information generation unit 209 estimates the above-mentioned overall fairness score at this point in time using the fairness score transition and additional fairness contribution indicated by object 505. A breakdown of how fairness has improved may also be disclosed. The additional fairness contribution corresponds to the fairness score added by the accuracy determination unit 205 and the fairness determination unit 206. As already explained, the fairness score added in the fact check is attributable to the expert, and this expert is given a fairness score.

The hardware configuration for realizing this system is shown in Figure 11. The CPU 413 starts the program for this system and executes the processing of the functional blocks in Figure 3. Data used for calculations is temporarily stored in memory 415. When high-speed calculations are required, data can be transferred to the memory of the GPU 414, which can support the processing performed by each functional block. Display information generated by this system is displayed on the display device 411, allowing the system operator to preview it. Databases such as the article DB 3001 that stores news articles can be stored on the storage device 412. This system can be configured using multiple computers via the communication interface 417, and the actual calculation processing can be executed on an instance on the cloud.

This server may be provided as an on-premise server or implemented on a public cloud. When implemented on a public cloud, the CPU 413, GPU 414, and memory 415 may be dynamically allocated and configured to scale according to the scale of calculations. The system (fairness management unit) may be implemented on the same hardware as the system (backend and frontend), or on separate hardware. Here, a split implementation is assumed, and is represented by adding the symbol -A to the end.

The fairness reward distribution unit 210 preferably realizes its functions using the fairness score management unit 211 and display devices 411-A to 417-A. In particular, communication between the system (backend and frontend) and the system (fairness management unit) occurs via communication I/Fs 417 and 417-A, forming a blockchain network N for exchanging data blocks. A fairness management ledger for each news article and a fairness contribution management ledger for each user are stored in the storage device 421-A of the system (fairness management unit), and updates are made via the blockchain by approving transactions related to the increase or decrease in fairness contribution values output to the blockchain by the fairness judgment unit in the system (backend and frontend).

The news viewing website GUI is transferred from the server to the user via communication I/F 417. The user's terminal also has a similar hardware configuration. Any terminal with a similar hardware configuration that has web browsing functionality, such as a PC, smartphone, or game console, can be used as a network participant terminal for this system.

With this, we can provide a system where users can receive rewards in the form of tokens if they contribute to the fairness of news articles. When users post comments on social media or news sites, the system collects information and processes it to ensure fair information.

2. Second Embodiment In the first embodiment, a case where the system actively retrieves data is described. In the second embodiment, a case where the user provides information or notifies the system of information related to the user's reaction, and data is collected passively from the system's perspective, is described.

Specifically, as explained in the first embodiment regarding operations on the interface 508, the fairness information UI 208 enables information provision through the information provision interface 508. The processing in this case is explained below.

For example, the accuracy determination unit 205 requires information from primary sources to ensure the accuracy of the information in the original news article of buzz news. As shown in FIG. 13, after receiving information from the user via the fairness information UI 208 (S1301), the accuracy determination unit 205 determines whether useful information has been obtained regarding the primary source (S1302).

The determination of whether or not information is useful as a primary source may be performed automatically by the system or manually. For example, the accuracy determination unit 205 determines whether or not the primary source being determined is a government-approved reference, and if it determines that the primary source is a government-approved reference, it analyzes the primary source in the same manner as in Figure 5 (S1303).

As a result of analyzing the primary source, the accuracy determination unit 205 determines whether the content of the news article has been altered (S1304). If the accuracy determination unit 205 determines that the content of the news article has not been altered (S1304; No), it increases the fairness score of the news article (S1305). If the accuracy determination unit 205 determines that the content of the news article has not been altered (S1304; Yes), it decreases the fairness score of the news article (S1306). In response, the accuracy determination unit 205 recalculates and updates the reporter's fairness score (S1307, S1308), and also assigns a fairness score to the information provider (S1309).

The original data used to generate opinion clusters, which serves as the source data for the fairness determination unit 206, can also be input directly from the information provision interface 508. In this system, fairness is defined as presenting a variety of opinions and visualizing their position within them, so even if a large number of extreme opinions are posted, there will be no impact.

If the fairness score of an article changes as a result of this processing, the fairness reward distribution unit 210 will award fairness contribution points to users who played a role in forming the opinion cluster. By distributing fairness contribution point rewards within the same opinion cluster, the system can be made robust against opinion manipulation through multiple posts by a few users.

Furthermore, while we have explained how the system of the present invention selects experts for the fact-checking unit 207 from an expert database or a fairness contribution ledger for each user, it is also possible for experts not selected by the system to provide fact-checking information directly using the information provision interface 508. However, self-nominated fact-checking is permitted only if they have a proven track record in the news field in question. A proven track record is assumed to be one in which they have been certified by a third-party organization, such as having previously published a peer-reviewed paper recognized by a prestigious academic society. As described above, it is advisable to provide a system in which information posted by users can supplement information that the system has not been able to collect, thereby improving the fairness of news. The present system has been described above using the drawings, but as explained using Figures 1-3, the fair news distribution system 1000 is a news distribution system that distributes predetermined news by having a computer having a processor and memory execute a program, and includes a buzz news detection unit 201 that monitors news sites on a network and detects buzz news in which quotes or comments on news articles published on the news site satisfy a predetermined condition (for example, whether the number of times the news article is cited or the number of comments is equal to or exceeds a predetermined threshold), a sentiment information collection unit 202 that collects transaction data including comments on the buzz news and cumulative values of the comments that are posted on the network in association with the buzz news, and a sentiment information collection unit 203 that analyzes the comments included in the transaction data, and if the number of clusters obtained by clustering the opinions included in the comments satisfies a predetermined condition (for example, the number of clusters is equal to or exceeds a predetermined threshold), the news article to which the comment is attached is a news article with high fairness. a related article collection unit 204 that collects related articles including primary sources of the news article before it is edited by the media from the network; an accuracy determination unit 205 that determines the accuracy of the news article based on the news article determined by the emotion analysis unit and the related articles collected by the related article collection unit; and a fairness score that quantifies the fairness of the news article determined by the emotion analysis unit and the accuracy of the news article determined by the accuracy determination unit. The system includes a fairness determination unit 206 that outputs a fairness score, a fairness management ledger 801 that associates the fairness scores with the news articles, a fairness contribution management ledger 9101 that associates the users with fairness contributions, which are fairness scores given to users when they perform actions that contribute to an increase or decrease in the fairness score, and a fairness reward distribution unit 210 that uses the fairness management ledger and the fairness contribution management ledger to reward users who perform actions that increase the fairness of the news articles. This ensures transparency in the position of news article reporters and other senders, making it possible to deliver accurate and fair news.

Furthermore, as explained in Figure 2, S301, etc., the buzz news detection unit detects the buzz news when the number of citations or comments on the news article reaches a predetermined threshold as the predetermined condition, and therefore fairness can be determined based on the number of citations or comments on the news article.

Furthermore, as explained in Figure 5, S506, etc., the accuracy determination unit compares the information from the primary source with the gist of the news article, and if the content has not been altered, adds the fairness score. This makes it possible to confirm that accurate news articles are being published without compromising the gist of the original news article.

Furthermore, as explained in Figure 5, S502, etc., the accuracy determination unit adds a fairness score to the news article if the news article contains a description disclosing a primary source and the primary source can actually be accessed and viewed. This ensures the objectivity of the news article, that it is based on a primary source, and makes it possible to calculate a fairness score that takes this objectivity into account.

In addition, as explained in Figure 5, S509, etc., the system includes a fact-checking unit 207 that requests a specified expert to fact-check the news article, and if the result of the fact-check is a positive evaluation of the content of the news article, adds a fairness score to the news article. This makes it possible to ensure that the content of the news article is fair.

Furthermore, as explained using Figure 7 etc., the fairness determination unit calculates a fairness score by adding a higher fairness score the greater the number of opinion clusters, so it is possible to confirm that the news article is fair and takes into account a wider variety of opinions.

In addition, as explained using Figure 4 etc., the system includes a display information generation unit 209 that calculates the current overall fairness score based on the raw score of the news article and the increase or decrease in the fairness contribution rate, and displays the calculated overall fairness score on a computer screen, allowing users of the system to easily check the current fairness score for a news article.

Furthermore, as explained using Figure 4 etc., the display information generation unit displays on the screen a graph showing the time progression from the raw score of the news article to the overall fairness score, allowing users of the system to easily check the time progression of the fairness score.

Furthermore, as explained using Figure 4 etc., the display information generation unit reads the fairness contribution management ledger and displays on the screen a breakdown of users who contributed to an increase or decrease in the fairness contribution, allowing users of the system to easily check which users contributed to an increase or decrease in the fairness contribution.

As described with reference to FIG. 4 and the like, the display information generation unit displays an interface on the screen that allows a user to view information that contributes to improving fairness,
Furthermore, as explained using Figure 4, etc., the accuracy determination unit determines the accuracy of the news article using the primary source obtained via the interface, so that accuracy can be determined using the primary source provided by the user.

Furthermore, as explained using Figure 10 etc., the fairness score management unit accepts processes related to the increase or decrease of the fairness contribution from users via the network (e.g., provision of primary sources, donations, fact-checking), associates the user with the news article that is the subject of the fairness contribution, and the amount of the increase or decrease, and manages them as tokens exchanged on the blockchain. This allows fairness contributions to be issued using tokens circulating on the blockchain.

Furthermore, as explained using Figure 10 etc., the token has a transaction ID defined for each process related to the increase or decrease, and the fairness score management unit manages the increase or decrease in the fairness score for each transaction ID, so that the increase or decrease in the fairness score can be managed each time a user performs an action such as providing a primary source, making a donation, or fact-checking.

Furthermore, as explained using Figure 10 etc., the fairness reward distribution unit awards the reward based on the fairness score, which increases or decreases for each token identified by the transaction ID, so that it is possible to award a reward according to the fairness score, which increases or decreases each time the user performs the action.

Furthermore, as explained using FIG. 10 etc., the reward distribution unit may manage the tokens in a format that allows exchange for other currencies, based on the amount of increase or decrease in the fairness contribution. This allows the fairness contribution to be reflected as the value of other currencies.

In this way, this system can provide an ecosystem for the distribution of fair news. Furthermore, because this system uses a blockchain framework, it is possible to track the ownership and rights of the data used in news articles, ensuring reliability and transparency. Furthermore, the rights themselves can be bought and sold through a marketplace.

The present invention is not limited to the above-described embodiments, and in the implementation stage, the components may be modified and embodied within the scope of the gist of the invention, or multiple components disclosed in the above-described embodiments may be combined as appropriate.

201 Buzz news detection unit 202 Sentiment information collection unit 203 Sentiment analysis unit 204 Related article collection unit 205 Accuracy determination unit 206 Fairness determination unit 207 Fact check unit 208 Fairness information user interface (UI)
209 Display information generation unit 210 Fairness reward distribution unit 211 Fairness score management unit 301 Buzz news detection 302 Reaction collection 303 Sentiment analysis processing 304 News-related data collection 305 Fact check request 306 Submission of correction article 307 Fact check result 308 Article accuracy/fairness judgment processing 309 Reference information source creation/update 310 Reference information source site provision 311 Donation to reference information source 312 Token distribution 313 User fairness contribution update processing 411 Display device 412 Storage device 413 CPU
414 GPUs
415 Memory 416 User interface 417 Communication interface 501 Verification information 502 Sentiment analysis information 503 Raw fairness score of original article 504 Progress of fact check 505 Time series change in fairness score 506 User behavior history that influenced the fairness score 507 Donation reception interface 508 Information reception interface 509 Backlog reference interface

Claims (15)

A news distribution system in which a computer executes a program to distribute predetermined news,
a buzz news detection unit that monitors news sites on the network and detects buzz news in which quotes or comments on news articles published on the news sites satisfy predetermined conditions;
a sentiment information collecting unit that collects transaction data including comments on the buzz news posted on the network in association with the buzz news and cumulative values of the comments;
a sentiment analysis unit that analyzes the comments included in the transaction data, and determines that the news article to which the comment is attached is a news article with high fairness if the number of clusters obtained by clustering the opinions included in the comments satisfies a predetermined condition;
a related article collection unit that collects related articles, including primary sources published by governments or companies, about the news article from the network;
an accuracy determination unit that determines accuracy of the news article based on the news article determined by the sentiment analysis unit and the related articles collected by the related article collection unit;
a fairness determination unit that calculates a fairness score by quantifying the fairness of the news article determined by the sentiment analysis unit and the accuracy of the news article determined by the accuracy determination unit;
a fairness score management unit that manages a fairness management ledger that associates the fairness scores with the news articles, and a fairness contribution management ledger that associates the users with fairness contributions, which are fairness scores that are assigned to users when the users engage in evaluation behavior that is defined as behavior that contributes to an increase or decrease in the fairness score, and that compensates for deficiencies in the requirements of not being altered from the primary source and reducing bias by reaching a conclusion after considering arguments from multiple perspectives, or behavior that enhances the requirements to a certain extent;
a fairness reward distribution unit that uses the fairness management ledger and the fairness contribution management ledger to reward users who have taken actions that increase fairness for the news article;
A news distribution system comprising: 2. The news distribution system according to claim 1,
the buzz news detection unit detects the buzz news when the number of citations or the number of comments on the news article reaches a predetermined threshold as the predetermined condition;
A news distribution system characterized by: 2. The news distribution system according to claim 1,
The accuracy determination unit compares the information of the primary source with the gist of the news article, and if the content has not been altered, adds the fairness score.
A news distribution system characterized by: 2. The news distribution system according to claim 1,
The accuracy determination unit adds a fairness score to the news article if the news article contains a description disclosing a primary source and the primary source is actually accessible and viewable.
A news distribution system characterized by: 2. The news distribution system according to claim 1,
a fact-checking unit that requests a predetermined expert to fact-check the news article, and adds a fairness score to the news article when the result of the fact-check is a positive evaluation of the content of the news article;
A news distribution system comprising: 2. The news distribution system according to claim 1,
the fairness determination unit calculates a fairness score by adding a higher fairness score as the number of opinion clusters increases;
A news distribution system characterized by: 2. The news distribution system according to claim 1,
a display information generation unit that calculates a current overall fairness score based on the raw score of the news article and the increase or decrease in the fairness contribution rate, and displays the calculated overall fairness score on a computer screen;
A news distribution system comprising: 8. The news distribution system according to claim 7,
the display information generation unit displays on the screen a graph showing a time progression from the raw score of the news article to the overall fairness score.
A news distribution system characterized by: 8. The news distribution system according to claim 7,
the display information generation unit reads the fairness contribution management ledger and displays on the screen a breakdown of users who contributed to an increase or decrease in the fairness contribution rate.
A news distribution system characterized by: 8. The news distribution system according to claim 7 ,
the display information generation unit displays, on the screen, an interface that provides information that contributes to improving fairness from a user;
the accuracy determination unit determines the accuracy of the news article using the primary source obtained via the interface;
A news distribution system characterized by: 2. The news distribution system according to claim 1,
The fairness score management unit receives processing related to an increase or decrease in the fairness contribution rate from a user via the network, associates the user with the news article that is the subject of the fairness contribution rate, and the amount of the increase or decrease, and manages them as tokens exchanged on a blockchain.
A news distribution system characterized by: The news distribution system according to claim 11,
The token has a transaction ID defined for each process related to the increase or decrease, and the fairness score management unit manages the increase or decrease of the fairness score for each transaction ID.
A news distribution system characterized by: The news distribution system according to claim 12 ,
the fairness reward distribution unit awards the reward based on the fairness score, which increases or decreases for each of the tokens identified by the transaction ID;
A news distribution system characterized by: The news distribution system according to claim 13,
The fairness reward distribution unit manages the tokens in a form that allows exchange with other currencies based on the amount of increase or decrease in the fairness contribution rate.
A news distribution system characterized by: A news distribution method carried out in a news distribution system that distributes predetermined news using a computer having a processor and a memory, comprising:
The system monitors news sites on the network and detects buzz news in which quotes or comments on news articles published on the news sites satisfy predetermined conditions;
Collecting transaction data including comments on the buzz news posted on the network in association with the buzz news and cumulative values of the comments;
analyzing the comments included in the transaction data, and if the number of clusters obtained by clustering the opinions included in the comments satisfies a predetermined condition, determining that the news article to which the comment is attached is a news article with high fairness;
Collecting related articles, including primary sources published by governments or companies, about the news article from the network;
determining accuracy of the news article based on the determined news article and the collected related articles;
calculating a fairness score that quantifies the determined fairness of the news article and the determined accuracy of the news article;
A fairness management ledger that associates the fairness scores with the news articles, and a fairness contribution management ledger that associates users with fairness contributions, which are fairness scores given to users when the user falls under an evaluation behavior that is defined as behavior that contributes to an increase or decrease in the fairness score, and that compensates for deficiencies in the requirements that the news article has not been altered from the primary source and that bias is reduced by reaching a conclusion after considering arguments from multiple perspectives, or behavior that enhances the requirements to a certain extent, and that contributes to an increase or decrease in the fairness score, are managed;
using the fairness management ledger and the fairness contribution management ledger to reward users who have performed actions that increase the fairness of the news article;
A news distribution method comprising: