JP7287992B2 - Information processing device, information processing system, information processing method, and program - Google Patents

Description

The present invention relates to an information processing device, an information processing system, an information processing method, and a program.

In recent years, services have been provided in which news distributors collect news and distribute the collected news. News collected by news distributors may include fake news, so-called fake news. In order to prevent the transmission of fake news, conventionally, there is a technique of learning the characteristics of manually discovered fake news and automatically detecting fake news based on the learning result (see, for example, Patent Document 1).

Japanese Unexamined Patent Application Publication No. 2020-113179

However, it is not easy to spot fake news from a large number of news. Therefore, the news distributed by news distributors includes various news, from low-credit news to high-credit news. There is a problem that it is difficult for users who have received news delivered from such various news distributors to uniformly trust the content of the news even if they see the content of the news.

The present invention has been made in consideration of such circumstances, and provides an information processing apparatus, an information processing system, an information processing method, and a program that enable users to trust the content of news that is distributed. is one of the purposes.

According to one aspect of the present invention, a collection unit that collects texts of news and titles attached to the news, and an analysis unit that clusters the texts and the titles based on words included in the texts and the titles. and an analysis unit that obtains a degree of matching between the text and the title based on the result of the clustering.

ADVANTAGE OF THE INVENTION According to one aspect of the present invention, users can be made to trust the content of news to be distributed.

2 is a diagram showing an example of transmission of information on news handled by an information processing apparatus 100 in an information processing system M. FIG. 2 is a diagram showing an example of the configuration of a news distribution server 2; FIG. It is an example of the top page G1 displayed on the terminal device 10 . 3 is a diagram showing an example of a news page G2 displayed on the terminal device 10; FIG. 1 is a diagram illustrating an example of a configuration of an information processing device 100; FIG. FIG. 6 is a diagram conceptually showing the functions of the trained model 151. As shown in FIG. 4 is a sequence diagram showing an example of the flow of processing executed by the information processing system M; FIG.

Hereinafter, embodiments of an information processing apparatus, an information processing system, an information processing method, and a program according to the present invention will be described with reference to the drawings.

[overview]
FIG. 1 is a diagram showing an example of transmission of information on news handled by an information processing apparatus 100 in an information processing system M. As shown in FIG. The information processing system M includes, for example, one or more poster terminals 1 , news distribution servers 2 , terminal devices 10 , and information processing devices 100 . The information processing device 100 is connected to a plurality of poster terminals 1, a news distribution server 2, and a plurality of terminal devices 10 via a network. Networks include WANs (Wide Area Networks), LANs (Local Area Networks), the Internet, cellular networks, and the like.

A contributor uses contributor terminal 1 to create news about various events occurring in the world, and posts the created news to news distribution server 2 . A contributor who posts news using contributor terminal 1 may be, for example, a legal entity that posts news, such as so-called media, or a natural person other than a legal entity. The news distribution server 2 distributes news posted by the poster terminal 1 to the terminal device 10 . The news distribution server 2 is an example of a news distribution device.

The terminal device 10 displays, for example, news distributed by the news distribution server 2 via the network. The user who occupies the terminal device 10 can know the content of the news posted by the contributor by displaying the news on the terminal device 10 .

The information processing device 100 acquires news distributed to the news distribution server 2 . The information processing apparatus 100 derives information indicating the degree of matching between the title and text of the acquired news, and determines the degree of reliability that the acquired news is fake news. The information processing device 100 transmits information on the derived degree of matching and the determined degree of reliability to the news distribution server 2 .

[News distribution server 2]
The news distribution server 2 is implemented by, for example, one or more processors. FIG. 2 is a diagram showing an example of the configuration of the news distribution server 2. As shown in FIG. The news distribution server 2 includes, for example, a communication section 210 and a processing section 220 . The communication unit 210 includes, for example, a communication interface such as NIC. The communication unit 210 communicates with the terminal device 10, the information processing device 100, and the like via a network.

The processing unit 220 includes, for example, a collection unit 221, a determination unit 222, and a distribution unit 223. Each component of the processing unit 220 is realized by executing a program (software) by a hardware processor such as a CPU (Central Processing Unit). Some or all of these components are hardware (circuit part; circuitry) or by cooperation of software and hardware. The program may be stored in advance in a storage device (a storage device with a non-transitory storage medium) such as a HDD (Hard Disk Drive) or flash memory, or may be stored in a detachable storage device such as a DVD or CD-ROM. It may be stored in a storage medium (non-transitory storage medium) and installed in the storage device by mounting the storage medium in the drive device. All the above components need not be implemented by a single processor, and distributed processing by a plurality of processors may be performed as appropriate.

The collection unit 221 collects multiple pieces of news posted by multiple poster terminals 1 . The collection unit 221 collectively notifies the determination unit 222 of the collected news. The determination unit 222 selects the content of the news notified by the collection unit 221, and classifies a plurality of news according to genre.

The determination unit 222 transmits a plurality of pieces of news classified by the collection unit 221 to the information processing device 100 . The determining unit 222 may transmit the news before being classified by the collecting unit 221 to the information processing device 100 . The information processing device 100 derives the matching degree of the news transmitted by the news distribution server 2 , determines the degree of reliability, and transmits it to the news distribution server 2 . The news distribution server 2 determines the priority of the classified news based on the reliability information transmitted by the information processing device 100 .

The determining unit 222 adds meta information indicating the classified genre to the news. The determining unit 222 gives the degree of matching transmitted by the information processing device to the news. The determining unit 222 notifies the distributing unit 223 of the news information together with the assigned priority, meta information, and degree of matching transmitted by the information processing apparatus 100 . The distribution unit 223 distributes the news information notified by the determination unit 222 to the terminal device 10 .

[Terminal device 10]
The terminal device 10 is, for example, a smart phone, and includes a touch panel that serves both as a display section for displaying images and as an input section. The terminal device 10 may be any device that has at least a communication function and a display function, such as a mobile phone other than a smart phone, a tablet terminal, a notebook or desktop personal computer, or a PDA (Personal Digital Assistant).

In the terminal device 10, for example, a UA (User Agent) such as an application program or a browser operates. The terminal device 10 is distributed by the news distribution server 2 by, for example, operating a UA that displays news and transmitting a request to the news distribution server 2 . Then, the terminal device 10 can display the delivered news on the touch panel.

When the user of the terminal device 10 wants to receive news delivered by the news delivery server 2, the user activates (operates) the UA. The user of the terminal device 10 causes the terminal device 10 to transmit a news distribution request to the news distribution server 2 by operating the terminal device 10 in which the UA is operating.

The news distribution server 2 distributes a top page as news content to the terminal device 10 in response to a news distribution request transmitted from the terminal device 10 . The terminal device 10 displays the top page distributed by the news distribution server 2 on the touch panel. The top page is content that includes a list of front images including news titles and thumbnails.

FIG. 3 is an example of the top page G1 displayed on the terminal device 10. As shown in FIG. As illustrated, the terminal device 10 displays a genre G11 and a front image G12 on the top page G1. The genre G11 is displayed above the top page G1. The genre G11 includes the content of the genre given to the news as meta information. For the genre G11, genres such as "major", "entertainment", "sports", "domestic", "economy", and "international" are displayed in tabs. When one genre G11 is selected from a plurality of genres G11, the tab of the selected genre is darkened, and a plurality of news front images G12 to which the meta information of the selected genre is added are sequentially displayed.

The front image G12 includes a thumbnail G13, a title G14, and a matching degree G15. The title G14 included in the front image G12 is restricted in the number of characters. Therefore, the title G14 of the front image G12 may be a simplified version of the news title included in the body of the news. The front image G12 is displayed from the top in the order of the display priority assigned to the news (high ranking order).

When the front image G12 is operated, the terminal device 10 transmits to the news distribution server 2 a designation request corresponding to the news indicated by the front image G12. For example, when receiving a specified request transmitted from the terminal device 10 , the news delivery server 2 delivers the body of the news corresponding to the specified request to the terminal device 10 . The terminal device 10 to which the text of the news has been distributed by the news distribution server 2 displays the news image on the touch panel. The news distribution server 2 may distribute the text of the news to the terminal device 10 together with the top page G1.

FIG. 4 is a diagram showing an example of the news page G2 displayed on the terminal device 10. As shown in FIG. As shown, the news page G2 includes an enlarged image G21, a title G22, a matching degree G23, a front text G24, and a continue display button G25. The enlarged image G21 is, for example, an enlarged image of the thumbnail G13 included in the front image G12. Since a plurality of pieces of news are displayed in a narrow area in the front image G12, the thumbnail G13 is a small image. The news page G2 can have a large area for displaying images. Therefore, the enlarged image G21 included in the news page G2 can be an image larger than the thumbnail G13.

A title G22 is displayed below the enlarged image G21. The title G22 has, for example, the same content as the title G14 included in the front image G12. The title G14 included in the front image G12 may not be able to contain a long sentence when the display space is narrow. Therefore, the title G14 included in the front image G12 is a simplified version of the title G22.

A matching degree G23 is displayed between the enlarged image G21 and the title G22. The degree of matching G23 is the same degree of matching as displayed on the front page. A front text G24 is displayed below the title G22. The front text G24 is text indicating the first few lines of the news text. A display continuation button G25 is displayed below the front text G24. By operating the display continuation button G25, the text of the news is displayed. When the text of the news is long, the display of the terminal device 10 is scrolled to sequentially display the full text of the news. The title and enlarged image are also displayed together with the text of the news displayed here.

[Configuration of information processing apparatus 100]
The information processing device 100 is implemented by, for example, one or more processors. FIG. 5 is a diagram showing an example of the configuration of the information processing apparatus 100. As shown in FIG. The information processing device 100 includes, for example, a communication unit 110, a processing unit 120, and a storage unit 150. The communication unit 110 includes, for example, a communication interface such as NIC. The communication unit 110 communicates with the news distribution server 2 and the like via a network.

The processing unit 120 includes, for example, a collection unit 121, an analysis unit 122, a derivation unit 123, a generation unit 124, and a determination unit 125. Each component of the processing unit 120 is implemented by, for example, a hardware processor such as a CPU executing a program (software). Some or all of these components may be realized by hardware (including circuitry) such as LSI, ASIC, FPGA, GPU, etc., or by cooperation of software and hardware. good too. The program may be stored in advance in a storage device (a storage device with a non-transitory storage medium) such as a HDD (Hard Disk Drive) or flash memory, or may be stored in a detachable storage device such as a DVD or CD-ROM. It may be stored in a storage medium (non-transitory storage medium) and installed in the storage device by mounting the storage medium in the drive device. All the above components need not be implemented by a single processor, and distributed processing by a plurality of processors may be performed as appropriate.

Information processing apparatus 100 further includes storage unit 150 . The storage unit 150 is implemented by, for example, a storage device such as an HDD, flash memory, EEPROM (Electrically Erasable Programmable Read Only Memory), ROM (Read Only Memory), RAM (Random Access Memory). The storage unit 150 stores a learned model 151 in addition to various programs such as firmware and application programs. The storage unit 150 may be a NAS (Network Attached Storage) that can be accessed from the information processing apparatus 100 via a network or the like instead of a device attached to the information processing apparatus 100 .

The collection unit 121 acquires multiple pieces of news distributed by the news distribution server 2 . The collection unit 121 extracts the title and text of each piece of news collected. The collection unit 121 notifies the analysis unit 122 of the titles and texts of the collected news.

The analysis unit 122 performs, for example, natural language processing on each of the title and text of the news, and extracts words included in the title and text. The analysis unit 122 vectorizes and clusters the title and text based on the presence of words contained in the title and text to generate a title vector and text vector. The analysis unit 122 generates, for example, a title vector, which is a sparse vector in which vector elements corresponding to words included in the title are set to 1, and other vector elements are set to 0. Then, the analysis unit 122 clusters the titles based on the positions of the title vectors in the vector space.

The analysis unit 122 similarly generates a text vector, which is a sparse vector in which, for example, vector elements corresponding to words included in the text are set to 1, and vector elements that do not are set to zero. Then, the analysis unit 122 clusters the text based on the positions of the text vectors on the vector space. The analysis unit 122 notifies the derivation unit 123 of the title vector and text vector obtained as a result of clustering.

The analysis unit 122 may use any method for clustering. The analysis unit 122 uses, for example, a single link method (shortest distance method), a complete link method (longest distance method), a group average method, a Word method (minimum variance method), a centroid method (center of gravity method), a weighted average method, You may use well-known methods, such as the median method and the K mean method.

The derivation unit 123 derives the matching degree between the news title and the text based on the clustering result of the analysis unit 122 . The degree of matching is an index value representing the rate at which the title of the news matches the content of the text. For example, news with a title that accurately and concisely describes the content of the main text is derived with a high degree of matching, and although the title is exaggerated, the content of the main text is far from the title. In some cases, the degree of matching is derived low.

The derivation unit 123 calculates, for example, the cosine similarity between the title vector and the text vector generated by the analysis unit 122 as the matching degree between the news title and text. The derivation unit 123 derives the calculated cosine similarity as the matching degree between the news title and text. The degree of matching between news titles and text may be derived by other methods. The derivation unit 123 transmits the derived degree of matching to the news distribution server 2 .

The generation unit 124 generates the learned model 151 based on the news titles and texts extracted by the collection unit 121 . The generation unit 124 generates an optimized trained model by performing machine learning using, for example, news titles and texts extracted from the news collected by the collection unit 121 as learning data.

FIG. 6 is a diagram conceptually showing the functions of the trained model 151. As shown in FIG. The trained model 151 uses, for example, the title and text of news as input data, and output data as to whether or not the news is fake news. The trained model 151 has, for example, an input layer, an intermediate layer and an output layer.

The input layer receives news title and text data. The output layer outputs a result indicating whether the news is fake news. The intermediate layer has, for example, a multilayer neural network connecting the input layer and the output layer. As a trained model, for example, Transformer and Vert are used. As pre-processing for processing by the trained model, processing for narrowing down words by TF-IDF (Term Frequency-Inverse Document Frequency) or the like may be performed.

For example, the generation unit 124 treats data obtained when news is determined to be fake news by some method as correct data, and adjusts the parameters of the machine learning model on which the trained model 151 is based. The generation unit 124 uses input data to the input layer as learning data and data to be output from the output layer as correct data, and optimizes the machine learning model by adjusting the parameters of the intermediate layer by back propagation or the like. By doing so, a trained model 151 is generated. The generation unit 124 stores the generated learned model 151 in the storage unit 150 .

The determination unit 125 reads the learned model 151 from the storage unit 150, for example. The determination unit 125 inputs the news titles and texts extracted by the collection unit 121 to the learned model 151 as input data, thereby determining the reliability that the news collected by the collection unit 121 is fake news. Further, for example, the determination unit 125 determines the reliability to be higher as the degree of matching derived by the derivation unit 123 is lower. The determination unit 125 transmits the determined reliability to the news distribution server 2 .

[Processing sequence]
Next, a series of processes executed by the information processing system M will be described. FIG. 7 is a sequence diagram showing an example of the flow of processing executed by the information processing system M. As shown in FIG. First, the poster terminal 1 posts the created news to the news distribution server 2 (step S1). There are a plurality of poster terminals 1, and news is posted to the news distribution server 2 from the plurality of poster terminals 1. - 特許庁Each contributor terminal 1 posts a plurality of news items to the news distribution server 2 .

The news distribution server 2 collects the news posted by the contributor terminal 1 in the collection unit 221, and then the determination unit 222 classifies the news collected by the collection unit 221 (step S11). The news distribution server 2 transmits the news classified by the determination unit 222 to the information processing device 100 (step S2).

The information processing apparatus 100 to which the news has been transmitted collects the news in the collection unit 121 and extracts the title and text of the collected news. Subsequently, the analysis unit 122 clusters the news titles and texts extracted by the collection unit 121 (step S21) to generate title vectors and text vectors.

Subsequently, the derivation unit 123 derives the matching degree between the news title and the text using the title vector and text vector generated by the analysis unit 122 (step S22). Subsequently, the generating unit 124 generates an optimized trained model by performing machine learning using the news title and text as learning data (step S23).

Subsequently, the determination unit 125 reads the learned model 151 from the storage unit 150 . The determination unit 125 inputs the news titles and texts extracted by the collection unit 121 to the learned model 151 as input data. Thus, the determination unit 125 determines the reliability that the news collected by the collection unit 121 is fake news (step S24). After that, the information processing apparatus 100 transmits the degree of derivation derived by the derivation unit 123 and the degree of reliability determined by the determination unit 125 to the news distribution server 2 (step S3).

The determination unit 222 of the news distribution server 2 determines the priority of the news based on the reliability transmitted from the information processing device 100 (step S12). Subsequently, the news distribution server 2 distributes a plurality of news items to the terminal device 10 by adding meta information indicating the sorted genre together with the priority order and degree of matching between the title and the text (step S4).

The terminal device 10 that receives the delivered news displays the delivered news (step S31). When displaying news, the terminal device 10 classifies the news based on the meta information attached to the news. Furthermore, the terminal device 10 arranges and displays news from the top in descending order of the priority assigned to the news. Furthermore, the terminal device 10 displays the degree of matching given to the news together with the news. Thus, a series of processes executed by the information processing system M ends.

The information processing apparatus 100 according to the embodiment clusters news titles and texts to generate title vectors and text vectors, and derives the degree of matching between the titles and texts. It can be seen that when the degree of matching between the title and the text is low, there is a tendency for the news to be more likely to be fake news. Therefore, by deriving the matching degree between the title and the text and distributing the news according to the derived matching degree, it is possible to make the user trust the contents of the distributed news.

Further, the information processing apparatus 100 determines the reliability that the news is fake news based on the derived degree of matching, and the lower the degree of matching, the higher the reliability. You can clearly recognize the possibility of fake news. Furthermore, using a learned model that determines whether the news is fake news by machine learning using the text and title as learning data and the news when the news is judged to be fake news as correct data. , to determine the reliability. Therefore, the reliability of fake news can be determined easily and accurately.

Furthermore, the information processing system M displays the matching degree of the news when displaying the news on the terminal device 10 . Therefore, the user who occupies the terminal device 10 can roughly predict whether the text of the news is worth reading by checking the degree of matching.

[Modification]
In the above embodiment, the determination unit 222 determines the priority based on the reliability of the news being fake news. On the other hand, the determining unit 222 may determine the order of priority based on the degree of matching between the news title and text. In this case, the determination unit 222 may determine a higher priority as the degree of matching is higher. Furthermore, the determination unit 222 may determine the priority based on the degree of reliability that the news is fake news and the degree of matching between the title and text of the news.

In the above-described embodiment, the news distribution server 2 distributes the news to the terminal device 10 after giving the degree of matching to the news. On the other hand, the news distribution server 2 may distribute the news to the terminal device 10 by adding reliability to the news instead of or in addition to the degree of matching. In this case, the reliability that the news is fake news is displayed on the terminal device 10 together with the news. The news distribution server 2 may distribute news to the terminal device 10 to which the degree of matching and the degree of reliability are assigned.

In the above embodiment, the determination unit 125 uses the learned model 151 to determine the reliability that the news is fake news. On the other hand, the determination unit 125 may determine the reliability of news based on the degree of matching between the title and text of the news. The determination unit 125 may, for example, use the reciprocal of the degree of matching between the title and text of the news as the degree of reliability that the news is fake news.

In the above embodiment, the information processing device 100 includes the generating unit 124 and generates the trained model 151 in the information processing device 100 . On the other hand, the information processing apparatus 100 may not include the generation unit 124 and may not generate the trained model 151 . In this case, the information processing apparatus 100 acquires the separately generated learned model 151 and stores it in the storage unit 150, for example.

Fake news includes types such as propaganda, simple hoaxes, malicious information, misinformation, disinformation, and misinformation. The information processing apparatus 100 may derive the reliability of each of these types of fake news as the reliability of fake news.

As described above, the mode for carrying out the present invention has been described using the embodiments, but the present invention is not limited to such embodiments at all, and various modifications and replacements can be made without departing from the scope of the present invention. can be added.

1... contributor terminal 2... news distribution server 10... terminal device 100... information processing apparatus 110, 210... communication units 120, 220... processing units 121, 221... collection unit 122... analysis unit 123... derivation unit 124... generation unit 125 ... determination section 150 ... storage section 151 ... learned model 222 ... determination section 223 ... delivery section G1 ... top page G2 ... news page G11 ... genre G12 ... front image G13 ... thumbnail G14 ... title G15 ... matching degree G21 ... enlarged image G22 ... Title G23 ... Matching degree G24 ... Front text G25 ... Display continuation button M ... Information processing system

Claims (14)

a collection unit that collects texts of news provided via a network and titles given to the news;
an analysis unit that clusters the text and the title based on words contained in the text and the title, respectively;
a derivation unit that derives a degree of matching between the text and the title to be displayed on a user terminal together with the text and the title, based on the clustering result;
Information processing equipment. A determination unit that determines the reliability that the news is fake news based on the degree of matching,
The determination unit determines the reliability to be larger as the degree of matching is lower.
The information processing device according to claim 1 . The judging unit judges whether the news is fake news by machine learning using the title and the text as learning data and correct data when the news is judged to be fake news. Determining the reliability using a trained model that
The information processing apparatus according to claim 2. A trained model for determining whether or not the news is fake news by machine learning using the title and the text as learning data and the news when the news is determined to be fake news as correct data. a generator that generates
The information processing apparatus according to claim 2 or 3. an information processing apparatus according to any one of claims 1 to 4;
a news distribution device that collects news provided via a network, classifies the collected news, and distributes it to a terminal device;
The derivation unit provides the degree of matching to the news distribution device,
The news distribution device causes the terminal of the user to display the degree of matching together with the title and the text.
Information processing system. The news distribution device includes a determination unit that determines the priority of the classified news,
The determination unit determines the priority based on the degree of matching provided by the derivation unit.
The information processing system according to claim 5. The determination unit determines the priority higher as the degree of matching is higher.
The information processing system according to claim 6. The news distribution device distributes the news to which the degree of matching is attached.
The information processing system according to any one of claims 5 to 7. an information processing apparatus according to any one of claims 2 to 4;
a news distribution device that collects news provided via a network, classifies the collected news, and distributes it to a terminal device;
The determination unit provides the reliability to the news distribution device,
Information processing system. The news distribution device includes a determination unit that determines the priority of the classified news,
The determination unit determines the priority based on the reliability provided by the information processing device.
The information processing system according to claim 9. The determination unit determines the priority to be lower as the reliability is higher.
The information processing system according to claim 10. The news distribution device distributes the news with the reliability attached.
The information processing system according to any one of claims 9 to 11. the computer
collecting the text of the news provided via the network and the title attached to the news,
clustering the text and the title based on words contained in the text and the title, respectively;
Based on the results of the clustering, derive the degree of matching between the text and the title to be displayed on the user's terminal together with the text and the title;
Information processing methods. to the computer,
collecting texts of news provided via a network and titles given to the news;
Clustering the text and the title based on words contained in the text and the title, respectively;
Based on the result of the clustering, derive the degree of matching between the text and the title to be displayed together with the text and the title on the user's terminal ;
program.

Images