Information distribution system, information distribution method and program

Abstract

To distribute summary information optimized for each user.SOLUTION: The present invention relates to an information distribution system that distributes summary information to one or more users, and the information distribution system has: a summary information acquisition part which acquires the summary information including a summary and a classification of acquired information using a pre-trained large-scale language model; distribution part which distributes the summary information to the users; and a recommendation part which updates the users to which the summary information is distributed by the distribution part according to reaction information on the users to the distributed summary information.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to an information distribution system, an information distribution method, and a program. [Background technology]

[0002] For example, there is a service known in the art that allows users to specify in advance keywords about which they want to collect information, and then notifies them by email when information containing those keywords is made public on the Internet or elsewhere (see, for example, non-patent document 1). [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] "Wikipedia", [online], [searched June 5, 2023], Internet<https: / / ja.wikipedia.org / wiki / Google_%E3%82%A2%E3%83%A9%E3%83%BC%E3%83%88> Summary of the Invention [Problem to be solved by the invention]

[0004] However, in order to notify users of only the information they are interested in from the information published on various platforms such as news, the web, blogs, etc., it was necessary to specify appropriate keywords. For example, in the case of keyword matching, there were cases where a lot of information that the user is not interested in (noise) was included, or information that the user is interested in was often omitted.

[0005] An object of the present disclosure is to provide an information distribution system, an information distribution method, and a program for distributing summary information optimized for each user. [Means for solving the problem]

[0006] The present disclosure has the following configuration.

[0007] [1] An information delivery system for delivering summary information to one or more users, comprising: a summary information acquisition unit that acquires the summary information including the main points and classification of the acquired information by using a trained large-scale language model; a distribution unit that distributes the summary information to the user; a recommendation unit that updates the users to whom the summary information is to be distributed by the distribution unit based on reaction information of the users to the distributed summary information; An information distribution system having the above configuration.

[0008] [2] The summary information acquisition unit inputs the acquired information to the large-scale language model together with a command statement specifying a condition for the large-scale language model to generate the summary information, and acquires the summary information including a gist and a classification of the acquired information. An information distribution system according to [1],

[0009] [3] The distribution unit provides the recommendation unit with an operation of the user with respect to the distributed summary information as the reaction information. An information distribution system according to [1] or [2], characterized in that:

[0010] [4] The user's operation is an operation showing a favorable reaction to the distributed summary information or an operation of adding a comment to the distributed summary information. An information distribution system as described in [3], characterized by:

[0011] [5] The recommendation unit updates the users to whom the summary information is to be distributed by the distribution unit based on the reaction information collected by the group to which the users belong. An information distribution system according to any one of claims [1] to [4],

[0012] [6] An information distribution method for an information distribution system that distributes summary information to one or more users, comprising: a summary information acquisition step of acquiring the summary information including the gist and classification of the acquired information by using a trained large-scale language model; a delivery step of delivering the summary information to the user; a recommendation step of updating the users to whom the summary information is to be delivered in the delivery step based on reaction information of the users to the delivered summary information; An information distribution method having the above structure.

[0013] [7] A computer that distributes summary information to one or more users, a summary information acquisition step of acquiring the summary information including the gist and classification of the acquired information by using a trained large-scale language model; a delivery step of delivering the summary information to the user; a recommendation step of updating the users to whom the summary information is to be delivered in the delivery step based on reaction information of the users to the delivered summary information; A program for executing the above. Effect of the Invention

[0014] According to the present disclosure, summary information optimized for each user can be delivered. [Brief description of the drawings]

[0015] [Figure 1] 1 is a configuration diagram of an example of an information distribution system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram illustrating an example of a hardware configuration of a computer according to the present embodiment. [Diagram 3] 1 is a functional configuration diagram of an example of an information distribution system 1 according to an embodiment of the present invention. [Figure 4] FIG. 4 is an explanatory diagram of an example of a command statement according to the embodiment. [Diagram 5] 11 is an explanatory diagram of an example of a condition to be added to a command statement according to the embodiment; [Figure 6] 10 is a flowchart illustrating an example of a process of the information distribution system according to the embodiment. [Figure 7] FIG. 11 is an explanatory diagram of an example of summary information. [Figure 8] FIG. 11 is an explanatory diagram of an example of summary information. [Figure 9] 13 is a flowchart illustrating an example of a process in step S14. [Figure 10] 13 is a flowchart showing an example of the process of steps S18 to S20. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] Next, an embodiment of the present invention will be described in detail, but the present invention is not limited to the following embodiment.

[0017] <System configuration> Fig. 1 is a configuration diagram of an example of an information distribution system 1 according to this embodiment. The information distribution system 1 in Fig. 1 includes an information providing device 10, an information summarizing device 12, a recommendation device 14, a distribution device 16, and one or more user terminals 18. The information providing device 10, the information summarizing device 12, the recommendation device 14, the distribution device 16, and the user terminals 18 are connected to each other so as to be able to communicate data with each other via a communication network 20 such as a local area network (LAN) or the Internet.

[0018] The information providing device 10 is an information processing device or a website that provides various information such as news or papers. The information providing device 10 is, for example, a news site, arXiv, or ChemRxiv. For example, arXiv and ChemRxiv are examples of websites where pre-peer-reviewed papers are stored and made public.

[0019] The information summarizing device 12 is an information processing device or a website that summarizes summary information including the main points and classification of information using a trained large-scale language model. For example, the information summarizing device 12 can be realized by using the summarizing power of ChatGPT, an artificial intelligence chatbot service. The information summarizing device 12 summarizes the information acquired from the information providing device 10 into summary information including the main points and classification.

[0020] The recommendation device 14 is an information processing device or website that selects summary information that is likely to be of interest to a user or members of a group (team, department, etc.) to which the user belongs by updating the users to whom summary information is distributed by the distribution device 16 as described below.

[0021] The distribution device 16 is an information processing device, a website, or the like that distributes summary information to a user terminal 18 associated with a user. In addition to the function of distributing summary information, the distribution device 16 has a function of accepting a user's operation on the distributed summary information and providing it as reaction information. The user's operation on the distributed summary information is an operation showing a favorable reaction to the distributed summary information (for example, "like" or "good," which are examples of an intention to show sympathy or favor), or an operation of adding a comment to the distributed summary information. For example, the distribution device 16 can be realized by using Teams (registered trademark) or Slack (registered trademark), which are examples of communication tools.

[0022] The user terminal 18 is an information processing device such as a PC, a tablet terminal, or a smartphone operated by a user to whom summary information is delivered. The recommendation device 14 updates the content of the summary information delivered to the user based on the user's reaction information to the summary information delivered to the user terminal 18 (such as an operation showing a favorable reaction to the summary information or an operation of adding a comment to the delivered summary information).

[0023] The information summarization device 12 uses a trained large-scale language model to summarize information into summary information including key points and classifications. The trained large-scale language model is a natural language processing model trained using a large amount of text data. For example, the trained large-scale language model can be adapted to natural language processing such as text generation by fine-tuning a pre-trained trained model.

[0024] The information providing device 10, the information summarizing device 12, the recommending device 14, and the distribution device 16 are information processing devices such as a PC or a workstation. The configuration of the information distribution system 1 in FIG. 1 is an example. There are various system configuration examples of the information distribution system 1 depending on the use and purpose. For example, the information providing device 10, the information summarizing device 12, the recommending device 14, and the distribution device 16 may be realized by a plurality of computers, or may be realized as a cloud computing service. Furthermore, the recommending device 14 and the distribution device 16 may be configured as an integrated device. The information summarizing device 12, the recommending device 14, and the distribution device 16 may be configured as an integrated device.

[0025] <Hardware configuration> The information providing device 10, information summarizing device 12, recommendation device 14, distribution device 16, and user terminal 18 in FIG. 1 are realized by, for example, a computer 500 having the hardware configuration shown in FIG.

[0026] Fig. 2 is a diagram showing an example of a hardware configuration of a computer 500 according to this embodiment. The computer 500 in Fig. 2 includes an input device 501, a display device 502, an external I / F 503, a RAM 504, a ROM 505, a CPU 506, a communication I / F 507, and a HDD 508, all of which are connected to each other via a bus B. The input device 501 and the display device 502 may be connected for use.

[0027] The input device 501 is a touch panel, operation keys or buttons, a keyboard, a mouse, etc., used by the user to input various signals. The display device 502 is composed of a display such as a liquid crystal or organic EL display for displaying a screen, a speaker for outputting sound data such as voice and sound, etc. The communication I / F 507 is an interface for the computer 500 to perform data communication.

[0028] The HDD 508 is an example of a non-volatile storage device that stores programs and data. The stored programs and data include an OS, which is basic software that controls the entire computer 500, and applications that provide various functions on the OS. Note that the computer 500 may use a drive device (such as a solid-state drive (SSD)) that uses a flash memory as a storage medium instead of the HDD 508.

[0029] The external I / F 503 is an interface with an external device. The external device may include a recording medium 503a. This allows the computer 500 to read and / or write data from and to the recording medium 503a via the external I / F 503. The recording medium 503a may include a flexible disk, a CD, a DVD, an SD memory card, a USB memory, and the like.

[0030] The ROM 505 is an example of a non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. The ROM 505 stores programs and data such as BIOS, OS settings, and network settings that are executed when the computer 500 is started. The RAM 504 is an example of a volatile semiconductor memory (storage device) that temporarily retains programs and data.

[0031] The CPU 506 is a calculation device that reads out programs and data from storage devices such as the ROM 505 and the HDD 508 onto the RAM 504 and executes processing to realize the overall control and functions of the computer 500. The computer 500 according to this embodiment can realize various functions of the information providing device 10, the information summarizing device 12, the recommendation device 14, and the distribution device 16, which will be described later, by executing programs. The CPU 506 may read out the program via the external I / F 503 from a recording medium 503a in which the program is stored, and execute the program. Alternatively, the CPU 506 may read out the program from another communication network by data communication via the communication I / F 507, and execute the program.

[0032] <Functional configuration> The functional configuration of the information distribution system 1 according to this embodiment will be described.

[0033] Fig. 3 is a functional configuration diagram of an example of the information distribution system 1 according to this embodiment. Note that, in the configuration diagram of Fig. 3, parts that are not necessary for explaining this embodiment are appropriately omitted.

[0034] The information providing device 10 shown in Fig. 3 has an information providing unit 50. The information summarizing device 12 has an information summarizing unit 60, a summary information providing unit 62, and a trained large-scale language model 64. The recommending device 14 has an information acquiring unit 30, a summary information acquiring unit 32, a delivery request unit 34, a reaction information acquiring unit 36, a recommending unit 38, a command statement storage unit 40, a delivery destination storage unit 42, and a user information storage unit 44. The delivery device 16 has a delivery unit 70, a reaction information processing unit 72, and a reaction information providing unit 74.

[0035] The information providing unit 50 of the information providing device 10 provides various information such as news or papers to the recommendation device 14. The information acquiring unit 30 of the recommendation device 14 acquires various information such as news or papers from the information providing device 10. The summary information acquiring unit 32 reads out, for example, a command statement shown in FIG. 4 from the command statement storage unit 40.

[0036] Fig. 4 is an explanatory diagram of an example of a command statement according to this embodiment. Fig. 4(A) shows an example of a command statement in Japanese that specifies conditions for the trained large-scale language model 64 to generate summary information. Fig. 4(B) shows an example of a command statement in English that specifies conditions for the trained large-scale language model 64 to generate summary information. The command statement in Fig. 4 shows a prompt of ChatGPT as an example.

[0037] For example, in the command sentence in FIG. 4, rules are set as conditions for summary information to be generated by the trained large-scale language model 64 in a task of summarizing the main points of an English research paper in three points and outputting them in Japanese. The first rule is set to output a Japanese translation of the paper title, output the field of the content, express it concisely, and display unknown words and possible names of people, except for the title, in English. The second rule is set as a format to be strictly followed in the output, with the Japanese translation of the paper title, field: field of paper, main point 1: main point 1, main point 2: main point 2, and main point 3: main point 3.

[0038] Fig. 5 is an explanatory diagram of an example of a condition to be added to the command statement according to the present embodiment. By adding the command statement in Fig. 5 to the command statement in Fig. 4, for example, the field of the paper can be classified into the field of the content designated by the user, such as chemistry, image analysis, natural language processing, or other.

[0039] 4 to the information summarizing device 12, and causes the information summarizing device 12 to summarize, in accordance with the command statement, the summary information including, for example, the main points and classification of the information acquired by the information acquiring unit 30. The summary information acquiring unit 32 acquires the summary information summarized in accordance with the command statement from the information summarizing device 12. The delivery request unit 34 reads out delivery destination information from the delivery destination storage unit 42. The delivery request unit 34 requests the delivery device 16 to deliver the summary information acquired from the information summarizing device 12 to the user in accordance with the read delivery destination information.

[0040] The reaction information acquisition unit 36 ​​acquires, from the distribution device 16, user reaction information regarding the summary information distributed to the user by the distribution device 16. The reaction information acquisition unit 36 ​​stores the acquired user reaction information in, for example, the user information storage unit 44. The recommendation unit 38 updates the destination information in the destination storage unit 42 based on the acquired user reaction information so that summary information likely to be of interest to the user is distributed to the user. For example, the recommendation unit 38 can determine that the more operations indicating favorable reactions to the distributed summary information and the more operations of adding comments to the distributed summary information, the more likely the content of the summary information is to be of interest to the user.

[0041] The user information storage unit 44 may store information about members of a group to which the user belongs. The recommendation unit 38 may update the destination information in the destination storage unit 42 based on the acquired user reaction information so that summary information that may be of interest to members of the group to which the user belongs is delivered to the user.

[0042] The information summarization unit 60 of the information summarization device 12 receives information to be summarized from the recommendation device 14 together with, for example, the command sentence of Fig. 4 from the recommendation device 14. The information summarization unit 60 summarizes summary information including the main points and classification of the received information according to the received command sentence using a trained large-scale language model 64. The summary information providing unit 62 provides the summary information summarized by the information summarization unit 60 to the recommendation device 14.

[0043] The distribution unit 70 of the distribution device 16 distributes summary information to users in response to a request from the recommendation device 14. The reaction information processing unit 72 accepts user operations on the summary information distributed to the users as user reaction information. The user operations on the summary information accepted by the reaction information processing unit 72 include operations showing favorable reactions to the distributed summary information and operations of adding comments to the distributed summary information. The reaction information providing unit 74 provides the user reaction information accepted by the reaction information processing unit 72 to the recommendation device 14.

[0044] <Processing> The following describes the processing of the information distribution system 1 according to this embodiment. Fig. 6 is a flowchart of an example of the processing of the information distribution system according to this embodiment.

[0045] In step S10, the recommendation unit 38 of the recommendation device 14 initializes the destination information in the destination storage unit 42 so that random summary information is delivered to the user. The initialization of the destination information is not limited to the process of initializing the destination information so that random summary information is delivered to the user, but may be the process of initializing the destination information so that summary information in a field previously set by the user or an administrator is delivered to the user.

[0046] In step S12, the information acquisition unit 30 of the recommendation device 14 acquires various types of information such as news or papers from the information providing device 10. For example, the information acquisition unit 30 may acquire information such as abstract information from the information providing device 10 by using an API (Application Programming Interface) provided by the information providing device 10.

[0047] In step S14, the summary information acquisition unit 32 of the recommendation device 14 transmits the information acquired by the information acquisition unit 30 to the information summarizing device 12 together with, for example, the command statement of FIG. 4. The information summarizing unit 60 of the information summarizing device 12 receives the information to be summarized from the recommendation device 14 together with, for example, the command statement of FIG. 4. The information summarizing unit 60 uses the trained large-scale language model 64 to summarize, for example, summary information of FIG. 7 including the main points and classification of the received information according to the received command statement of FIG. 4. The summary information of FIG. 7 is a summary of a paper (https: / / arxiv.org / abs / 2303.06145v1) as an example of information to be summarized according to the command statement of FIG. 4.

[0048] FIG. 7 is an explanatory diagram of an example of summary information. The summary information in FIG. 7 is an example of summary information generated according to the condition specification performed in the command statement in FIG. 4. When the command statement in FIG. 5 is added to the command statement in FIG. 4, the field included in the summary information in FIG. 7 is classified into the fields of the content specified by the user, such as chemistry, image analysis, natural language processing, or others, as shown in FIG. 8. The field included in the summary information in FIG. 8 is a summary of a paper (https: / / arxiv.org / abs / 2303.16109v1) as an example of information to be summarized, according to the command statements in FIG. 4 and FIG. 5. The summary information providing unit 62 of the information summarizing device 12 provides the summary information summarized by the information summarizing unit 60 to the recommendation device 14.

[0049] In step S16, the delivery request unit 34 of the recommendation device 14 reads out delivery destination information from the delivery destination storage unit 42. The delivery destination information from the delivery destination storage unit 42 is updated by repeating the processes of steps S12 to S20 in Fig. 6 so that summary information that is likely to be of interest to the user is delivered to the user. The delivery request unit 34 requests the delivery device 16 to deliver the summary information acquired from the information summary device 12 to the user in accordance with the delivery destination information read out from the delivery destination storage unit 42.

[0050] The distribution unit 70 of the distribution device 16 distributes summary information to a user in response to a request from the recommendation device 14. For example, a user can check the distributed summary information by operating the user terminal 18. The user can also perform an operation to show a favorable reaction to the distributed summary information, such as "like" or "good." The user can also perform an operation to add a comment to the distributed summary information. Hereinafter, an operation to show a favorable reaction by the user and an operation to add a comment are referred to as a user's reaction.

[0051] Summary information that receives many reactions from users can be regarded as summary information that the users are interested in. Note that summary information that receives many reactions from users may be counted by group to which the users belong, and summary information that receives many reactions within a group may be regarded as summary information that the users belonging to that group are interested in.

[0052] In step S18, the reaction information processing unit 72 of the distribution device 16 accepts, as reaction information, the degree of reaction to the summary information distributed to the user. The reaction information acquisition unit 36 ​​of the recommendation device 14 acquires, from the distribution device 16, the user's reaction information to the summary information distributed to the user by the distribution device 16.

[0053] In step S20, the recommendation unit 38 of the recommendation device 14 updates the destination information in the destination storage unit 42 based on the acquired user reaction information so that summary information likely to be of interest to the user is delivered to the user. For example, the recommendation unit 38 can determine that the content of summary information likely to be of interest to the user is more likely the higher the similarity between the summary information that has been delivered and the summary information that has been subjected to the most operations showing favorable reactions to the summary information and the most operations of adding comments to the summary information.

[0054] After the process of step S20, the information distribution system 1 according to the present embodiment returns to the process of step S12, thereby repeating the processes of steps S12 to S20.

[0055] The process of step S14 in Fig. 6 is performed, for example, according to the procedure shown in Fig. 9. Fig. 9 is a flowchart showing an example of the process of step S14.

[0056] In step S30, the summary information acquisition unit 32 of the recommendation device 14 reads, for example, the command statement of Fig. 4 from the command statement storage unit 40. In step S32, the summary information acquisition unit 32 transmits the information acquired by the information acquisition unit 30 to the information summary device 12 together with, for example, the command statement of Fig. 4.

[0057] The information summarizing unit 60 of the information summarizing device 12 receives information to be summarized from the recommendation device 14 together with, for example, the command sentence of Figure 4. In step S34, the information summarizing unit 60 inputs the information to be summarized and, for example, the command sentence of Figure 4 to the trained large-scale language model 64.

[0058] In step S36, the trained large-scale language model 64 outputs summary information obtained by summarizing the information to be summarized, for example, according to the command sentence in Fig. 4. The summary information providing unit 62 of the information summarizing device 12 transmits the summary information summarized and output by the trained large-scale language model 64 to the recommendation device 14.

[0059] The processing of steps S18 to S20 in Fig. 6 is performed, for example, according to the procedure shown in Fig. 10. Fig. 10 is a flowchart showing an example of the processing of steps S18 to S20.

[0060] In step S50, the reaction information processing unit 72 of the distribution device 16 accepts, as reaction information, the degree of reaction to the summary information distributed to the user. The reaction information acquisition unit 36 ​​of the recommendation device 14 acquires, from the distribution device 16, the user's reaction information to the summary information distributed to the user by the distribution device 16.

[0061] In step S52, the recommending section 38 of the recommending device 14 selects operations of users who have responded favorably and operations of users who have added comments from the acquired user response information.

[0062] In step S54, the recommendation unit 38 scores the operations of the users who showed favorable reactions selected in step S52 and the operations of the users who added comments, for example, based on the number of times, and considers that summary information with a higher score has content that is more interesting to the user, and summary information with a lower score has content that is less interesting to the user.

[0063] The recommendation unit 38 updates the delivery destination information in the delivery destination storage unit 42 so that summary information that is likely to interest the user is delivered to the user. For example, the recommendation unit 38 updates the association between the content of summary information included in the delivery destination information and the delivery destination so that summary information that is highly similar to summary information deemed to be of interest to the user is delivered.

[0064] The information delivery system 1 according to this embodiment classifies information fields using the summarization ability of the trained large-scale language model 64, and acquires user reaction information to the summary information using communication tools, etc. The information delivery system 1 according to this embodiment can build a recommendation system based on the user reaction information to the acquired summary information, and can deliver the summary information to users from the recommendation system.

[0065] Therefore, according to the information distribution system 1 of this embodiment, the outline of various information such as the latest news or papers can be easily grasped in Japanese, and summary information of information that matches one's interests can be delivered.

[0066] Although the present embodiment has been described above, it will be understood that various changes in form and details can be made without departing from the spirit and scope of the claims. [Explanation of symbols]

[0067] 1 Information distribution system 10 Information provision device 12 Information Summarizer 14 Recommendation device 16 Broadcasting Device 18 User terminal 30 Information acquisition department 32 Summary information acquisition unit 34 Delivery request section 36 Reaction information acquisition unit 38 Recommendation Department 50 Information provision department 60 Information Summary Section 62 Summary information provision department 64 large-scale pre-trained language models 70 Distribution Department 72 Reaction information processing section 74 Reaction Information Department 500 Computers

Claims

1. An information delivery system for delivering summary information to one or more users, comprising: a summary information acquisition unit that acquires the summary information including the main points and classification of the acquired information by using a trained large-scale language model; a distribution unit that distributes the summary information to the user; a recommendation unit that updates the users to whom the summary information is to be distributed by the distribution unit based on reaction information of the users to the distributed summary information; An information distribution system having the above configuration.

2. the summary information acquisition unit inputs the acquired information to the large-scale language model together with a command statement specifying a condition for the large-scale language model to generate the summary information, and acquires the summary information including a gist and a classification of the acquired information.

2. The information distribution system according to claim 1,

3. The distribution unit provides the recommendation unit with an operation of the user with respect to the distributed summary information as the reaction information.

3. The information distribution system according to claim 1 or 2,

4. The user's operation is an operation showing a favorable reaction to the distributed summary information or an operation of adding a comment to the distributed summary information.

4. The information distribution system according to claim 3,

5. The recommendation unit updates the users to whom the summary information is to be distributed by the distribution unit based on the reaction information collected for each group to which the users belong.

3. The information distribution system according to claim 1 or 2,

6. An information distribution method for an information distribution system that distributes summary information to one or more users, comprising the steps of: a summary information acquisition step of acquiring the summary information including the gist and classification of the acquired information by using a trained large-scale language model; a delivery step of delivering the summary information to the user; a recommendation step of updating the users to whom the summary information is to be delivered in the delivery step based on reaction information of the users to the delivered summary information; An information distribution method having the above structure.

7. A computer for delivering summary information to one or more users, a summary information acquisition step of acquiring the summary information including the gist and classification of the acquired information by using a trained large-scale language model; a delivery step of delivering the summary information to the user; a recommendation step of updating the users to whom the summary information is to be delivered in the delivery step based on reaction information of the users to the delivered summary information; A program for executing the above.

Images