Information processing device, information processing method, program

Abstract

We provide a system for efficiently creating meeting minutes. [Solution] The information processing device includes a first output means that outputs an instruction to estimate the field of speech in a meeting, an acquisition means that acquires the field of speech estimated based on the instruction output by the first output means, and a second output means that outputs the field of speech acquired by the acquisition means, the content of the speech, and an instruction to create meeting minutes based on the field of speech and the content of the speech. The acquisition means acquires the meeting minutes created based on the instruction output by the second output means.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and a program.

Background Art

[0002] With the recent improvement in the accuracy of speech recognition tools and the spread of generative AI (Artificial Intelligence), there is a desire to automatically create meeting minutes from the text converted using a speech recognition tool by recording or transcribing a meeting. Patent Document 1 discloses a technique for creating a summary sentence by extracting summary speech from the classification of speech and the identification of progress stages in order to easily grasp important matters in speech at meetings and lectures. Patent Document 2 discloses a technique in which a plurality of templates called prototypes of meeting minutes are prepared in advance, and a person in charge manually inputs necessary matters and links them to task management such as decision-making matters.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] Meeting minutes are generally prepared in different ways depending on their purpose. For example, minutes of a management meeting might focus on numerical targets, while minutes of a development meeting might focus on schedule management. When converting important statements into short sentences or written language, as in Patent Document 1, it is necessary to devise a method for determining important statements according to the format of the meeting. Another possible solution for changing how meeting minutes are summarized is a template system, but Patent Document 2 does not reduce the workload of creating meeting minutes, even including the pre-creation of templates. Unless the information to be entered can be automatically filled in using a template, the workload of the template system will not be reduced. The present invention aims to provide a mechanism for efficiently creating meeting minutes. [Means for solving the problem]

[0005] The information processing device of the present invention comprises: a first output means for outputting an instruction for estimating the field of speech in a meeting; an acquisition means for acquiring the field of speech estimated based on the instruction output by the first output means; and a second output means for outputting the field of speech acquired by the acquisition means, the content of the speech, and an instruction for creating minutes of the meeting based on the field of speech and the content of the speech, wherein the acquisition means acquires minutes of the meeting created based on the instruction output by the second output means. [Effects of the Invention]

[0006] According to the present invention, it is possible to provide a mechanism for efficiently creating meeting minutes. [Brief explanation of the drawing]

[0007] [Figure 1] This figure shows an example of the functional configuration of an information processing device. [Figure 2] This figure shows an example of the hardware configuration of an information processing device. [Figure 3] A flowchart illustrating an example of processing performed by an information processing device. [Figure 4] This is a flowchart illustrating an example of speech recognition processing. [Figure 5] This is a flowchart illustrating an example of the process for creating meeting minutes. [Figure 6] This diagram illustrates an example of keywords used in meeting minutes. [Figure 7] This diagram illustrates an example of a prompt related to keyword estimation. [Figure 8] This flowchart shows an example of the process for extracting meeting minutes. [Figure 9] This diagram illustrates an example of a prompt related to agenda extraction. [Figure 10] This diagram illustrates an example of a prompt related to extracting detailed information. [Figure 11] This diagram illustrates an example of a prompt related to question and answer extraction. [Figure 12] This diagram illustrates an example of a prompt for extracting decisions and other relevant information. [Figure 13] This flowchart shows an example of speech summarization processing. [Figure 14] This diagram illustrates an example of a prompt related to speech summarization. [Figure 15] This is a diagram showing an example of meeting minutes. [Modes for carrying out the invention]

[0008] Embodiments of the present invention will be described below with reference to the drawings.

[0009] Figure 1 is a block diagram showing an example of the functional configuration of an information processing device in one embodiment of the present invention. The information processing device 100 in this embodiment includes a speech recognition unit 101, a data management unit 102, a meeting minutes creation unit 103, a meeting minutes management unit 104, and a presentation unit 105.

[0010] The speech recognition unit 101 converts the speech data included in the video and audio data 110 of a meeting or the like into text by speech recognition, and creates a text including the content of the speech in the meeting or the like. That is, the speech recognition unit 101 performs speech-to-text conversion on the video and audio data 110 of a meeting or the like to create a text. The data management unit 102 manages the text (speech-to-text conversion result) created by the speech recognition unit 101, the original video and audio data 110, and the like.

[0011] The minutes creation unit 103 creates minutes or the like based on the text created by the speech recognition unit 101. For example, the minutes creation unit 103 extracts the topics in the meeting, the detailed content for each topic, question-and-answer sessions, and decisions from the text created by the speech recognition unit 101 to create minutes. Also, for example, the minutes creation unit 103 creates a speech summary (summary of the content of the speech) based on the text created by the speech recognition unit 101.

[0012] The minutes management unit 104 manages the minutes and the like created by the minutes creation unit 103. The presentation unit 105 presents the data managed by the data management unit 102, the information created by the minutes creation unit 103 managed by the minutes management unit 104, and the like to users and the like.

[0013] FIG. 2 is a block diagram showing an example of the hardware configuration of the information processing apparatus in the present embodiment. As shown in FIG. 2, the information processing apparatus 100 has a CPU (Central Processing Unit) 201, a ROM (Read Only Memory) 202, a RAM (Random Access Memory) 203, an input controller 205, a video controller 206, a memory controller 207, and a communication I / F controller 208 connected via a system bus 204.

[0014] The CPU 201 comprehensively controls each device and controller connected to the system bus 204.

[0015] ROM202 or external memory211 holds the BIOS (Basic Input / Output System) and OS (Operating System), which are control programs executed by the CPU201, as well as computer-readable and executable programs and various necessary data (including data tables) for implementing the processes described later.

[0016] RAM203 functions as the main memory, work area, etc., of the CPU201. The CPU201 loads the necessary programs, etc., from ROM202 or external memory211 into RAM203 and executes the loaded programs to perform various operations. The functions of the first output means, second output means, and acquisition means are realized when the CPU201 reads and executes a program from ROM202 or external memory211 into RAM203.

[0017] The input controller 205 controls input from the input device 209, such as a keyboard or a pointing device like a mouse. If the input device is a touch panel, the user can give various instructions by pressing (touching with a finger, etc.) icons, cursors, or buttons displayed on the touch panel. The touch panel may be a multi-touch screen or other touch panel capable of detecting the positions of multiple fingers touching it.

[0018] The video controller 206 controls the display to an external output device such as the display 210. The display includes the display of a notebook computer integrated with the main unit. The external output device is not limited to a display; for example, it may be a projector. Furthermore, for the aforementioned touch-enabled device, an input device is also provided. The video controller 206 can control the video memory (VRAM) for display control, and can utilize a portion of the RAM 203 as the video memory area, or it is possible to provide a separate dedicated video memory.

[0019] The memory controller 207 controls access to the external memory 211. The external memory 211 can be an external storage device (hard disk drive or solid-state drive) that stores boot programs, various applications, font data, user files, editing files, and various other data, a flexible disk (FD), or a CompactFlash® memory connected to a PCMCIA card slot via an adapter.

[0020] The communication interface controller 208 connects to and communicates with external devices via a network and performs communication control processing over the network. For example, it can handle communication using TCP / IP, telephone lines such as ISDN, and mobile communication networks such as mobile phones.

[0021] The CPU 201 enables display on the display 210 by, for example, performing the process of expanding (rasterizing) outline fonts into the display information area in RAM 203. The CPU 201 also enables user input via a mouse cursor (not shown) on the display 210.

[0022] Next, the processing performed by the information processing device 100 in this embodiment will be described. The processing in each flowchart described below is realized, for example, by the CPU 201 of the information processing device 100 loading a program stored in the ROM 202 or external memory 211, etc., into the RAM 203 and executing it.

[0023] Figure 3 is a flowchart showing an example of a process performed by the information processing device 100 in this embodiment. In step S301, the CPU 201 performs speech recognition processing on video and audio data such as meetings, and creates text related to the audio data contained in the video and audio data. In other words, in step S301, the CPU 201 transcribes the video and audio data such as meetings and creates text. The speech recognition processing in step S301 will be explained with reference to Figure 4.

[0024] Figure 4 is a flowchart showing an example of the speech recognition process performed in step S301 of Figure 3. In step S401, CPU201 converts the audio data contained in video and audio data such as meetings into text (transcribes it) using a speech recognition tool, and creates text related to the audio data.

[0025] In step S402, the CPU 201 determines whether the speech recognition tool used for speech recognition has a speaker splitting function. If the CPU 201 determines that the speech recognition tool has a speaker splitting function (YES in step S402), the process proceeds to step S403. On the other hand, if the CPU 201 determines that the speech recognition tool does not have a speaker splitting function (NO in step S402), the process proceeds to step S404.

[0026] In step S403, the CPU 201 associates the speaker information with the transcribed speech content and registers it in the data management unit 102. In step S404, the CPU 201 registers the transcribed speech content in the data management unit 102. After registering with the data management unit 102 in step S403 or step S404, CPU 201 returns to the flowchart process shown in Figure 3.

[0027] Returning to Figure 3, in step S302, the CPU 201 performs a meeting minutes creation process based on the results of the speech recognition process registered in the data management unit 102 during the speech recognition process in step S301, and creates the meeting minutes. Referring to Figure 5, the meeting minutes creation process in step S302 will be explained. The meeting minutes creation process using a generation AI will be explained below, but the generation AI may be located inside or outside the information processing device. If it is located outside, the meeting minutes creation unit outputs instructions to the generation AI and obtains the processing results from the generation AI to realize the meeting minutes creation process.

[0028] Figure 5 is a flowchart showing an example of the meeting minutes creation process performed in step S302 of Figure 3. In step S501, the CPU201 estimates meeting minutes keywords based on the utterances from the meeting that have been transcribed using speech recognition processing. Meeting minutes keywords are groups of keywords that are predefined as key points for creating meeting minutes, as exemplified in Figure 6. In the example shown in Figure 6, multiple keywords (phrases) that may be included in utterances at a meeting related to a theme (field) are defined as groups of words for each theme (field). Keywords (phrases) are, for example, words related to the agenda of the meeting or the attributes of the meeting participants. In keyword estimation, the CPU201 targets the data of the transcribed utterances and estimates which group of keywords is most frequently included. Then, the CPU201 selects the group that contains the most estimated keywords as the group (field) related to the meeting minutes.

[0029] In step S501, keyword estimation is performed using a generative AI (Artificial Intelligence) that employs a trained model, such as a language model. The CPU 201 performs keyword estimation by inputting a prompt to the generative AI that includes, for example, the transcribed content of the meeting utterances, the minutes keywords for each group, and an instruction to estimate the group (field) related to the meeting utterances based on the minutes keywords, as shown in Figure 7(a). The CPU 201 then acquires the estimated group (field) information related to the utterances, which is generated and output by the generative AI based on the instructions in the prompt, as shown in Figure 7(b) as an example. Note that the estimation method can also utilize natural language processing techniques such as pattern matching or morphological analysis.

[0030] Furthermore, if the meeting for which minutes are to be created is clearly defined, users can also specify keyword groups. In that case, for example, if it is clearly a development meeting, the user can select the "development" group as the keyword group. Alternatively, keyword groups may be set based on the intended audience for whom the minutes will be shared. For example, in the case of minutes for a meeting between development and intellectual property, the "development" group may be selected as the keyword group for the development minutes, and the "intellectual property" group may be selected as the keyword group for the intellectual property minutes. In this case, the system may also be configured to automatically retrieve the affiliation of the minute-taker and select a keyword group based on that affiliation.

[0031] Next, in step S502, the CPU 201 performs a meeting minutes extraction process based on the speech content of the meeting that has been transcribed by speech recognition processing and the groups (fields) related to the speech content acquired in step S501. The meeting minutes extraction process in step S502 will be explained with reference to Figure 8.

[0032] Figure 8 is a flowchart showing an example of the meeting minutes extraction process performed in step S502 of Figure 5. In step S801, the CPU 201 extracts the agenda from the transcribed meeting speech content. For example, the agenda is extracted by estimation using a generative AI that employs a trained model such as a language model. The CPU 201 extracts the agenda by inputting a prompt to the generative AI that includes the transcribed meeting speech content, the keyword group obtained in step S501, and an instruction to output the meeting agenda, as shown in Figure 9(a). The CPU 201 then obtains information about the meeting agenda, such as an example shown in Figure 9(b), which is generated and output by the generative AI based on the prompt's instructions. The obtained information about the meeting agenda is stored in the minutes management unit 104 as meeting minutes information. In this way, by including the keyword group obtained in step S501 as a hint in the prompt given to the generative AI, it becomes possible to extract agenda items that take keywords into consideration during the agenda extraction process, making it possible to extract agenda items that are suitable for the target audience viewing the meeting and minutes.

[0033] In step S802, the CPU 201 extracts detailed information about each agenda item extracted in step S801. For example, the CPU 201 extracts details such as implementation details, future plans, and numerical targets. For example, the extraction of detailed information is performed by estimation using a generative AI that uses a trained model such as a language model. The CPU 201 extracts detailed information by inputting prompts to the generative AI that include instructions to output detailed information for each agenda item, as shown in Figure 10(a) as an example, and obtains information about the detailed information that is generated and output by the generative AI based on the instructions of the prompts, as shown in Figure 10(b) as an example. The obtained information about the detailed information is stored in the minutes management unit 104 as information for the meeting minutes.

[0034] In the example described above, the generating AI is first instructed to extract the agenda items in step S801, and then to list the details of each agenda item, including the implementation details, results, and future plans. However, it is not limited to this, and in step S802, a prompt may be entered into the generating AI to instruct it to extract details again, including the content spoken in the meeting and the extracted agenda items. Also, similar to step S801, when instructing the generating AI to extract details for each agenda item, the keyword group obtained in step S501 may be included as a hint in the prompt given to the generating AI. This may include not only the keyword group obtained in step S501, but also keyword groups set by the user. This makes it possible to extract detailed content that takes keywords into consideration when extracting details for each agenda item, and to extract content that is suitable for the target audience who will be viewing the meeting and minutes.

[0035] In step S803, the CPU 201 extracts questions and answers from the transcribed meeting speech. The CPU 201 extracts statements related to questions and answers, consisting of question sentences and their corresponding response parts, from the transcribed meeting speech. For example, the extraction of questions and answers is performed by estimation using a generation AI that uses a trained model such as a language model. The CPU 201 extracts questions and answers by inputting a prompt to the generation AI that includes the transcribed meeting speech, the keyword group obtained in step S501, and an instruction to output the questions and answers from the meeting, as shown in Figure 11(a) as an example. The CPU 201 then obtains information about the questions and answers from the meeting, which is generated and output by the generation AI based on the instructions in the prompt, as shown in Figure 11(b) as an example. The obtained information about the questions and answers from the meeting is stored in the minutes management unit 104 as meeting minutes information.

[0036] Furthermore, if speaker information and the transcribed utterance are associated in step S403, the system may be configured to output the extracted question-and-answer portion in association with the questioner and responder. If no question-and-answer is found, the system can be instructed to output "No question-and-answer," thereby preventing hallucination if no question-and-answer exists in the transcribed utterance. Additionally, by including the keyword group obtained in step S501 as a hint in the prompt given to the generating AI when instructing it to extract question-and-answer, it becomes possible to extract question-and-answer that takes keywords into consideration, making it possible to extract question-and-answer that is suitable for the target audience viewing the meeting or minutes.

[0037] Furthermore, keyword groups corresponding to the speaker's information (e.g., affiliation and role) can be set as hints in the prompts given to the generating AI, and the AI ​​can be instructed to extract questions and answers based on those hints. For example, if the questioner, "Mr. Suzuki," has the affiliation and role of "Affiliation: Development, Role: Management," then by using the keyword group for development (plan, personnel, schedule, request, consultation, cost, etc.) and the keyword group for management (goal, report, approval, etc.) as hints to extract questions and answers, it becomes possible to extract questions and answers that take into account keywords likely to be relevant to Mr. Suzuki's questions and answers. This makes it possible to estimate keywords likely to be included in each individual's utterance and achieve comprehensive extraction of questions and answers.

[0038] Furthermore, in the example shown in Figure 11(a), the instruction is "Write down the speaker and a summary of the content of the statement for each question and answer session from the statement," but it is also possible to give instructions in two stages, such as, "A statement is a pair of (speaker) and content of the statement. Please perform the following two tasks. First, find all question and answer pairs from the statement. Next, for all the question and answer sessions you have found, write down the questioner: summary of the question, and the respondent: summary of the answer. Do not write down item names such as "questioner" or "content," just write down the name and content."

[0039] In step S804, the CPU 201 extracts decisions and other information from the transcribed meeting speeches. The CPU 201 extracts decisions made at the meeting and matters related to future plans from the transcribed meeting speeches. For example, decisions and other information are extracted by estimation using a generation AI that uses a trained model such as a language model. The CPU 201 extracts decisions and other information by inputting a prompt to the generation AI that includes the transcribed meeting speeches, the keyword group obtained in step S501, and an instruction to output the decisions made at the meeting, as shown in Figure 12(a) as an example. The CPU 201 then obtains information about the decisions made at the meeting, as shown in Figure 12(b) as an example, which is generated and output by the generation AI based on the instructions in the prompt. The obtained information about the decisions made at the meeting is stored in the minutes management unit 104 as meeting minutes information. In this way, by including the keyword group obtained in step S501 as a hint in the prompt given to the generating AI, it becomes possible to extract decisions and other matters that take keywords into consideration, and to extract decisions and other matters that are suitable for the target audience who will be viewing the meeting minutes.

[0040] As described above, information regarding items that may be included in the meeting minutes is created in each of the processes from steps S801 to S804. After executing the process in step S804, the process returns to the flowchart shown in Figure 3. In the explanation above, each process from steps S801 to S804 is implemented using a generative AI with a trained model such as a language model, but some or all of these processes may be implemented using natural language processing such as pattern matching or morphological analysis.

[0041] Furthermore, while the flowcharts shown in Figures 5 and 8 perform keyword estimation for one transcribed speech (transcription result), in meetings where multiple different topics are discussed, it is possible to accommodate various types of meetings by implementing a process that performs keyword estimation for each topic extraction process. For example, in a meeting where each department reports on various topics, keyword estimation can be performed for each topic, or keywords can be set for each reporting department, enabling extraction processing that is appropriate for the topic.

[0042] Returning to Figure 5, in step S503, the CPU 201 performs speech summarization processing based on the utterances in the meeting that have been transcribed by speech recognition processing and the groups (fields) related to the utterances obtained in step S501. Speech summarization processing is the process of concisely summarizing each utterance. Generally, human speech includes fillers such as "um" and "uh," and often contains redundant utterances when written. For example, the person creating the minutes of a meeting may want to concisely review the utterances in the meeting, and speech summarization processing is performed to reduce this review work. Refer to Figure 13 to explain the utterance summarization processing in step S503.

[0043] Figure 13 is a flowchart showing an example of the speech summarization process performed in step S503 of Figure 5. In step S1301, the CPU 201 performs filler removal from the transcribed speech content of the meeting. Some speech recognition tools used in step S401 can remove fillers contained in the speech, but depending on the performance of the tool, some cannot, so the process in step S1301 should be performed. Filler removal in step S1301 can be performed, for example, using natural language processing such as pattern matching or morphological analysis, or using generative AI with a trained model such as a language model.

[0044] In step S1302, the CPU 201 creates a speech summary based on the transcribed meeting speech content, which was filtered in step S1301. For example, in the speech summary creation process, if conventional natural language processing is used, an extractive summarization method that extracts the content as is may be used, or if a generative AI using a trained model such as a language model is used, an abstractive summarization method that generates new content based on the content may be used. The format of the speech summary may be a bulleted list for each speech, or a summary of the entire speech. By creating a speech summary, the person creating the minutes can easily check the content of the speech. The created speech summary is stored in the minutes management unit 104 as information for the meeting minutes.

[0045] When generating a summary of a speech using the generation AI, the CPU 201 creates the speech summary by inputting a prompt to the generation AI that includes, for example, the textual content of the meeting after filler removal, the keyword group acquired in step S501, and an instruction to summarize the utterance, as shown in Figure 14(a). The CPU 201 then acquires information about the speech summary, which is generated and output by the generation AI based on the instructions in the prompt, as shown in Figure 14(b) as an example. The acquired information about the speech summary is stored in the minutes management unit 104 as information for the meeting minutes. In this way, by including the keyword group acquired in step S501 as a hint in the prompt given to the generation AI, it becomes possible to create a summary that takes keywords into consideration when summarizing speeches, and it becomes possible to efficiently summarize content that is suitable for the target audience viewing the meeting and minutes.

[0046] Furthermore, keyword groups based on speaker information (e.g., affiliation and role) can be set as hints in prompts given to the generating AI, and the AI ​​can be instructed to summarize the utterance based on these hints. This makes it possible to summarize the utterance, including its key points, according to the keywords that are likely to be included in each individual's utterance.

[0047] Returning to Figure 3, in step S303, the CPU 201 presents the minutes created in the minutes creation process in step S302 to the user via the presentation unit 105. Figure 15 shows an example of the created minutes. The user can obtain the minutes automatically created by the information processing device 100. The presentation unit 105 may also be equipped with a minutes editing function, allowing the user to edit the minutes created and presented by the information processing device 100 as a template to create the final minutes. The CPU 201 may also refer to and present audio data managed by the data management unit 102, etc., in addition to the created minutes and speech summaries, at the request of the user. Furthermore, in this embodiment, the minutes creation process was performed based on predetermined minutes items, but the system may be configured to change the minutes items according to the group (field) related to the minutes inferred in step S501, for example.

[0048] Furthermore, in this embodiment, keyword groups are included as hints in the prompts input to the generating AI, but the method is not limited to the example described above. Any method that allows the generating AI to perform the extraction process described above while considering the estimated keyword groups may be used.

[0049] As described above, by obtaining information on groups (fields) related to the content of speech in a meeting from the transcribed speech content, and then issuing instructions to create meeting minutes based on the obtained groups (fields) related to the speech content, it is possible to create meeting minutes from audio data of meetings, etc., thereby reducing the workload of the minute-taker when creating minutes based on audio data. Furthermore, since information on each item of the meeting minutes is obtained based on the obtained groups (fields) related to the speech content, it is possible to create meeting minutes that are appropriate for the target audience who will be viewing the meeting and minutes.

[0050] In the embodiment described above, when creating meeting minutes using a generative AI, the minutes are created using a generative AI that uses a pre-trained model such as a language model. However, it is also possible to create meeting minutes using a generative AI that uses a multimodal model capable of receiving audio data as a pre-trained model. When using a multimodal model as a pre-trained model, it becomes possible to input prompts containing audio data from the meeting into the generative AI, and the meeting minute creation process can be performed without executing the speech recognition process in step S301 shown in Figure 3.

[0051] Although embodiments of the present invention have been described above, the present invention can take the form of, for example, a system, apparatus, method, program, or recording medium. Specifically, it may be applied to a system consisting of multiple devices, or to an apparatus consisting of a single device.

[0052] Furthermore, the program in the present invention is a program that allows a computer to execute the flowchart processing method described above, and the recording medium in the present invention stores the program that allows a computer to execute the flowchart processing method described above.

[0053] As described above, it goes without saying that the object of the present invention can also be achieved by supplying a recording medium containing a program that realizes the functions of the embodiments described above to a system or device, and by having the computer (or CPU or MPU) of that system or device read and execute the program stored on the recording medium.

[0054] In this case, the program read from the recording medium itself realizes the novel function of the present invention, and the recording medium on which the program is recorded constitutes the present invention. Examples of recording media that can be used to supply the program include flexible disks, hard disk drives, solid-state drives, optical disks, magneto-optical disks, CD-ROMs, CD-Rs, DVD-ROMs, magnetic tapes, non-volatile memory cards, ROMs, EEPROMs, silicon disks, etc.

[0055] Furthermore, it goes without saying that the functions of the aforementioned embodiments are realized not only by the computer executing the program it has read, but also by the operating system (OS) running on the computer performing some or all of the actual processing based on the instructions of that program, thereby realizing the functions of the aforementioned embodiments.

[0056] Furthermore, it goes without saying that this also includes cases where, after a program read from a recording medium is written to the memory of a function expansion board inserted into a computer or a function expansion unit connected to a computer, the CPU or other components of the function expansion board or function expansion unit perform some or all of the actual processing based on the instructions of the program code, and the functions of the aforementioned embodiments are realized through that processing.

[0057] Furthermore, the present invention may be applied to a system consisting of multiple devices or to a device consisting of a single device. It goes without saying that the present invention can also be applied when the results are achieved by supplying a program to a system or device. In this case, by reading a recording medium containing a program for achieving the present invention into the system or device, the system or device can enjoy the effects of the present invention.

[0058] Furthermore, by downloading and reading the program for achieving the present invention from a server, database, etc. on a network using a communication program, the system or device can enjoy the effects of the present invention. It should be noted that configurations combining the aforementioned embodiments and their variations are also included in the present invention.

[0059] It should be noted that the embodiments described above are merely examples of how the present invention can be implemented, and the technical scope of the present invention should not be interpreted as being limited by them. In other words, the present invention can be implemented in various forms without departing from its technical concept or its main features.

[0060] The disclosure of this embodiment includes the following configurations and methods, etc. (Composition 1) A first output means that outputs instructions for estimating the field of speech in a meeting, An acquisition means for acquiring the field of the utterance estimated based on the instruction output by the first output means, The system includes a second output means that outputs the field of the utterance acquired by the acquisition means, the content of the utterance, and an instruction to create minutes of the meeting based on the field of the utterance and the content of the utterance. The information processing device is characterized in that the acquisition means acquires the minutes of the meeting, which were created based on the instructions output by the second output means. (Configuration 2) The information processing device according to configuration 1, characterized in that the first output means outputs the content of an utterance in a meeting, a group of words defined for each field that may be included in the utterance, and an instruction to estimate the field of the utterance based on the group of words. (Composition 3) The information processing device according to configuration 2, characterized in that the aforementioned term is a term related to the agenda of a meeting or the attributes of the meeting participants. (Composition 4) The information processing device according to configuration 1 or 2, characterized in that the minutes of the meeting include information on at least one of the following items: agenda, decisions, summaries of statements, and questions and answers from the meeting. (Composition 5) The information processing device according to any one of configurations 1 to 4, characterized in that the instruction output by the first output means is output as a prompt to the trained model. (Composition 6) The information processing device according to configuration 4, characterized in that information regarding the items of the meeting minutes is created by outputting the instruction output by the second output means as a prompt to the trained model. (Composition 7) The information processing device according to configuration 5 or 6, characterized in that the trained model is a language model. (Composition 8) The information processing device according to configuration 5 or 6, characterized in that the trained model is a multimodal model. (Method 1) An information processing method performed by an information processing device, A first output step outputs instructions for estimating the area of ​​speech in a meeting, A first acquisition step of acquiring the field of the utterance estimated based on the instruction output by the first output means, A second output step outputs the field of the utterance obtained in the first acquisition step, the content of the utterance, and an instruction to create minutes of the meeting based on the field of the utterance and the content of the utterance. An information processing method characterized by comprising a second acquisition step of acquiring minutes of the meeting prepared based on instructions output by the second output means. (Program 1) A program for causing at least one computer to function as one of the means of the information processing device described in any one of items 1 to 8. [Explanation of Symbols]

[0061] 100 Information Processing Devices 101 Voice Recognition Unit 102 Data Management Department 103 Minutes Preparation Department 104 Minutes Management Department 105 Presentation section

Claims

1. A first output means that outputs instructions for estimating the field of speech in a meeting, An acquisition means for acquiring the field of the utterance estimated based on the instruction output by the first output means, The system includes a second output means that outputs the field of the utterance acquired by the acquisition means, the content of the utterance, and an instruction to create minutes of the meeting based on the field of the utterance and the content of the utterance. The acquisition means is characterized by acquiring the minutes of the meeting, which were created based on the instructions output by the second output means.

2. The information processing apparatus according to claim 1, characterized in that the first output means outputs the content of an utterance in a meeting, a group of words defined for each field that may be included in the utterance, and an instruction to estimate the field of the utterance based on the group of words.

3. The information processing apparatus according to claim 2, characterized in that the aforementioned terms are terms related to the agenda of the meeting or the attributes of the meeting participants.

4. The information processing device according to claim 2, characterized in that the minutes of the meeting include information on at least one of the following items: agenda items, decisions made, summaries of statements, and questions and answers from the meeting.

5. The information processing device according to claim 1, characterized in that the instruction output by the first output means is output as a prompt to the trained model.

6. The information processing device according to claim 4, characterized in that information relating to the items of the meeting minutes is created by outputting the instruction output by the second output means as a prompt to the trained model.

7. The information processing device according to claim 5 or 6, characterized in that the trained model is a language model.

8. The information processing device according to claim 5 or 6, characterized in that the trained model is a multimodal model.

9. An information processing method performed by an information processing device, A first output step outputs instructions for estimating the area of ​​speech in a meeting, A first acquisition step of acquiring the field of the utterance estimated based on the instruction output by the first output means, A second output step outputs the field of the utterance obtained in the first acquisition step, the content of the utterance, and an instruction to create minutes of the meeting based on the field of the utterance and the content of the utterance. An information processing method characterized by comprising a second acquisition step of acquiring minutes of the meeting prepared based on instructions output by the second output means.

10. A program for causing at least one computer to function as each of the means of the information processing apparatus described in any one of claims 1 to 6.

Images