Meeting minutes generation method, device, terminal and computer-readable storage medium

Abstract

The present application relates to the field of computer technology. The embodiment of the present application discloses a method, device, terminal and computer-readable storage medium for generating meeting minutes. The method includes obtaining meeting data for generating meeting minutes, extracting characteristic information from the meeting data to obtain characteristic text data; constructing a meeting minutes prompt word template for the characteristic text data to obtain a specific meeting minutes prompt word template corresponding to the meeting data, constructing a meeting minutes prompt word through the specific meeting minutes prompt word template and the characteristic text data; inputting the meeting minutes prompt word into a meeting minutes generation model, and outputting target meeting minutes for the meeting minutes prompt word based on the meeting minutes generation model. This solves the technical problem that the generation efficiency of meeting minutes generated by manual recording is low and requires a lot of time.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, terminal and computer-readable storage medium for generating meeting minutes. Background Art

[0002] In daily meetings, it is usually necessary to manually organize and summarize the content of the meeting to obtain meeting minutes. However, the generation efficiency of meeting minutes generated by manual recording is low and it takes a lot of time. Summary of the invention

[0003] The embodiments of the present application provide a method, device, terminal and computer-readable storage medium for generating meeting minutes, which can solve the technical problem that the generation efficiency of meeting minutes generated by manual recording is low and requires a lot of time.

[0004] In a first aspect, an embodiment of the present application provides a method for generating meeting minutes, the method comprising:

[0005] Acquire meeting data for which meeting minutes are to be generated, and perform feature information extraction processing on the meeting data to obtain feature text data;

[0006] Performing a conference minutes prompt word template construction process on the characteristic text data to obtain a specific conference minutes prompt word template corresponding to the conference data, and constructing a conference minutes prompt word through the specific conference minutes prompt word template and the characteristic text data;

[0007] The meeting minutes prompt words are input into the meeting minutes generation big model, and the target meeting minutes for the meeting minutes prompt words are output based on the meeting minutes generation big model.

[0008] Optionally, the step of constructing a conference minutes prompt word template for the characteristic text data to obtain a specific conference minutes prompt word template corresponding to the conference data includes:

[0009] Determine the meeting key points attribute corresponding to the characteristic text data;

[0010] A specific meeting minutes prompt word template corresponding to the meeting data is generated based on the meeting key point attributes.

[0011] Optionally, determining the attributes of meeting points corresponding to the characteristic text data includes:

[0012] Determine the meeting key points element corresponding to the characteristic text data, extract common attributes for the meeting key points element, and obtain the meeting common key points attributes;

[0013] Extracting proprietary attributes from the meeting key points elements to obtain proprietary meeting key points attributes;

[0014] The conference key point attributes are determined based on the conference general key point attributes and the conference specific key point attributes.

[0015] Optionally, the generating a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attributes includes:

[0016] Generate a basic template for meeting minutes reminder words based on the general meeting key points attributes;

[0017] A proprietary description text is generated based on the proprietary key point attributes of the conference, and description editing processing is performed on the basic template of the conference minutes prompt words based on the proprietary description text to obtain a specific conference minutes prompt word template corresponding to the conference data.

[0018] Optionally, the conference data includes audio data, and the extracting feature information from the conference data to obtain feature text data includes:

[0019] Conduct voice monitoring of conference voice;

[0020] When the result of the voice monitoring is that the conference voice is in progress, real-time audio data is obtained from the conference data, and when it is determined that the duration corresponding to the audio data is greater than or equal to the preset duration, the audio data is subjected to stage audio data caching processing to obtain current stage audio data; the current stage audio data is subjected to text conversion processing to obtain stage text data, and the step of performing voice monitoring on the conference voice is returned to be executed;

[0021] When the result of the voice monitoring is that the conference voice is in an ending state, text data is determined based on the stage text data, and text feature extraction processing is performed on the text data to obtain feature text data.

[0022] Optionally, after performing text conversion processing on the current stage audio data to obtain stage text data, the method further includes:

[0023] Sending the stage text data to the audio-to-text display service, and receiving the stage text data based on the audio-to-text display service;

[0024] Based on the audio-to-text display service, using the stage text data, performing display content addition processing on the display content of the audio-to-text display service to determine target display content;

[0025] The determining of text data based on the stage text data comprises:

[0026] Text data is determined based on the target display content.

[0027] Optionally, the performing text conversion processing on the current stage audio data to obtain stage text data includes:

[0028] Performing voice extraction processing on the current stage audio data to obtain voice audio data;

[0029] Performing voice user division processing on the human voice audio data to determine the voice user corresponding to each human voice audio sub-data in the human voice audio data;

[0030] The human voice audio sub-data corresponding to each voice user is processed into text to obtain user text data corresponding to each voice user, and the stage text data is determined based on the user text data corresponding to each voice user.

[0031] In a second aspect, an embodiment of the present application provides a device for generating meeting minutes, the device comprising:

[0032] A processing module, adapted to obtain conference data for generating conference minutes, and extract characteristic information from the conference data to obtain characteristic text data;

[0033] A construction module, adapted to perform a conference minutes prompt word template construction process on the characteristic text data, obtain a specific conference minutes prompt word template corresponding to the conference data, and construct a conference minutes prompt word through the specific conference minutes prompt word template and the characteristic text data;

[0034] The meeting minutes generation module is adapted to input the meeting minutes prompt words into the meeting minutes generation big model, and output target meeting minutes for the meeting minutes prompt words based on the meeting minutes generation big model.

[0035] In a third aspect, an embodiment of the present application provides a terminal, the terminal comprising:

[0036] Processor; and

[0037] A memory arranged to store computer executable instructions, which when executed cause the processor to perform any of the methods described above.

[0038] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores one or more programs, and when the one or more programs are executed by a processor, implement any of the methods described above.

[0039] The beneficial effects brought by the technical solutions provided in some embodiments of the present application include at least: first, obtaining the meeting data for the meeting minutes to be generated, then extracting the feature text data corresponding to the meeting data, and then realizing the preliminary filtering of the redundant information corresponding to the meeting data, and then constructing and processing the meeting minutes prompt word template for the feature text data, so that a dedicated prompt word template can be constructed for the feature text data, that is, a specific meeting minutes prompt word template, so that the meeting minutes generation large model can fully understand the feature information corresponding to the learning feature text data for the meeting minutes prompt word template, and make the meeting minutes generation large model for the specific meeting minutes prompt word template more targeted to the feature text data, and then construct the meeting minutes prompt word through the specific meeting minutes prompt word template and the feature text data, input the meeting minutes prompt word into the meeting minutes generation large model, process the feature text data based on the indication information corresponding to the specific meeting minutes prompt word template in the meeting minutes prompt word, and output the target meeting minutes for the meeting minutes prompt word. Thereby solving the technical problem that the generation efficiency of the meeting minutes generated by manual recording is low and it takes a lot of time. Moreover, the meeting minutes generation model in the present application is more targeted when generating meeting minutes based on a specific meeting minutes prompt word template for feature text data, which is different from the related art in which the meeting minutes generation process is generated using the same template. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 An exemplary system architecture diagram of a method for generating meeting minutes provided in an embodiment of the present application;

[0042] Figure 2 A flowchart of a method for generating meeting minutes provided in an embodiment of the present application;

[0043] Figure 3 A schematic diagram of a process for determining a specific meeting minutes prompt word template provided in an embodiment of the present application;

[0044] Figure 4 A schematic diagram of a process for determining the attributes of meeting points provided in an embodiment of the present application;

[0045] Figure 5 A schematic diagram of a process for generating a specific meeting minutes prompt word template provided in an embodiment of the present application;

[0046] Figure 6 A schematic diagram of an interface of a basic template for a meeting minutes reminder provided in an embodiment of the present application;

[0047] Figure 7 A schematic diagram of a process for obtaining text data provided in an embodiment of the present application;

[0048] Figure 8 A schematic diagram of a process for determining text data in a stage provided in an embodiment of the present application;

[0049] Fig. 9 A schematic diagram of the structure of a meeting minutes generating device provided in an embodiment of the present application;

[0050] Fig.10 A schematic diagram of the structure of a terminal provided in an embodiment of the present application. DETAILED DESCRIPTION

[0051] In order to make the features and advantages of the embodiments of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of them. Based on the embodiments in the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the embodiments of the present application.

[0052] In related technologies, with the rapid development of the Internet and communication technologies, conferences that use voice communication over the Internet or telephone lines are becoming more and more common, such as audio conferences and video conferences. However, since a large amount of information in the process of voice communication needs to be recorded and organized, it is necessary to manually process and generate meeting minutes corresponding to the voice communication. However, the generation of traditional meeting minutes usually requires a lot of manpower and time, which is inefficient. Important information in the meeting may be omitted or misunderstood, resulting in a decrease in the accuracy of the meeting minutes. In addition, traditional speech-to-text technology may have difficulties in processing multi-person audio conferences, such as being unable to distinguish the voices of multiple speakers, and traditional meeting minutes are often just a simple listing of the content of the meeting, lacking the ability to understand semantics and generate summaries.

[0053] In order to solve the related technical problems, the embodiment of the present application provides a method for generating meeting minutes, which includes: obtaining meeting data for generating meeting minutes, performing feature information extraction processing on the meeting data to obtain feature text data; performing meeting minutes prompt word template construction processing on the feature text data to obtain a specific meeting minutes prompt word template corresponding to the meeting data, and constructing meeting minutes prompt words through the specific meeting minutes prompt word template and the feature text data; inputting the meeting minutes prompt words into the meeting minutes generation big model, and outputting the target meeting minutes for the meeting minutes prompt words based on the meeting minutes generation big model. The artificial intelligence big model corresponding to the meeting minutes generation big model assists users in intelligent office work, and then generates the corresponding target meeting minutes, thereby solving the technical problem that the generation efficiency of meeting minutes generated by manual recording is low and requires a lot of time.

[0054] See also Figure 1 , Figure 1 An exemplary system architecture diagram of a method for generating meeting minutes provided in an embodiment of the present application.

[0055] like Figure 1 As shown, the system architecture may include a terminal 101, a network 102, and a server 103. The network 102 is used to provide a medium for a communication link between the terminal 101 and the server 103. The network 102 may include various types of wired communication links or wireless communication links, for example, the wired communication link includes an optical fiber, a twisted pair, or a coaxial cable, and the wireless communication link includes a Bluetooth communication link, a Wi-Fi (Wireless-Fidelity) communication link, or a microwave communication link.

[0056] The terminal 101 can interact with the server 103 through the network 102 to receive messages from the server 103 or send messages to the server 103, or the terminal 101 can interact with the server 103 through the network 102 to receive messages or data sent by other users to the server 103. The terminal 101 can be hardware or software. When the terminal 101 is hardware, it can be various terminals, including but not limited to smart watches, smart phones, tablet computers, laptop portable computers and desktop computers. When the terminal 101 is software, it can be installed in the terminals listed above, which can be implemented as multiple software or software modules (for example: used to provide distributed services), or it can be implemented as a single software or software module, which is not specifically limited here.

[0057] The server 103 may be a business server that provides various services. It should be noted that the server 103 may be hardware or software. When the server 103 is hardware, it may be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the server 103 is software, it may be implemented as multiple software or software modules (for example, for providing distributed services), or as a single software or software module, which is not specifically limited here.

[0058] In an embodiment of the present application, terminal 101 can obtain meeting data for meeting minutes to be generated, perform feature information extraction processing on the meeting data to obtain feature text data; perform meeting minutes prompt word template construction processing on the feature text data to obtain a specific meeting minutes prompt word template corresponding to the meeting data, and construct meeting minutes prompt words through the specific meeting minutes prompt word template and the feature text data; input the meeting minutes prompt words into a meeting minutes generation big model, and output a target meeting minutes for the meeting minutes prompt words based on the meeting minutes generation big model.

[0059] It should be understood that the number of the above terminals, networks and servers is only illustrative, and any number of terminals, networks and servers may be used according to implementation requirements. Of course, in an embodiment provided in the present application, the system architecture may not include a server.

[0060] See also Figure 2 , Figure 2 A flowchart of a method for generating meeting minutes provided in an embodiment of the present application. The execution subject of the embodiment of the present application can be a terminal that executes the method for generating meeting minutes, or a processor in the terminal that executes the method for generating meeting minutes, or a meeting minutes generation service in the terminal that executes the method for generating meeting minutes. For the convenience of description, the specific execution process of the method for generating meeting minutes is introduced below by taking the execution subject being a processor in a terminal as an example.

[0061] like Figure 2 As shown, the method for generating meeting minutes includes:

[0062] S202: Acquire meeting data for generating meeting minutes, and perform feature information extraction on the meeting data to obtain feature text data.

[0063] Among them, the meeting data for generating the meeting minutes may include audio data, and the meeting data may be audio data generated by real-time recording during the meeting, or audio data generated by recording a historical meeting that has ended. Whether it is an audio file recorded in real time on site or an audio file generated by recording a meeting that has been completed in the past, it can be used as meeting data. Of course, the meeting data for generating the meeting minutes may also include text data recorded by the user for the meeting. In this case, text conversion processing of the audio data corresponding to the meeting can be avoided, simplifying the process.

[0064] When the conference data includes audio data, feature information extraction processing is performed on the conference data in S202 to obtain feature text data, which may be:

[0065] The audio data is subjected to text conversion processing, so that the audio information corresponding to the audio data is converted into text information, and then the timing information of the text information corresponding to the audio information is determined based on the timing information of the audio information, and finally the text data corresponding to the audio data is determined based on the text information and the timing information corresponding to the text information. Here, the timing information refers to time sequence information, which can be specifically represented by a timestamp. After the text data is determined, the text data is subjected to text feature information extraction processing to obtain feature text data. Specifically, the text data can be subjected to text feature information extraction processing through a large model for text feature information extraction. The text feature information extraction processing is used to perform item clustering and item information association on each text information in the text data, and then the associated text information is subjected to information fusion, so as to obtain feature text data corresponding to the text data. The large model for text feature information extraction can be obtained after model training based on the basic large language model.

[0066] Exemplarily, based on the text feature information extraction model, the item labels corresponding to each text information in the text data are determined, and then the items in the text data are clustered based on the item labels to obtain the text information corresponding to each item label. The text information corresponding to the item label can be understood as the key information, key content and topic corresponding to the item. Specifically, the key information may include but is not limited to the executor, item execution time and item execution location corresponding to the item, the key content includes but is not limited to the item execution content and the related information of the item, and the topic includes but is not limited to the item name corresponding to the item. Then, the text information corresponding to the item label is processed for item information association, so as to splice the item global information of the item corresponding to the item label based on the scattered information of the item corresponding to the item label.

[0067] After that, the item association information of the items corresponding to each item label is determined. For example, if item A is the current arrangement for the target event and item B is the future plan for the target event, then item A and item B are related to each other in terms of time arrangement based on the target event. After determining the item association information of the items corresponding to each item label, the item global information corresponding to each item label is associated based on the item association information, thereby obtaining feature text data corresponding to the text data.

[0068] When the conference data includes text data, the audio data can be avoided from being converted into text, thereby simplifying the processing flow. In S202, feature information extraction is performed on the conference data to obtain feature text data, which can be specifically:

[0069] The feature text data is obtained by extracting text feature information from the text data. Specifically, the text feature information extraction large model can be used to extract text feature information from the text data. The text feature information extraction process is used to cluster the items in the text data and associate the item information. The associated text information is then fused to obtain the feature text data corresponding to the text data. The text feature information extraction large model can be obtained after model training based on the basic large language model.

[0070] S204: Performing conference minutes prompt word template construction processing on the characteristic text data to obtain a specific conference minutes prompt word template corresponding to the conference data, and constructing conference minutes prompt words through the specific conference minutes prompt word template and the characteristic text data.

[0071] Among them, after determining the characteristic text data, an exclusive prompt word template can be constructed for the characteristic text data. In actual scenarios, the audio data corresponding to different meetings involve different subdivisions and the focus of different audio data is also different. Therefore, in order to accurately generate meeting minutes for different audio data, it is necessary to construct an exclusive prompt word template for the characteristic text data, that is, a specific meeting minutes prompt word template. It should be understood that different meeting data can correspond to different specific meeting minutes prompt word templates, and the meeting data corresponds to the specific meeting minutes prompt word template one by one.

[0072] Exemplarily, the meeting key points attributes may be extracted from the feature text data, that is, the meeting key points attributes contained in the feature text data may be determined. The meeting key points attributes may refer to the key attributes corresponding to the feature text data. The meeting key points attributes may include general meeting key points attributes and meeting-specific key points attributes.

[0073] The general key attributes of a meeting refer to the key attributes that are generally applicable. When determining the general key attributes of a meeting corresponding to the feature text data, the general attribute matching processing can be performed on the general key attribute database of the meeting based on the feature text data, so as to determine the general key attributes of the meeting corresponding to the feature text data. The general key attributes of a meeting in the general key attribute database of the meeting include but are not limited to the basic information attributes of the meeting, the purpose and agenda attributes of the meeting, the main discussion content attributes, the resolution and action item attributes, the important information and key points attributes, the opinions and suggestions attributes of the participants, the next step plan attributes, etc.

[0074] The conference-specific key point attributes may refer to unique or specific key attributes. When determining the conference-specific key point attributes corresponding to the feature text data, the conference-specific key point attribute database may be subjected to a unique attribute matching process based on the feature text data, thereby determining the conference-specific key point attributes corresponding to the feature text data. The conference-specific key point attributes in the conference-specific key point attribute database include, but are not limited to, domain attributes, scenario attributes, etc. Domain attributes include, but are not limited to, scientific research domain attributes, education domain attributes, and financial domain attributes, etc. Scenario attributes include, but are not limited to, specific scenario attributes corresponding to the feature text data, such as software development scenario attributes, software testing scenario attributes, market launch scenario attributes, etc.

[0075] After determining the attributes of the meeting points, a specific meeting minutes prompt word template is generated for the meeting points attributes, so that different feature text data can generate corresponding specific meeting minutes prompt word templates in a targeted manner, and then the meeting minutes generation model can fully understand the text information corresponding to the feature text data for the specific meeting minutes prompt word template, and make the meeting minutes generation model more targeted in processing the feature text data for the specific meeting minutes prompt word template.

[0076] After determining the specific meeting minutes prompt word template, the specific meeting minutes prompt word template and the feature text data can be processed by prompt word splicing, so as to obtain the meeting minutes prompt word for the feature text data. Here, since the meeting minutes prompt word is obtained by extracting feature information from the text data corresponding to the audio data, constructing the specific meeting minutes prompt word template, and splicing the prompt word, the length of the meeting minutes prompt word is greatly reduced, and can be smoothly input into the meeting minutes generation large model without exceeding the length limit range of the prompt word received by the meeting minutes generation large model.

[0077] In addition, after determining the prompt words for the meeting minutes, the prompt words for the meeting minutes can be edited in response to the prompt word editing operation of the user, so as to edit and update the prompt words for the meeting minutes based on the needs of the user. The editing process includes but is not limited to adding, deleting and changing.

[0078] S206: Input the meeting minutes prompt words into the meeting minutes generation big model, and output the target meeting minutes for the audio data based on the meeting minutes generation big model.

[0079] Among them, after determining the meeting minutes prompt words, the meeting minutes prompt words are input into the meeting minutes generation large model, and the meeting minutes generation large model is obtained through meeting minutes generation training based on the basic LLM (Large Language Model). After receiving the meeting minutes prompt words, the meeting minutes generation large model processes the feature text data based on the indication information corresponding to the specific meeting minutes prompt word template in the meeting minutes prompt words, thereby generating the target meeting minutes for the audio data, and then outputs the target meeting minutes for the audio data based on the meeting minutes generation large model.

[0080] The embodiment provided by the present application provides a method for generating meeting minutes. The method first obtains the meeting data of the meeting minutes, then extracts the feature text data corresponding to the meeting data, and then realizes the preliminary filtering of the redundant information corresponding to the meeting data. Then, the meeting minutes prompt word template is constructed and processed for the feature text data, so that an exclusive prompt word template can be constructed for the feature text data, that is, a specific meeting minutes prompt word template, so that the meeting minutes generation large model can fully understand the feature information corresponding to the learning feature text data based on the specific meeting minutes prompt word template, and make the meeting minutes generation large model based on the specific meeting minutes prompt word template more targeted in processing the feature text data, then the meeting minutes prompt words are constructed through the specific meeting minutes prompt word template and the feature text data, the meeting minutes prompt words are input into the meeting minutes generation large model, the feature text data is processed based on the indication information corresponding to the specific meeting minutes prompt word template in the meeting minutes prompt words, and the target meeting minutes for the meeting minutes prompt words are output. Thus, the technical problem that the generation efficiency of meeting minutes generated by manual recording is low and a lot of time is consumed is solved. Moreover, the large model for generating meeting minutes in the present application is more targeted when generating meeting minutes for feature text data based on a specific meeting minutes prompt word template, which is different from the related art in which the meeting minutes generation process is generated using the same template.

[0081] See also Figure 3 , Figure 3 A flow chart of determining a specific meeting minutes prompt word template provided in an embodiment of the present application. Figure 3 As shown, in S204, the conference minutes prompt word template is constructed for the characteristic text data to obtain a specific conference minutes prompt word template corresponding to the conference data, including:

[0082] S302: Determine the attributes of meeting points corresponding to the feature text data.

[0083] The meeting key point attribute may refer to a key attribute corresponding to the feature text data.

[0084] Exemplarily, the feature text data may be input into a meeting key point attribute determination model, and the meeting key point attribute determination model may output the meeting key point attribute corresponding to the feature text data.

[0085] Among them, the meeting key points attribute determination model can be obtained based on the initial basic large language model through meeting key points attribute training. Specifically, sample feature text data is obtained, and the sample feature text data carries the corresponding sample meeting key points attributes. The sample feature text data is input into the initial basic large language model, and the sample feature text data is processed for meeting key points attribute recognition based on the initial basic large language model to obtain the reference meeting key points attributes corresponding to the sample feature text data. Then, a target model loss function is constructed based on a first attribute parameter corresponding to the reference meeting key points attribute and a second attribute parameter corresponding to the sample meeting key points attribute. Then, a first numerical value corresponding to the reference meeting key points attribute and a second numerical value corresponding to the sample meeting key points attribute are substituted into the target model loss function to obtain a target loss value. Then, the initial basic large language model is trained based on the target loss value until the model training is completed to obtain the meeting key points attribute determination model.

[0086] Exemplarily, the meeting key point attribute matching processing may be performed on the feature text data based on the meeting key point attribute database to obtain the meeting key point attributes corresponding to the feature text data.

[0087] Among them, the characteristic text data can be clustered to determine the sub-item information of each item corresponding to the characteristic text data, and the sub-item information can be the information corresponding to the sub-item divided by time, component, structure or level, etc. Then, the sub-item information of each item is attributed to obtain the sub-item attributes, and the attribute retrieval of the meeting key points attribute database is performed based on the sub-item attributes to retrieve and determine the meeting key points attributes with a similarity greater than a preset similarity, so as to obtain the meeting key points attributes corresponding to the characteristic text data.

[0088] S304: Generate a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attributes.

[0089] The corresponding description text can be generated for the meeting key point attributes, and then the description texts corresponding to the meeting key point attributes are spliced ​​to determine the specific meeting minutes prompt word template corresponding to the meeting data. Specifically, the description text corresponding to the meeting key point attribute is used to describe the meeting minutes requirement specification of the meeting key point attribute.

[0090] In the embodiment provided in the present application, by determining the meeting key points attributes corresponding to the characteristic text data, and then generating a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attributes, the specific meeting minutes prompt word template includes the meeting minutes requirement specifications corresponding to the meeting key points attributes, thereby making the meeting minutes generation model more targeted when generating meeting minutes for the characteristic text data based on the specific meeting minutes prompt word template.

[0091] See also Figure 4 , Figure 4 A flow chart of determining the attributes of meeting points provided in an embodiment of the present application. Figure 4 As shown, the attributes of the meeting points corresponding to the characteristic text data are determined in S302, including:

[0092] S402: Determine the meeting key points element corresponding to the feature text data, extract common attributes for the meeting key points element, and obtain meeting common key points attributes.

[0093] Among them, the key elements of the meeting include but are not limited to meeting objectives, agenda, participants, materials, decisions, action plans, time management, etc.

[0094] Exemplarily, after extracting the meeting key points elements, the element association information between the meeting key points elements is determined, and then the general meeting key points attributes corresponding to the meeting key points elements are determined based on the meeting key points elements and the element association information. For example, based on the element association information, it is determined that the meeting key points elements belong to one or more of the following: basic meeting information attributes, meeting purpose and agenda attributes, main discussion content attributes, resolution and action item attributes, important information and key points attributes, participant opinions and suggestions attributes, next step plan attributes, etc.

[0095] Exemplarily, after extracting the meeting key points element, the meeting general key points attribute corresponding to the meeting key points element may also be determined based on the meeting general key points attribute model.

[0096] The general meeting key points attribute model can be obtained through the general meeting key points attribute training based on the first initial basic large language model. Specifically, a sample meeting key points element is obtained, and the sample meeting key points element carries the corresponding sample meeting key points attribute. The sample meeting key points element is input into the first initial basic large language model, and the sample meeting key points attribute recognition processing is performed on the sample meeting key points element based on the first initial basic large language model to obtain the reference meeting key points attribute corresponding to the sample meeting key points element. Then, a first target model loss function is constructed based on the third attribute parameter corresponding to the reference meeting key points attribute and the fourth attribute parameter corresponding to the sample meeting key points attribute. Then, the third numerical value corresponding to the reference meeting key points attribute and the fourth numerical value corresponding to the sample meeting key points attribute are substituted into the first target model loss function to obtain a first target loss value. Then, model training is performed on the first initial basic large language model based on the first target loss value until the model training is completed to obtain the general meeting key points attribute model.

[0097] Of course, it is also possible to perform attribute matching on the conference key point elements based on the conference general key point attribute database to determine the conference general key point attributes corresponding to the conference key point elements.

[0098] Among them, the meeting key points elements can be clustered to determine the clustering information of each item corresponding to the meeting key points elements. The clustering information can be information that the meeting key points elements are divided according to time, component, structure or level. Then, the clustering information of each item is divided by attributes to obtain clustering attributes, and attribute retrieval is performed on the general meeting key points attribute database based on the clustering attributes, and the similarity between the clustering attributes and the general meeting key points attributes in the general meeting key points attribute database is calculated. The general meeting key points attributes with a similarity greater than a preset similarity are retrieved to determine that the general meeting key points attributes corresponding to the meeting key points elements are obtained. Here, the clustering attributes can be vectorized to represent the clustering attribute vector corresponding to the clustering attributes, and the general meeting key points attributes in the general meeting key points attribute database can be represented by the general meeting key points attribute vector. When calculating the similarity between the clustering attributes and the general meeting key points attributes in the general meeting key points attribute database, the vector similarity between the clustering attribute vector and the general meeting key points attribute vector in the general meeting key points attribute database can be calculated, and the similarity between the clustering attributes and the general meeting key points attributes in the general meeting key points attribute database can be represented by the vector similarity.

[0099] S404: Extracting proprietary attributes of the meeting key points elements to obtain proprietary attributes of the meeting key points.

[0100] Exemplarily, after extracting the meeting key points elements, element association information between the meeting key points elements is determined, and then one or more of the domain attributes, scene attributes, etc. corresponding to the meeting key points elements are determined based on the meeting key points elements and the element association information.

[0101] The meeting-specific key points attribute model can be obtained by training the second initial basic large language model with the meeting-specific key points attributes. Specifically, a sample meeting key points element is obtained, and the sample meeting key points element carries the corresponding sample meeting-specific key points attributes. The sample meeting key points element is input into the second initial basic large language model, and the sample meeting key points element is identified with the meeting-specific key points attributes based on the second initial basic large language model to obtain the reference meeting-specific key points attributes corresponding to the sample meeting key points element. Then, a second target model loss function is constructed based on the fifth attribute parameter corresponding to the reference meeting-specific key points attribute and the sixth attribute parameter corresponding to the sample meeting-specific key points attribute. Then, the fifth numerical value corresponding to the reference meeting-specific key points attribute and the sixth numerical value corresponding to the sample meeting-specific key points attribute are substituted into the second target model loss function to obtain the second target loss value. Then, the second initial basic large language model is trained based on the second target loss value until the model training is completed to obtain the meeting-specific key points attribute model.

[0102] Of course, it is also possible to perform attribute matching on the conference key point elements based on the conference-specific key point attribute database to determine the conference-specific key point attributes corresponding to the conference key point elements.

[0103] Among them, the meeting key points elements can be clustered to determine the clustering information of each item corresponding to the meeting key points elements. The clustering information can be information that divides the meeting key points elements according to time, components, structure or level. Then, the clustering information of each item is divided into attributes to obtain clustering attributes, and attribute retrieval is performed on the meeting-specific key points attribute database based on the clustering attributes. The similarity between the clustering attributes and the meeting-specific key points attributes in the meeting-specific key points attribute database is calculated, and the meeting-specific key points attributes with a similarity greater than a preset similarity are retrieved to determine, thereby obtaining the meeting-specific key points attributes corresponding to the meeting key points elements. Here, the clustering attributes can be vectorized to determine the clustering attribute vector corresponding to the clustering attributes, and the meeting-specific key points attributes in the meeting-specific key points attribute database can be represented by the meeting-specific key points attribute vector. When calculating the similarity between the clustering attributes and the conference-specific key point attributes in the conference-specific key point attribute database, the vector similarity between the clustering attribute vector and the conference-specific key point attribute vector in the conference-specific key point attribute database can be calculated, and the similarity between the clustering attributes and the conference-specific key point attributes in the conference-specific key point attribute database can be characterized by the vector similarity.

[0104] S406: Determine the attributes of the meeting key points based on the general meeting key point attributes and the specific meeting key point attributes.

[0105] When determining the general key point attributes and the specific key point attributes of the meeting, the general key point attributes and the specific key point attributes of the meeting may be concatenated to obtain the key point attributes of the meeting.

[0106] In the embodiment provided in the present application, the general attributes of the meeting key points are extracted from the general attributes of the meeting key points elements to obtain the general attributes of the meeting, and the specific attributes of the meeting key points are extracted from the specific attributes of the meeting key points elements to obtain the specific attributes of the meeting. Finally, the general attributes of the meeting key points and the specific attributes of the meeting are used to determine the attributes of the meeting key points, so that the attributes of the meeting key points include both the general attributes of the meeting and the specific attributes of the meeting. In the process of generating a specific meeting minutes prompt word template based on the meeting key points attributes, the specific meeting minutes prompt word template can be generated in a targeted manner from the general dimension and the special (specific) dimension.

[0107] See also Figure 5 , Figure 5 A schematic diagram of a process for generating a specific meeting minutes prompt word template provided in an embodiment of the present application. Figure 6 , Figure 6 This is a schematic diagram of an interface of a basic template for a meeting minutes reminder provided in an embodiment of the present application. Figure 5 and Figure 6 As shown, in S304, a specific meeting minutes prompt word template corresponding to the meeting data is generated based on the meeting key points attributes, including:

[0108] S502: Generate a basic template for the reminder words of the meeting minutes based on the general key points attributes of the meeting.

[0109] Among them, after determining the general key points attributes of the meeting corresponding to the characteristic text data, the corresponding general description text is generated based on the general key points attributes of the meeting. The general description text corresponding to the general key points attributes of the meeting is used to describe the meeting minutes requirements and specifications of the general key points attributes of the meeting. Then, a basic template for meeting minutes prompt words is generated based on the general description texts of the general key points attributes of each meeting.

[0110] For example, the basic template prompt for the meeting minutes is as follows: Figure 6 As shown, the general key points attributes of the meeting corresponding to the feature text data at this time include basic meeting information attributes, meeting purpose and agenda attributes, main discussion content attributes, resolution and action item attributes, important information and key points attributes, participant opinions and suggestions attributes and next step plan attributes.

[0111] The general description text corresponding to the basic information attribute of the meeting can be "Basic information of the meeting: including the meeting theme, time, location, participants, etc." The general description text of the purpose and agenda attribute of the meeting can be "Purpose and agenda of the meeting: briefly explain the purpose of the meeting and list the various topics on the meeting agenda." The general description text of the main discussion content attribute can be "Main discussion content: record the main content and viewpoints of the meeting discussion. It can be summarized one by one in the order of the agenda, including discussions related to each topic, questions raised, solutions, etc." The general description text of the resolution and action item attribute can be "Resolutions and action items: record the decisions made in the meeting. The general description text of the important information and key points attribute can be "Important information and key points: Some important information and key points can be highlighted in the minutes so that readers can quickly understand the core content of the meeting." The general description text of the opinions and suggestions of the participants can be "Record the opinions and suggestions put forward by the participants, which can be highlighted in the minutes." The general description text of the next step plan attribute can be "Next step plan: Outline the next plan and actions after the meeting, including unresolved issues, topics that need further discussion, etc."

[0112] At the same time, a connection description can also be generated, such as "I need a meeting minutes that records the discussion content, decisions, action plans and participants of our most recent meeting. I hope the minutes can clearly summarize the content of the meeting and facilitate participants to review and follow up on work progress. The following is the completed text record of this meeting: {text}. Now please help me compile the minutes of this meeting, which must include the following table content:", where the text in {text} is used to fill in the feature text data. In addition, the general description text corresponding to the general key points attributes of each meeting is numbered.

[0113] S504: Generate a proprietary description text based on the conference-specific key point attributes, and perform description editing processing on the conference minutes prompt word basic template based on the proprietary description text to obtain a specific conference minutes prompt word template corresponding to the conference data.

[0114] Among them, after determining the meeting-specific key points attributes corresponding to the characteristic text data, the corresponding proprietary description text is generated based on the meeting-specific key points attributes. The proprietary description text corresponding to the meeting-specific key points attributes is used to describe the meeting minutes requirement specifications of the meeting-specific key points attributes. Then, based on the proprietary description texts of each meeting-specific key points attributes, the meeting minutes prompt word basic template is described and edited. For example, the meeting minutes prompt word basic template can be added or modified based on the proprietary description text.

[0115] Exemplarily, when the attributes of the conference-specific key points include domain attributes and scene attributes, when the domain attributes are specifically educational attributes, and the scene attributes are specifically video education software application scene attributes, the domain attributes and scene attributes can be associated to jointly generate corresponding proprietary description texts, such as the proprietary description text "proprietary meaning or proprietary advantage, outlining the proprietary advantages or proprietary meaning of this video education software in the corresponding scene in the education field", etc. At this time, the proprietary description text can be added to the basic template of the meeting minutes prompt words, and the corresponding general description words in the basic template of the meeting minutes prompt words can be replaced based on the proper nouns corresponding to the domain attributes or scene attributes, such as the participants can be replaced with the video education software application discussion personnel, so that the large model for generating meeting minutes can fully understand and learn the description corresponding to the specific meeting minutes prompt word template.

[0116] In the embodiment provided in the present application, a basic template for meeting minutes prompt words is generated using the general key points attributes of the meeting, and then a proprietary description text is generated using the specific key points attributes of the meeting. The basic template for meeting minutes prompt words is described and edited using the proprietary description text to obtain a specific meeting minutes prompt word template corresponding to the meeting data, thereby making it possible to generate a specific meeting minutes prompt word template for the feature text data from the general dimension and the dedicated (specific) dimension, so that the processing of the feature text data by the large meeting minutes generation model based on the specific meeting minutes prompt word template is more targeted.

[0117] See also Figure 7 , Figure 7 A schematic diagram of a process for obtaining text data provided in an embodiment of the present application. Figure 7 As shown, the conference data includes audio data. In S202, feature information extraction is performed on the conference data to obtain feature text data, including:

[0118] S702: Perform voice monitoring on the conference voice.

[0119] Among them, voice monitoring of conference voice is used to determine whether the conference voice is in the conference voice ongoing state or the conference voice ended state. For example, when the conference voice is conducted through conference software, conference voice ongoing information can be obtained from the conference software, so as to determine whether the conference voice is ended based on the conference voice ongoing information, and then judge whether the conference voice is in the conference voice ongoing state or the conference voice ended state.

[0120] For example, when the conference voice is conducted in the conference room, it can be determined based on keyword monitoring whether the conference voice is in the conference voice ongoing state or the conference voice ended state. For example, when the user voice is an associated description of the end of the meeting, and the user voice is not detected within a predetermined time after the associated description is monitored, it indicates that the conference voice is in the ended state; otherwise, it is in the conference voice ongoing state.

[0121] Of course, when performing voice monitoring on the conference voice, the user can determine that the current state is the end of the conference voice based on the physical button or virtual button of the conference voice end state; and the user can determine that the current state is the progress of the conference voice based on the physical button or virtual button of the conference voice progress state.

[0122] S704: When the result of voice monitoring is that the conference voice is in progress, real-time audio data is obtained from the conference data. When it is determined that the duration corresponding to the audio data is greater than or equal to the preset duration, the audio data is cached in the stage audio data to obtain the current stage audio data; the current stage audio data is converted into text to obtain the stage text data, and the step of performing voice monitoring on the conference voice is returned.

[0123] Among them, when it is monitored that the current state is conference voice, the audio is recorded in real time to generate conference data, and then the real-time audio data is obtained from the conference data. When the duration of the real-time audio data recorded is greater than or equal to the preset duration (such as 1 second), in order to reduce the computational pressure of text conversion processing and improve the accuracy of text recognition, the real-time audio data is cached in stages to obtain the current stage audio data. The duration corresponding to the current stage audio data is usually the preset duration.

[0124] Afterwards, the current stage audio data is converted into text to obtain the stage text data corresponding to the current stage audio data, and S702 is simultaneously executed: the step of performing voice monitoring on the conference voice.

[0125] S706: When the result of the voice monitoring is that the conference voice is in an ending state, the text data is determined based on the stage text data, and text feature extraction processing is performed on the text data to obtain feature text data.

[0126] When it is determined that the current state is the end of the conference voice, the text data is determined based on all the stage text data. For example, if the first stage text data is "Hello everyone", the second stage text data is "XXX", ..., and the Nth stage text data is "The meeting is over", then the text data determined based on the stage text data is "Hello everyone, XXX ... The meeting is over".

[0127] When performing text feature extraction processing on text data to obtain feature text data, the text feature information extraction processing can be performed on the text data through a large text feature information extraction model. The text feature information extraction processing is used to cluster items and associate item information on each text information in the text data, and then fuse the associated text information to obtain feature text data corresponding to the text data. The large text feature information extraction model can be obtained after model training based on the basic large language model.

[0128] In the embodiment provided in the present application, by performing voice monitoring on the conference voice, when the result of the voice monitoring is that the conference voice is in progress, the real-time audio data is divided into stages, thereby reducing the computational pressure of text conversion processing and improving the accuracy of text recognition. When it is determined that the current state is the end of the conference voice, the text data is determined based on the text data of all stages, and then the text data is processed for text feature extraction to obtain feature text data.

[0129] In an embodiment provided by the present application, after S704 performs text conversion processing on the current stage audio data to obtain stage text data, it also includes:

[0130] Sending the phase text data to the audio-to-text display service, and receiving the phase text data based on the audio-to-text display service;

[0131] Based on the audio-to-text display service, the display content of the audio-to-text display service is newly processed by using the stage text data to determine the target display content;

[0132] Determining text data based on the stage text data in S706 includes:

[0133] The text data is determined based on the target display content.

[0134] Among them, after determining the current corresponding stage text data, the stage text data is sent to the audio-to-text display service, and the audio-to-text display service is used to display the stage text data. The audio-to-text display service receives the stage text data and determines the current display content of the audio-to-text display service. After that, the audio-to-text display service uses the stage text data to add display content. For example, if the current display content is "abc" and the stage text data is "def", the target display content is "abcdef", thereby updating the display content, determining and displaying the target display content including the current display content.

[0135] At the same time, since all conference voices are converted into phased text data in batches when the conference voice ends, and all can be displayed and viewed through the audio-to-text display service, the text data corresponding to the audio data can be determined based on the target display content.

[0136] Exemplarily, when the user is detected speaking, a microphone-opening callback event is triggered, i.e., the execution of steps S702 to S706 is triggered. The execution of steps S702 to S706 may be specifically as follows: obtaining real-time audio data may be obtaining the user's audio stream, setting a corresponding audio stream cache based on the user's audio stream, and setting a 1-second timer. When the duration of the audio stream recorded in the audio stream cache is less than 1 second, continue recording until the duration of the audio stream recorded in the audio stream cache is greater than or equal to 1 second. Then, the audio stream is taken out from the audio stream cache, and the audio stream is encoded into WAV (Waveform Audio File Format) format. The WAV format can ensure the quality of the audio processing. Then, the audio stream in WAV format is converted into based64 format to facilitate audio-to-text processing. Then, the audio-to-text interface is called to call the audio-to-text service, so as to convert the audio stream in based64 format into stage text data. Then, the stage text data is sent to the audio-to-text display service. The audio-to-text display service displays the stage text data based on the chat box.

[0137] In the embodiment provided by the present application, the current display content is processed by using the stage text data based on the audio-to-text display service, so that the text data of each stage, that is, the target display content, can be continuously displayed. In addition, in the embodiment provided by the present application, regardless of whether the meeting minutes are generated, the target display content can be displayed and viewed based on the audio-to-text display service.

[0138] See also Figure 8 , Figure 8 A schematic diagram of a process for determining text data in a stage provided in an embodiment of the present application. Figure 8 As shown, in S704, the current stage audio data is converted into text to obtain stage text data, including:

[0139] S802: Perform voice extraction processing on the audio data of the current stage to obtain voice audio data.

[0140] Among them, the audio data in the current stage is processed for voice extraction, so as to filter out background noise and obtain clean voice audio data.

[0141] S804: Perform voice user classification processing on the human voice audio data to determine the voice user corresponding to each human voice audio sub-data in the human voice audio data.

[0142] Before performing voice user segmentation processing on the human voice audio data, it is confirmed whether to perform voice user segmentation processing on the human voice audio data. When it is determined that the voice user segmentation processing is performed on the human voice audio data, the audio of different speakers is distinguished by the human voice separation technology. The human voice separation adopts the pyannote technology for speaker diarization. Here, the pyannote technology is used for human voice separation, so as to determine the voice user corresponding to each human voice audio sub-data in the human voice audio data. When it is determined not to perform voice user segmentation processing on the human voice audio data, the whisper model (such as whisper-large-v3) is directly used for audio to text processing, and different voice users will not be distinguished at this time. The whisper model is a general speech recognition model. It is trained on a large-scale and diverse audio data set and can perform multi-task processing such as multilingual speech recognition, speech translation and language recognition.

[0143] S806: Perform audio-to-text processing on the human voice audio sub-data corresponding to each voice user to obtain user text data corresponding to each voice user, and determine the stage text data based on the user text data corresponding to each voice user.

[0144] Among them, when the human voice audio sub-data corresponding to each voice user is processed from audio to text, the audio of different voice users can be distinguished by voice recognition technology and human voice separation technology. The voice recognition here adopts the whisper model (such as whisper-large-v3) for audio to text processing, so as to obtain the user text data corresponding to each voice user, wherein each user text data also includes timing information. Then, the stage text data is determined based on the user text data corresponding to each voice user, and then the text corresponding to each user's speech is determined.

[0145] In the embodiments provided in the present application, human voice audio data is divided into voice users and audio-to-text is processed to determine user text data corresponding to each voice user, so that the user text data corresponding to each voice user can be determined based on the stage text data, and by distinguishing voice users, feature text data can be extracted from the text data more accurately.

[0146] An embodiment of the present application further provides a computer storage medium, which can store multiple instructions, and the instructions are suitable for being loaded by a processor and executing the steps of any method in the above embodiments.

[0147] See also Fig. 9 , Fig. 9This is a schematic diagram of the structure of a meeting minutes generation device provided in an embodiment of the present application. Fig. 9 As shown, the conference minutes generating device 900 includes:

[0148] The processing module 910 is adapted to obtain the meeting data for generating the meeting minutes, and perform feature information extraction processing on the meeting data to obtain feature text data;

[0149] A construction module 920 is adapted to construct a conference minutes prompt word template for the characteristic text data, obtain a specific conference minutes prompt word template corresponding to the conference data, and construct a conference minutes prompt word through the specific conference minutes prompt word template and the characteristic text data;

[0150] The meeting minutes generation module 930 is adapted to input the meeting minutes prompt words into the meeting minutes generation big model, and output the target meeting minutes for the meeting minutes prompt words based on the meeting minutes generation big model.

[0151] Optionally, the building block 920 includes:

[0152] A determination unit adapted to determine the attributes of meeting points corresponding to the characteristic text data;

[0153] The generating unit is adapted to generate a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attributes.

[0154] Optionally, the determining unit includes:

[0155] The first determination subunit is adapted to determine the meeting key points element corresponding to the characteristic text data, extract the general attributes of the meeting key points element, and obtain the general key points attributes of the meeting;

[0156] The extraction subunit is adapted to extract the specific attributes of the key points of the meeting and obtain the specific key points attributes of the meeting;

[0157] The second determining subunit is adapted to determine the attributes of the key points of the meeting based on the general key point attributes of the meeting and the specific key point attributes of the meeting.

[0158] Optionally, the generating unit includes:

[0159] A generating subunit, adapted to generate a basic template of meeting minutes prompt words based on the general key points attributes of the meeting;

[0160] The conference minutes prompt word template determination subunit is suitable for generating a proprietary description text based on the conference-specific key point attributes, and performing description editing processing on the conference minutes prompt word basic template based on the proprietary description text to obtain a specific conference minutes prompt word template corresponding to the conference data.

[0161] Optionally, the processing module 910 includes:

[0162] A monitoring unit, suitable for voice monitoring of conference voice;

[0163] The first judgment unit is adapted to obtain real-time audio data from the conference data when the result of the voice monitoring is that the conference voice is in progress, and when it is determined that the duration corresponding to the audio data is greater than or equal to the preset duration, perform stage audio data caching processing on the audio data to obtain current stage audio data; perform text conversion processing on the current stage audio data to obtain stage text data, and return to execute the step of performing voice monitoring on the conference voice;

[0164] The second judgment unit is adapted to determine the text data based on the stage text data when the result of the voice monitoring is that the conference voice is in an end state, and to perform text feature extraction processing on the text data to obtain feature text data.

[0165] Optionally, the first judging unit further includes:

[0166] The sending subunit is adapted to send the phase text data to the audio-to-text display service, and receive the phase text data based on the audio-to-text display service;

[0167] A display adding unit is adapted to perform display content adding processing on the display content of the audio-to-text display service based on the audio-to-text display service by using the stage text data to determine the target display content;

[0168] The second judgment unit is further adapted to determine text data based on the target display content.

[0169] Optionally, the first judging unit includes:

[0170] The human voice audio data extraction subunit is adapted to perform human voice extraction processing on the audio data of the current stage to obtain human voice audio data;

[0171] A voice user determination subunit, adapted to perform voice user division processing on the human voice audio data, and determine the voice user corresponding to each human voice audio sub-data in the human voice audio data;

[0172] The stage text data determination subunit is suitable for performing audio-to-text processing on the human voice audio sub-data corresponding to each voice user to obtain the user text data corresponding to each voice user, and determining the stage text data based on the user text data corresponding to each voice user.

[0173] See also Fig.10 , Fig.10 A schematic diagram of the structure of a terminal provided in an embodiment of the present application. Fig.10 As shown, the terminal 1000 may include: at least one processor 1001 , at least one network interface 1004 , a user interface 1003 , a memory 1005 , and at least one communication bus 1002 .

[0174] The communication bus 1002 is used to realize the connection and communication between these components.

[0175] The user interface 1003 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 1003 may also include a standard wired interface and a wireless interface.

[0176] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0177] Among them, the processor 1001 may include one or more processing cores. The processor 1001 uses various interfaces and lines to connect various parts in the entire terminal 1000, and executes various functions and processes data of the terminal 1000 by running or executing instructions, programs, code sets or instruction sets stored in the memory 1005, and calling data stored in the memory 1005. Optionally, the processor 1001 can be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 1001 can integrate one or a combination of CPU (Central Processing Unit), GPU (Graphics Processing Unit) and modem. Among them, the CPU mainly processes the operating system, user interface and application program; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above-mentioned modem may not be integrated into the processor 1001, but implemented separately through a chip.

[0178] Among them, the memory 1005 may include RAM (Random Access Memory) and may also include ROM (Read-Only Memory). Optionally, the memory 1005 includes a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned method embodiments, etc.; the data storage area may store data involved in the above-mentioned method embodiments, etc. The memory 1005 may optionally be at least one storage device located away from the aforementioned processor 1001. As Fig.10 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a meeting minutes generation program.

[0179] exist Fig.10 In the terminal 1000 shown, the user interface 1003 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 1001 can be used to call the meeting minutes generation program stored in the memory 1005, and specifically perform the following operations:

[0180] Acquire the meeting data for which the meeting minutes are to be generated, and perform feature information extraction on the meeting data to obtain feature text data;

[0181] Performing conference minutes prompt word template construction processing on the characteristic text data to obtain a specific conference minutes prompt word template corresponding to the conference data, and constructing conference minutes prompt words through the specific conference minutes prompt word template and the characteristic text data;

[0182] The meeting minutes prompt words are input into the meeting minutes generation big model, and the target meeting minutes for the meeting minutes prompt words are output based on the meeting minutes generation big model.

[0183] Optionally, the processor 1001 performs a conference minutes prompt word template construction process on the characteristic text data, and when obtaining a specific conference minutes prompt word template corresponding to the conference data, specifically performs:

[0184] Determine the meeting key point attributes corresponding to the feature text data;

[0185] Generate a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attributes.

[0186] Optionally, when the processor 1001 determines the attributes of meeting points corresponding to the characteristic text data, it specifically performs:

[0187] Determine the meeting key points elements corresponding to the feature text data, extract common attributes for the meeting key points elements, and obtain common meeting key points attributes;

[0188] Extract the proprietary attributes of the meeting key points elements to obtain the proprietary key points attributes of the meeting;

[0189] The meeting key point attributes are determined based on the meeting common key point attributes and the meeting specific key point attributes.

[0190] Optionally, when the processor 1001 generates a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attribute, the processor 1001 specifically executes:

[0191] Generate a basic template for meeting minutes reminder words based on the general key points attributes of the meeting;

[0192] A proprietary description text is generated based on the proprietary key point attributes of the meeting, and a description editing process is performed on the basic template of the meeting minutes prompt words based on the proprietary description text to obtain a specific meeting minutes prompt word template corresponding to the meeting data.

[0193] Optionally, the conference data includes audio data, and when the processor 1001 performs feature information extraction processing on the conference data to obtain feature text data, specifically:

[0194] Conduct voice monitoring of conference voice;

[0195] When the result of voice monitoring is that the conference voice is in progress, real-time audio data is obtained from the conference data, and when it is determined that the duration corresponding to the audio data is greater than or equal to the preset duration, the audio data is cached for the stage audio data to obtain the current stage audio data; the current stage audio data is converted into text to obtain the stage text data, and the step of performing voice monitoring on the conference voice is returned;

[0196] When the result of the conference voice monitoring is that the conference voice is in an end state, the text data is determined based on the stage text data, and text feature extraction processing is performed on the text data to obtain feature text data.

[0197] Optionally, after performing text conversion processing on the current stage audio data to obtain stage text data, the processor 1001 further specifically performs:

[0198] Sending the phase text data to the audio-to-text display service, and receiving the phase text data based on the audio-to-text display service;

[0199] Based on the audio-to-text display service, the display content of the audio-to-text display service is newly processed by using the stage text data to determine the target display content;

[0200] Determine text data based on the stage text data, including:

[0201] The text data is determined based on the target display content.

[0202] Optionally, when the processor 1001 performs text conversion processing on the current stage audio data to obtain stage text data, specifically executes:

[0203] Performing voice extraction processing on the audio data of the current stage to obtain voice audio data;

[0204] Performing voice user division processing on the human voice audio data to determine the voice user corresponding to each human voice audio sub-data in the human voice audio data;

[0205] The human voice audio sub-data corresponding to each voice user is processed into text to obtain user text data corresponding to each voice user, and the stage text data is determined based on the user text data corresponding to each voice user.

[0206] In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0207] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0208] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules.

[0209] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0210] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of the present application.

[0211] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0212] The above is a description of a meeting minutes generation method, device, terminal and computer-readable storage medium provided in an embodiment of the present application. For technical personnel in this field, according to the ideas of the embodiments of the present application, there may be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the embodiments of the present application.

Claims

1. A method for generating meeting minutes, wherein: The method comprises: Acquire meeting data for which meeting minutes are to be generated, and perform feature information extraction processing on the meeting data to obtain feature text data; The conference data includes audio data, and the feature information extraction process for the conference data to obtain feature text data includes: Conduct voice monitoring of conference voice; When the result of the voice monitoring is that the conference voice is in progress, real-time audio data is obtained from the conference data, and when it is determined that the duration corresponding to the audio data is greater than or equal to the preset duration, the audio data is subjected to stage audio data caching processing to obtain current stage audio data; the current stage audio data is subjected to text conversion processing to obtain stage text data, and the step of performing voice monitoring on the conference voice is returned to be executed; When the result of the voice monitoring is that the conference voice is in an ending state, the text data is determined based on the stage text data, a large model is extracted based on the text feature information, the matter label corresponding to each text information in the text data is determined, and the text information in the text data is clustered based on the matter label to obtain the text information corresponding to each matter label, the text information corresponding to the matter label is associated with the matter information, and the associated information is fused to obtain the feature text data; Performing a conference minutes prompt word template construction process on the characteristic text data to obtain a specific conference minutes prompt word template corresponding to the conference data, and constructing a conference minutes prompt word through the specific conference minutes prompt word template and the characteristic text data; Inputting the meeting minutes prompt words into the meeting minutes generation big model, processing the feature text data based on the meeting minutes generation big model using the indication information corresponding to the specific meeting minutes prompt word template in the meeting minutes prompt words to obtain the target meeting minutes, and outputting the target meeting minutes for the meeting minutes prompt words based on the meeting minutes generation big model; The step of constructing a conference minutes prompt word template for the characteristic text data to obtain a specific conference minutes prompt word template corresponding to the conference data includes: Determine the meeting key points attribute corresponding to the characteristic text data; The step of determining the attributes of the meeting points corresponding to the characteristic text data includes: Determine the meeting key points elements corresponding to the characteristic text data, perform clustering processing on the meeting key points elements, determine clustering information of each item corresponding to the meeting key points elements, perform attribute division on the clustering information of each item, and obtain clustering attributes; Performing attribute retrieval on a general meeting key point attribute database based on the cluster attribute, calculating the similarity between the cluster attribute and the general meeting key point attribute in the general meeting key point attribute database, and retrieving and determining the general meeting key point attribute whose similarity is greater than a preset similarity; Performing attribute retrieval on the conference-specific key point attribute database based on the cluster attribute, calculating the similarity between the cluster attribute and the conference-specific key point attribute in the conference-specific key point attribute database, and retrieving and determining the conference-specific key point attribute whose similarity is greater than a preset similarity; Determining meeting key point attributes based on the meeting general key point attributes and the meeting specific key point attributes; Generate a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attributes; The step of generating a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attributes includes: Generate corresponding general description text based on the general key point attributes of the meeting, the general description text is used to describe the meeting minutes requirement specification of the general key point attributes of the meeting, and generate a basic template for the meeting minutes prompt words based on the general description texts of the general key point attributes of each meeting; Based on the meeting-specific key attributes, a corresponding proprietary description text is generated, and the proprietary description text is used to describe the meeting minutes requirement specifications of the meeting-specific key attributes. Based on the proprietary description text of each meeting-specific key attribute, the meeting minutes prompt word basic template is described and edited, and the description editing process includes adding or modifying, so as to obtain a specific meeting minutes prompt word template corresponding to the meeting data.

2. The method according to claim 1, wherein: After the text conversion process is performed on the current stage audio data to obtain the stage text data, the method further includes: Sending the stage text data to the audio-to-text display service, and receiving the stage text data based on the audio-to-text display service; Based on the audio-to-text display service, using the stage text data, performing display content addition processing on the display content of the audio-to-text display service to determine target display content; The determining of text data based on the stage text data comprises: Text data is determined based on the target display content.

3. The method according to claim 1, wherein: The step of converting the current stage audio data into text to obtain stage text data includes: Performing voice extraction processing on the current stage audio data to obtain voice audio data; Performing voice user division processing on the human voice audio data to determine the voice user corresponding to each human voice audio sub-data in the human voice audio data; The human voice audio sub-data corresponding to each voice user is processed into text to obtain user text data corresponding to each voice user, and the stage text data is determined based on the user text data corresponding to each voice user.

4. A device for generating meeting minutes, wherein: The meeting minutes generating device comprises: A processing module, adapted to obtain conference data for generating conference minutes, and extract characteristic information from the conference data to obtain characteristic text data; Wherein, the processing module includes: A monitoring unit, suitable for voice monitoring of conference voice; The first judgment unit is adapted to, when the result of the voice monitoring is that the conference voice is in progress, obtain real-time audio data from the conference data, and when it is determined that the duration corresponding to the audio data is greater than or equal to the preset duration, perform stage audio data caching processing on the audio data to obtain current stage audio data; perform text conversion processing on the current stage audio data to obtain stage text data, and return to execute the step of performing voice monitoring on the conference voice; The second judgment unit is adapted to determine the text data based on the stage text data when the result of the voice monitoring is the conference voice end state, extract a large model based on the text feature information, determine the matter label corresponding to each text information in the text data, and perform matter clustering on each text information in the text data based on the matter label to obtain the text information corresponding to each matter label, perform matter information association processing on the text information corresponding to the matter label, and perform information fusion on the associated information to obtain the feature text data; A construction module, adapted to perform a conference minutes prompt word template construction process on the characteristic text data, obtain a specific conference minutes prompt word template corresponding to the conference data, and construct a conference minutes prompt word through the specific conference minutes prompt word template and the characteristic text data; A meeting minutes generation module, adapted to input the meeting minutes prompt words into the meeting minutes generation macromodel, process the feature text data based on the meeting minutes generation macromodel using the indication information corresponding to the specific meeting minutes prompt word template in the meeting minutes prompt words to obtain target meeting minutes, and output the target meeting minutes for the meeting minutes prompt words based on the meeting minutes generation macromodel; Wherein, the building blocks include: A determination unit, adapted to determine the attributes of meeting points corresponding to the characteristic text data; The step of determining the attributes of the meeting points corresponding to the characteristic text data includes: Determine the meeting key points elements corresponding to the characteristic text data, perform clustering processing on the meeting key points elements, determine clustering information of each item corresponding to the meeting key points elements, perform attribute division on the clustering information of each item, and obtain clustering attributes; Performing attribute retrieval on a general meeting key point attribute database based on the cluster attribute, calculating the similarity between the cluster attribute and the general meeting key point attribute in the general meeting key point attribute database, and retrieving and determining the general meeting key point attribute whose similarity is greater than a preset similarity; Performing attribute retrieval on the conference-specific key point attribute database based on the cluster attribute, calculating the similarity between the cluster attribute and the conference-specific key point attribute in the conference-specific key point attribute database, and retrieving and determining the conference-specific key point attribute whose similarity is greater than a preset similarity; Determining meeting key point attributes based on the meeting general key point attributes and the meeting specific key point attributes; A generating unit, adapted to generate a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attribute; The step of generating a specific meeting minutes prompt word template corresponding to the meeting data based on the meeting key points attributes includes: Generate corresponding general description text based on the general key point attributes of the meeting, the general description text is used to describe the meeting minutes requirement specification of the general key point attributes of the meeting, and generate a basic template for the meeting minutes prompt words based on the general description texts of the general key point attributes of each meeting; Based on the meeting-specific key attributes, a corresponding proprietary description text is generated, and the proprietary description text is used to describe the meeting minutes requirement specifications of the meeting-specific key attributes. Based on the proprietary description text of each meeting-specific key attribute, the meeting minutes prompt word basic template is described and edited, and the description editing process includes adding or modifying, so as to obtain a specific meeting minutes prompt word template corresponding to the meeting data.

5. A terminal, wherein: The terminal includes: Processor; and A memory arranged to store computer executable instructions which, when executed, cause the processor to perform a method according to any one of claims 1 to 3.

6. A computer-readable storage medium, wherein: The computer-readable storage medium stores one or more programs, and when the one or more programs are executed by a processor, the method of any one of claims 1 to 3 is implemented.

Images