An innovative system and method for automatically generating meeting minutes and intelligently refining them

Abstract

The present application provides an innovative system and method for automatically generating meeting minutes and intelligently extracting them. The innovative system for automatically generating meeting minutes and intelligently extracting them includes: a speech recognition module for converting the speech content in the meeting into text in real time; a natural language processing module for performing in-depth analysis on the converted text, identifying important information, discussion topics, decision points and action items, and obtaining analysis results; a meeting minutes generation module for automatically generating structured meeting minutes based on the analysis results; and an intelligent content extraction module for extracting the most important content and accurately marking key decisions and subsequent action items. In the above technical solution, it can help users efficiently manage meeting content and improve work efficiency and decision-making quality.

Description

Background Art

[0002] In the operation of modern enterprises and organizations, meetings are a key link in information exchange and decision-making, and they frequently generate massive amounts of information. The traditional method of manually recording and organizing meeting content not only consumes a lot of manpower and time, but is also prone to information omissions and errors. In large-scale or multinational conference scenarios, the limitations of manual recording are more prominent, such as multilingual communication barriers and the difficulty in comprehensive recording due to the fast pace of meetings, which seriously affect the effective transformation and application of meeting results.

[0003] Currently, many companies still use the mode of manual recording and manual compilation of meeting minutes. This is not only inefficient, but also affected by the subjective factors of the recorder, making it difficult to ensure the accuracy and completeness of the information. When manually compiling complex meeting minutes, it is also easy to overlook key information and fail to accurately extract decision points and action items, resulting in the meeting minutes being unable to provide strong support for subsequent work. Summary of the invention

[0004] This application provides an innovative system and method for automatically generating meeting minutes and intelligently refining them, so as to help users efficiently manage meeting content and improve work efficiency and decision-making quality.

[0005] First, an innovative system for automatically generating meeting minutes and intelligently refining them is provided, including: Speech recognition module, used to convert speech content in meetings into text in real time; The natural language processing module is used to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points, and action items, and obtain analysis results; A meeting minutes generation module, used to automatically generate structured meeting minutes according to the analysis results; The intelligent content extraction module is used to extract the most important content and accurately mark key decisions and follow-up actions.

[0006] In the above technical solution, a speech recognition module is set up to convert the speech content in the meeting into text in real time; a natural language processing module is used to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points and action items, and obtain analysis results; a meeting minutes generation module is used to automatically generate structured meeting minutes based on the analysis results; an intelligent content extraction module is used to extract the most important content and accurately mark key decisions and subsequent action items; it can help users efficiently manage meeting content and improve work efficiency and decision-making quality.

[0007] In a specific embodiment, it also includes: The multimodal data processing module is used to collect and pre-process the multimodal data in the meeting and merge it with the text data after speech-to-text conversion.

[0008] In a specific embodiment, it also includes: The knowledge graph construction and application module is used to construct meeting-related knowledge graphs, and use knowledge graphs for semantic understanding, reasoning, and knowledge expansion during the generation of meeting minutes and the extraction of key content.

[0009] In a specific embodiment, it also includes: The multilingual translation and understanding optimization module is used to perform real-time multilingual adaptive translation and optimize the semantic understanding of the translated text.

[0010] In a specific implementation scheme, the meeting minutes generation module includes: Automatic summary generation unit, which is used to automatically generate summary content of the meeting through comprehensive analysis of meeting text, multimodal data and knowledge graph information; The structured content presentation unit is used to present the meeting minutes in a structured form.

[0011] In a specific implementation scheme, the intelligent content extraction module includes: Key content extraction unit, used to conduct in-depth analysis of meeting texts and extract the core content of the meeting; The decision point marking and tracking unit is used to automatically mark the decision points in the meeting and generate subsequent action item tracking records based on the decision points.

[0012] In the second aspect, an innovative method for automatically generating meeting minutes and intelligently refining them is provided, including the following steps: Use the speech recognition module to convert the speech content in the meeting into text in real time; Use the natural language processing module to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points and action items, and obtain analysis results; Using a meeting minutes generation module to automatically generate structured meeting minutes according to the analysis results; Use the intelligent content extraction module to extract the most important content and accurately mark key decisions and follow-up actions.

[0013] In the above technical solution, a speech recognition module is set up to convert the speech content in the meeting into text in real time; a natural language processing module is used to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points and action items, and obtain analysis results; a meeting minutes generation module is used to automatically generate structured meeting minutes based on the analysis results; an intelligent content extraction module is used to extract the most important content and accurately mark key decisions and subsequent action items; it can help users efficiently manage meeting content and improve work efficiency and decision-making quality.

[0014] In a specific embodiment, it also includes: The multimodal data processing module is used to collect and pre-process the multimodal data in the meeting, and then merge it with the text data after speech-to-text conversion.

[0015] In a specific embodiment, it also includes: Use the knowledge graph construction and application module to build the knowledge graph related to the meeting. In the process of generating meeting minutes and extracting key content, use the knowledge graph for semantic understanding, reasoning and knowledge expansion.

[0016] In a specific embodiment, it also includes: The multilingual translation and understanding optimization module is used to perform real-time multilingual adaptive translation and to optimize the semantic understanding of the translated text. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A structural diagram of the innovative system for automatically generating meeting minutes and intelligently refining provided in the embodiment of the present application; Figure 2 A flowchart of the innovative method for automatically generating meeting minutes and intelligently refining provided in the embodiment of the present application; Figure 3 A specific flow chart of another innovative method for automatically generating meeting minutes and intelligently refining provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] The present application is further described in detail below through the accompanying drawings and embodiments. Through these descriptions, the characteristics and advantages of the present application will become clearer and more specific.

[0019] The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise noted.

[0020] In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0021] To facilitate understanding of the automatic generation of meeting minutes and intelligent extraction innovation system and method provided by the embodiment of the present application, its application scenario is first explained. The automatic generation of meeting minutes and intelligent extraction innovation system and method provided by the embodiment of the present application are used to help users efficiently manage meeting content and improve work efficiency and decision-making quality. In the operation of modern enterprises and organizations, meetings are key links in information exchange and decision-making, and frequently generate massive amounts of information. The traditional way of relying on manual recording and collating meeting content not only consumes a lot of manpower and time, but is also prone to information omissions, errors and other problems. In large-scale meetings or multinational meetings, the limitations of manual recording are more prominent, such as multilingual communication barriers, fast meeting pace, etc., which make it difficult to record comprehensively, seriously affecting the effective transformation and application of meeting results. At present, many companies still use the mode of manual recording and manual arrangement of meeting minutes. This is not only inefficient, but also affected by the subjective factors of the recorder, and it is difficult to ensure the accuracy and completeness of the information. When manually arranging complex meeting minutes, it is also easy to ignore key information, and it is impossible to accurately extract decision points and action items, resulting in the inability of meeting minutes to provide strong support for subsequent work. To this end, the present application embodiment provides an innovative system and method for automatically generating meeting minutes and intelligently refining them, so as to help users efficiently manage meeting content, improve work efficiency and decision-making quality. The following is a detailed description of the embodiments in conjunction with specific drawings.

[0022] refer to Figures 1 to 3 , Figure 1 A structural diagram of the innovative system for automatically generating meeting minutes and intelligently refining provided in the embodiment of the present application; Figure 2 A flowchart of the innovative method for automatically generating meeting minutes and intelligently refining provided in the embodiment of the present application; Figure 3 A specific flow chart of another innovative method for automatically generating meeting minutes and intelligently refining provided in an embodiment of the present application.

[0023] exist Figure 1 In the embodiment of the present application, an innovative system for automatically generating meeting minutes and intelligently refining the minutes is provided, including: Speech recognition module, used to convert speech content in meetings into text in real time; The natural language processing module is used to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points, and action items, and obtain analysis results; A meeting minutes generation module, used to automatically generate structured meeting minutes according to the analysis results; The intelligent content extraction module is used to extract the most important content and accurately mark key decisions and follow-up actions.

[0024] In the above technical solution, a speech recognition module is set up to convert the speech content in the meeting into text in real time; a natural language processing module is used to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points and action items, and obtain analysis results; a meeting minutes generation module is used to automatically generate structured meeting minutes based on the analysis results; an intelligent content extraction module is used to extract the most important content and accurately mark key decisions and subsequent action items; it can help users efficiently manage meeting content and improve work efficiency and decision-making quality.

[0025] In a specific embodiment, it also includes: The multimodal data processing module is used to collect and pre-process the multimodal data in the meeting and merge it with the text data after speech-to-text conversion.

[0026] In a specific embodiment, it also includes: The knowledge graph construction and application module is used to construct meeting-related knowledge graphs, and use knowledge graphs for semantic understanding, reasoning, and knowledge expansion during the generation of meeting minutes and the extraction of key content.

[0027] In a specific embodiment, it also includes: The multilingual translation and understanding optimization module is used to perform real-time multilingual adaptive translation and optimize the semantic understanding of the translated text.

[0028] In a specific implementation scheme, the meeting minutes generation module includes: Automatic summary generation unit, which is used to automatically generate summary content of the meeting through comprehensive analysis of meeting text, multimodal data and knowledge graph information; The structured content presentation unit is used to present the meeting minutes in a structured form.

[0029] In a specific implementation scheme, the intelligent content extraction module includes: Key content extraction unit, used to conduct in-depth analysis of meeting texts and extract the core content of the meeting; The decision point marking and tracking unit is used to automatically mark the decision points in the meeting and generate subsequent action item tracking records based on the decision points.

[0030] Specifically, the automatic generation of meeting minutes and intelligent refining innovation system specifically includes: 1. Speech recognition module: Converts the speech content in the meeting into text in real time, supports multiple languages ​​and multiple audio tracks, and integrates advanced noise processing algorithms to ensure accurate speech recognition even in noisy environments.

[0031] 2. Natural language processing module: In-depth analysis of the converted text, combined with multimodal data and knowledge graphs, to identify important information, discussion topics, decision points and action items. Supports multilingual text processing and can automatically switch analysis models according to the meeting language environment.

[0032] 3. Meeting minutes generation module: Based on the analysis results, structured meeting minutes are automatically generated, supporting custom templates and formats. Multimodal data and knowledge graph information are used to enrich the content of meeting minutes and provide more comprehensive meeting records.

[0033] 4. Intelligent content extraction module: Through machine learning algorithms, it deeply analyzes meeting texts, multimodal data and knowledge graphs to extract the most important content, accurately mark key decisions and subsequent actions, and support the tracking and management of action items.

[0034] 5. Multimodal data processing module: collects and pre-processes multimodal data such as images and actions in the meeting, and integrates them with the text data after speech-to-text conversion to provide more comprehensive data support for natural language processing and meeting minutes generation.

[0035] 6. Knowledge graph construction and application module: Build a knowledge graph related to the meeting, and associate information such as people, events, and professional terms in the meeting. In the process of generating meeting minutes and extracting key content, use the knowledge graph for semantic understanding, reasoning, and knowledge expansion.

[0036] 7. Multi-language translation and understanding optimization module: Realize real-time multi-language adaptive translation, dynamically adjust translation models and strategies according to the conference context, and optimize the semantic understanding of the translated text to ensure that participants of different languages ​​have a consistent understanding of the conference content.

[0037] Furthermore, the speech recognition module comprises: 1. Speech-to-text conversion unit: Select advanced speech recognition technology, such as Google Speech-to-Text, IBM Watson Speech-to-Text or Microsoft Azure Speech Services, and integrate their API interfaces. By continuously optimizing the speech recognition model, the accuracy and real-time performance of speech-to-text conversion are improved, and speech recognition in multiple languages ​​is supported to meet the needs of meetings in different language environments.

[0038] 2. Noise processing and audio track separation unit: It uses advanced noise processing algorithms to effectively reduce the impact of environmental noise on speech recognition. It uses multi-audio track processing technology to accurately identify and separate the speech content of multiple speakers in a multi-speaking scenario to avoid content confusion and misrecognition.

[0039] Furthermore, the natural language processing module includes: 1. Text analysis and processing unit: Based on advanced natural language processing technologies such as BERT and GPT, the converted text is semantically analyzed. Combined with multimodal data (such as the expressions and actions of participants), it can more accurately identify important information, discussion topics and decision points in the meeting. It automatically marks speakers, time points and key discussion content to facilitate users to quickly understand the content of the meeting.

[0040] 2. Multi-language support and knowledge graph fusion unit: Supports the processing and analysis of multi-language texts, and automatically switches the analysis model according to the language environment of the meeting. Integrate the knowledge graph into the natural language processing process, and through the semantic understanding and reasoning ability of the knowledge graph, improve the depth of understanding of the meeting text and explore the potential information behind the text.

[0041] Furthermore, the meeting minutes generation module includes: 1. Automatic summary generation unit: Automatically generate summary content of the meeting through comprehensive analysis of meeting text, multimodal data and knowledge graph information. The summary covers the main discussion points, decision results, action items, and related background knowledge and extended information of the meeting. Users can customize the format and template of the summary according to their needs to generate meeting minutes that meet corporate standards and personal needs.

[0042] 2. Structured content presentation unit: Supports presenting meeting minutes in a structured form, such as a tree diagram, table, or list. Multimodal data and knowledge graph information are used to enrich structured content, making meeting minutes more intuitive and clear, and easier for users to understand and consult. Users can also edit and supplement meeting minutes to ensure that the final minutes meet actual needs.

[0043] Furthermore, the intelligent content extraction module includes: 1. Key content extraction unit: Use machine learning algorithms, combined with multimodal data and knowledge graphs, to conduct in-depth analysis of conference texts. Identify high-frequency words, subject words, key sentences, and important multimodal information to extract the core content of the conference. Through the association and reasoning of knowledge graphs, potential key information is mined to provide users with more comprehensive key content extraction results.

[0044] 2. Decision point marking and tracking unit: Automatically mark the decision points in the meeting and generate subsequent action item tracking records based on the decision points. Use the knowledge graph to perform semantic understanding and association analysis on the decision points, providing richer background information and decision basis for the decision points. Users can track the execution of these action items in the system to ensure that the meeting decisions are effectively implemented.

[0045] Furthermore, the multimodal data processing module includes: 1. Data collection and preprocessing unit: collect multimodal data such as images and actions in the meeting through cameras, sensors and other devices. Preprocess the collected data, including data cleaning, feature extraction and other operations, to convert the multimodal data into a format that can be used for analysis.

[0046] 2. Data fusion and analysis unit: The pre-processed multimodal data is fused with the text data after speech-to-text conversion, and the fusion algorithm based on the attention mechanism and other technologies are adopted to enable the natural language processing module to better utilize multimodal information for analysis. With the assistance of multimodal data, the ability to understand and analyze the content of the meeting is improved, providing more comprehensive data support for the generation of meeting minutes and the extraction of key content.

[0047] Furthermore, the knowledge graph construction and application module includes: 1. Knowledge graph construction unit: Collect entity information such as people, events, professional terms, and the relationships between them in the meeting to build a knowledge graph related to the meeting. Use natural language processing technology and machine learning algorithms to perform entity recognition and relationship extraction on the meeting text to continuously enrich and improve the knowledge graph.

[0048] 2. Knowledge graph application unit: In the process of generating meeting minutes and extracting key content, the knowledge graph is used for semantic understanding and reasoning. When a certain professional term or person is mentioned, the system can automatically associate it with the relevant explanations and background information in the knowledge graph and the relevant content involved in past meetings, providing richer background knowledge and contextual information for the meeting minutes. Through the intelligent expansion function of the knowledge graph, reference materials and research directions on related topics are automatically generated to assist decision-making.

[0049] Furthermore, the multilingual translation and understanding optimization module includes: 1. Real-time multi-language adaptive translation unit: Connect to the advanced neural network machine translation engine to achieve real-time multi-language translation. Using reinforcement learning algorithms, dynamically adjust translation models and strategies according to the meeting context, language habits of participants and real-time language switching to improve the accuracy and naturalness of translation.

[0050] 2. Semantic Understanding Optimization Unit: Optimize the semantic understanding of the translated text and use natural language processing technology to analyze the semantic structure and logical relationship of the translated text. Combined with knowledge graphs and multimodal data, it eliminates translation ambiguity and ensures that participants of different languages ​​have a consistent understanding of the conference content.

[0051] Furthermore, the system's scalability design includes: 1. Modular architecture: The system adopts a modular design, and each functional module (such as speech recognition, natural language processing, multimodal data processing, etc.) is independent of each other but works together. This architecture allows users to expand functions according to needs, such as adding new speech recognition engines, natural language processing models, or multimodal data acquisition devices, without affecting the overall operation of the system.

[0052] 2. Multi-platform support: The system supports deployment and operation on multiple platforms, including cloud platforms (such as Alibaba Cloud and Tencent Cloud), local servers, and mobile devices (such as iOS and Android systems). By developing corresponding client applications, it ensures that users can use the system conveniently on different devices and network environments.

[0053] Furthermore, the system availability design includes: 1. User-friendly interface: The system provides a simple and intuitive user interface. Users can complete the generation of meeting minutes, key content extraction, meeting follow-up tracking and other functions through simple operations. The interface design follows the principle of humanization and provides clear operation instructions and feedback information. The system also provides a multi-language interface to meet the needs of global teams.

[0054] 2. High availability architecture: The system adopts a high availability architecture design, supporting load balancing and automatic failover. The storage and processing nodes of key data are distributed in multiple geographical locations to avoid system downtime caused by single point failure. By real-time monitoring of the system's operating status, In this embodiment, by integrating speech recognition, natural language processing, machine learning, multimodal data processing and knowledge graph technology, automatic recording of meeting content, intelligent minutes generation, and in-depth extraction of key content are achieved, while solving the language barrier problem in multilingual meetings. The system adopts a modular design, and each module works together to support core functions such as speech-to-text conversion, multimodal data fusion analysis, automatic summary generation, content expansion based on knowledge graphs, decision point marking and tracking, and real-time multilingual adaptive translation and understanding optimization, helping users to efficiently manage meeting content and improve work efficiency and decision-making quality.

[0055] In the above technical solution, automated and multimodal recording of meeting content is achieved, minimizing human intervention; the efficiency and accuracy of meeting minutes generation are greatly improved, generating meeting minutes with rich content and clear logic; intelligent extraction technology is used, combined with multimodal data and knowledge graphs, to accurately identify and extract key content, decision points and action items in the meeting; real-time adaptive translation and understanding optimization in multilingual environments are supported to ensure smooth communication among global teams; intelligent association and expansion of meeting content are achieved through knowledge graphs to provide more comprehensive knowledge support for decision-making.

[0056] exist Figure 2In the embodiment of the present application, an innovative method for automatically generating meeting minutes and intelligently refining the minutes is provided, comprising the following steps: Use the speech recognition module to convert the speech content in the meeting into text in real time; Use the natural language processing module to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points and action items, and obtain analysis results; Using a meeting minutes generation module to automatically generate structured meeting minutes according to the analysis results; Use the intelligent content extraction module to extract the most important content and accurately mark key decisions and follow-up actions.

[0057] In the above technical solution, a speech recognition module is set up to convert the speech content in the meeting into text in real time; a natural language processing module is used to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points and action items, and obtain analysis results; a meeting minutes generation module is used to automatically generate structured meeting minutes based on the analysis results; an intelligent content extraction module is used to extract the most important content and accurately mark key decisions and subsequent action items; it can help users efficiently manage meeting content and improve work efficiency and decision-making quality.

[0058] In a specific embodiment, it also includes: The multimodal data processing module is used to collect and pre-process the multimodal data in the meeting, and then merge it with the text data after speech-to-text conversion.

[0059] In a specific embodiment, it also includes: Use the knowledge graph construction and application module to build the knowledge graph related to the meeting. In the process of generating meeting minutes and extracting key content, use the knowledge graph for semantic understanding, reasoning and knowledge expansion.

[0060] In a specific embodiment, it also includes: The multilingual translation and understanding optimization module is used to perform real-time multilingual adaptive translation and to optimize the semantic understanding of the translated text.

[0061] Specifically, refer to Figure 3 The innovative method for automatically generating meeting minutes and intelligently refining them includes the following steps: 1. Real-time meeting recording process 1. After the system is started, the user selects the conference voice input source, and the system automatically turns on the camera and sensor to collect multimodal data such as images and actions in the meeting.

[0062] 2. The speech recognition module processes the speech content in the meeting in real time, converts it into structured text, and performs preliminary noise reduction and multi-track separation on the text.

[0063] 3. The multimodal data processing module pre-processes the collected multimodal data and then integrates it with the text data after speech-to-text conversion.

[0064] 4. The integrated data is stored in the system's database, providing data support for subsequent meeting minutes generation and key content extraction.

[0065] 2. Meeting Minutes Generation Process 1. After the meeting, the system automatically calls the natural language processing module to conduct an in-depth analysis of the integrated data of the entire meeting (including text and multimodal data). During the analysis process, the natural language processing module combines the knowledge graph for semantic understanding and reasoning.

[0066] 2. The system automatically generates meeting minutes based on the analysis results. The meeting minutes include the main discussion points, decision results, and action items of the meeting, and also incorporate multimodal data (such as important expressions and action information) and background knowledge and extended information provided by the knowledge graph.

[0067] 3. Users can view the automatically generated meeting minutes in the system and adjust the format, edit and supplement the content as needed to ensure that the meeting minutes meet actual needs.

[0068] 3. Key content extraction process 1. The system uses machine learning algorithms, combined with multimodal data and knowledge graphs, to conduct in-depth analysis of conference fusion data. It identifies high-frequency words, topic words, key sentences, and important multimodal information, and extracts the core content of the meeting.

[0069] 2. The system automatically marks the key decision points and subsequent action items in the meeting, and uses the knowledge graph to perform semantic understanding and association analysis on the decision points, providing richer background information and decision-making basis for the decision points.

[0070] 3. Users can view and manage the extracted key content in the system, and adjust and optimize the key content and action items according to actual needs.

[0071] 4. Meeting follow-up process 1. After the system automatically generates meeting minutes, users can add decision points and action items in the minutes to the system's tracking module.

[0072] 2. The system will automatically remind and track the execution of each action item according to the predetermined schedule or user settings. By integrating with the company's existing project management tools or office systems, real-time monitoring of the progress of action items can be achieved.

[0073] 3. Users can view the execution progress of each action item in the system and make adjustments and optimizations based on actual conditions to ensure that meeting decisions are effectively implemented.

[0074] 5. Multilingual Translation and Understanding Optimization Process 1. During the meeting, the multilingual translation and understanding optimization module monitors the language changes in the meeting in real time. Once a language switch or multilingual speech is detected, the translation process is immediately started.

[0075] 2. Connect to the translation engine for real-time translation, and use reinforcement learning algorithms to dynamically adjust the translation strategy according to the meeting context to improve the accuracy and naturalness of the translation.

[0076] 3. Optimize the semantic understanding of the translated text, combine the knowledge graph and multimodal data to eliminate translation ambiguity, and ensure that participants of different languages ​​have a consistent understanding of the meeting content. The optimized translated text is stored and processed together with the original meeting data, providing unified multilingual data support for the generation of meeting minutes and the extraction of key content.

[0077] In the above technical solution, the beneficial effects of the innovative method of automatically generating meeting minutes and intelligently refining them include: 1. Speech Recognition Technology 1. Technology selection and optimization: According to project requirements and actual test results, select the most suitable speech recognition technology and integrate its API interface. Continue to pay attention to the development of speech recognition technology, regularly update and optimize speech recognition models, and improve the accuracy and real-time performance of speech-to-text conversion.

[0078] 2. Multi-track and noise processing: The advanced multi-track processing technology and noise suppression algorithm are used to ensure that the speech recognition module can still accurately identify and separate the speech content of each person when multiple people are speaking in a noisy environment, effectively reducing the impact of noise on speech recognition.

[0079] 2. Natural Language Processing Technology 1. Text analysis and multimodal fusion: Based on advanced natural language processing models such as BERT and GPT, semantic analysis of conference texts is performed. Combining multimodal data, a fusion algorithm based on the attention mechanism is used to integrate multimodal features into the natural language processing process, thereby improving the ability to understand and analyze conference content.

[0080] 2. Knowledge graph fusion and multi-language support: Integrate knowledge graph into natural language processing, and use the semantic understanding and reasoning capabilities of knowledge graph to enhance the depth of understanding of conference texts. Support multi-language text processing, automatically switch natural language processing models and knowledge graph resources according to the conference language environment, and ensure that conference content can be accurately analyzed in different language environments.

[0081] 3. Machine Learning Algorithms 1. Model training and optimization: For different functional modules (such as intelligent content extraction module, multi-language translation optimization module), corresponding machine learning algorithms are used for model training. A large amount of conference data and multimodal data are used to train and optimize the model to improve the adaptability and accuracy of the model in different conference scenarios. Users are supported to customize and train their own models to meet specific needs.

[0082] 2. Multimodal and knowledge graph applications: Integrate multimodal data and knowledge graph information into machine learning algorithms, and improve model performance through feature extraction of multimodal data and association analysis of knowledge graphs. For example, in the process of extracting key content, use the structural information of knowledge graphs to guide machine learning algorithms to mine potential key information.

[0083] 4. System Integration and Data Management 1. System integration: The system supports integration with the company's existing office systems (such as Office 365, Google Workspace) and project management tools (such as Jira, Trello) to achieve seamless data connection and synchronization. Users can directly use the meeting minutes tool in the existing office environment to improve work efficiency.

[0084] 2. Data management: Cloud storage and distributed database technology are used to ensure the security and reliability of conference data. Conference data is stored and managed in a classified manner, including voice data, text data, multimodal data, knowledge graph data, etc. The system supports automatic backup and recovery of data to ensure rapid recovery in the event of data loss or system failure.

[0085] 5. Multimodal Data Processing Technology 1. Data collection and preprocessing: Select appropriate cameras, sensors and other equipment to collect multimodal data, and develop corresponding data collection interfaces. Preprocess the collected multimodal data, including image preprocessing, motion feature extraction and other operations, and convert the multimodal data into a format that can be used for analysis.

[0086] 2. Data fusion algorithm: Use technologies such as fusion algorithms based on attention mechanisms to fuse multimodal data with text data. By adjusting the parameters of the fusion algorithm, the role of multimodal data in natural language processing and meeting minutes generation is optimized, and the system's ability to understand and analyze meeting content is improved.

[0087] 6. Knowledge Graph Technology 1. Algorithm construction: Use natural language processing technology and machine learning algorithms to perform entity recognition and relationship extraction, and build a knowledge graph related to the conference. Use a graph database (Neo4j) to store the knowledge graph to ensure efficient query and update of the knowledge graph.

[0088] 2. Application and expansion: In the process of generating meeting minutes and extracting key content, the knowledge graph is used for semantic understanding, reasoning and knowledge expansion. By continuously collecting and analyzing new meeting data, the knowledge graph is continuously enriched and improved to improve its application value.

[0089] 7. Multilingual Translation Technology 1. Translation engine connection: Connect with advanced neural network machine translation engines (such as Baidu Translate Open Platform, Tencent Cloud Translation, etc.) to achieve real-time multi-language translation. Select the appropriate translation model and language pair according to the meeting needs to ensure the accuracy and efficiency of the translation.

[0090] 2. Adaptive optimization: Using reinforcement learning algorithms, we dynamically adjust translation models and strategies based on the meeting context, the language habits of participants, and real-time language switching. We post-process the translated text, combine knowledge graphs and multimodal data to eliminate translation ambiguity, and improve the quality of the translated text.

[0091] Those skilled in the art will appreciate that the present application may be implemented as a system, method or computer program product.

[0092] Therefore, the present disclosure may be specifically implemented in the following forms, namely: it may be complete hardware, it may be complete software (including firmware, resident software, microcode, etc.), or it may be a combination of hardware and software, generally referred to herein as a "circuit", "module" or "system". In addition, in some embodiments, the present application may also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium may contain computer-readable program code.

[0093] Any combination of one or more computer-readable media can be used. Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this document, computer-readable storage media can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0094] Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art can change, modify, replace and modify the above embodiments within the scope of the present application. On this basis, the present application can be replaced and improved in many ways, all of which fall within the scope of protection of the present application.

Claims

1. An innovative system for automatically generating meeting minutes and intelligently refining them, characterized in that: include: Speech recognition module, used to convert speech content in meetings into text in real time; The natural language processing module is used to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points, and action items, and obtain analysis results; A meeting minutes generation module, used to automatically generate structured meeting minutes according to the analysis results; The intelligent content extraction module is used to extract the most important content and accurately mark key decisions and follow-up actions.

2. The automatic generation of meeting minutes and intelligent extraction innovation system according to claim 1 is characterized in that: Also includes: The multimodal data processing module is used to collect and pre-process the multimodal data in the meeting and merge it with the text data after speech-to-text conversion.

3. The automatic generation of meeting minutes and intelligent extraction innovation system according to claim 2 is characterized in that: Also includes: The knowledge graph construction and application module is used to construct meeting-related knowledge graphs, and use knowledge graphs for semantic understanding, reasoning, and knowledge expansion during the generation of meeting minutes and the extraction of key content.

4. The automatic generation of meeting minutes and intelligent extraction innovation system according to claim 3 is characterized in that: Also includes: The multilingual translation and understanding optimization module is used to perform real-time multilingual adaptive translation and optimize the semantic understanding of the translated text.

5. The automatic generation of meeting minutes and intelligent extraction innovation system according to claim 4 is characterized in that: The meeting minutes generation module includes: Automatic summary generation unit, which is used to automatically generate summary content of the meeting through comprehensive analysis of meeting text, multimodal data and knowledge graph information; The structured content presentation unit is used to present the meeting minutes in a structured form.

6. The automatic generation of meeting minutes and intelligent extraction innovation system according to claim 5 is characterized in that: The intelligent content extraction module includes: Key content extraction unit, used to conduct in-depth analysis of meeting texts and extract the core content of the meeting; The decision point marking and tracking unit is used to automatically mark the decision points in the meeting and generate subsequent action item tracking records based on the decision points.

7. An innovative method for automatically generating meeting minutes and intelligently refining them, characterized in that: The following steps are involved: Use the speech recognition module to convert the speech content in the meeting into text in real time; Use the natural language processing module to conduct in-depth analysis of the converted text, identify important information, discussion topics, decision points and action items, and obtain analysis results; Using a meeting minutes generation module to automatically generate structured meeting minutes according to the analysis results; Use the intelligent content extraction module to extract the most important content and accurately mark key decisions and follow-up actions.

8. The innovative method for automatically generating meeting minutes and intelligently refining according to claim 7 is characterized in that: Also includes: The multimodal data processing module is used to collect and pre-process the multimodal data in the meeting, and then merge it with the text data after speech-to-text conversion.

9. The innovative method for automatically generating meeting minutes and intelligently refining according to claim 8 is characterized in that: Also includes: Use the knowledge graph construction and application module to build the knowledge graph related to the meeting. In the process of generating meeting minutes and extracting key content, use the knowledge graph for semantic understanding, reasoning and knowledge expansion.

10. The innovative method for automatically generating meeting minutes and intelligently refining according to claim 9 is characterized in that: Also includes: The multilingual translation and understanding optimization module is used to perform real-time multilingual adaptive translation and to optimize the semantic understanding of the translated text.

Images