Conference memory enhancement method, device and medium for conference tablet

Abstract

The present application discloses a conference memory enhancement method, device and medium for a conference tablet, which relates to the technical field of intelligent conference equipment. The method includes: based on a preset data acquisition technology, real-time acquisition of multimodal data, and encoding the multimodal data into a spatiotemporal event stream according to the time sequence of the multimodal data; recognizing handwritten text on the conference tablet, and extracting key tags in the speech in combination with a pre-trained natural language processing model to construct a conference knowledge graph corresponding to the current meeting; comparing the conference knowledge graph with the historical conference knowledge graph to determine the conflicting content, and superimposing a virtual information layer on the physical interface of the conference tablet; based on the conference knowledge graph and the spatiotemporal time stream, querying the query association information corresponding to the semantic query request to determine the semantic contradiction, and displaying the historical conference evidence chain of the semantic contradiction based on the virtual information layer.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent conference equipment, and in particular to a conference memory enhancement method, device, and medium for a conference tablet. Background Art

[0002] As digital transformation and intelligent collaborative office converge, conference tablets, as the core carrier of enterprise-level intelligent interactive terminals, have gradually become key infrastructure for information sharing and real-time collaboration in modern meeting scenarios. With the surge in demand for hybrid working models and remote collaboration, the multimodal data generated during meetings is growing exponentially.

[0003] While current mainstream conference tablets have implemented basic content collection and storage capabilities, their technical architecture remains at the early stages of single-dimensional data recording. Specifically, existing technologies use independent storage mechanisms for handwriting traces, voice clips, and electronic documents, lacking associative mapping based on spatiotemporal features. For example, users cannot trace the location of whiteboard annotations corresponding to a particular voice discussion, resulting in a broken chain of meeting context information. Traditional keyword matching technology struggles to parse implicit logical relationships within meeting content. For example, for complex semantic queries such as "objections to project risk assessment," the system is unable to identify similar decision paths across meetings. Topics, decision items, and to-do tasks generated by different meetings exist as isolated nodes, failing to construct a cross-meeting knowledge evolution model and automatically identifying the progress of discussions on repeated topics. A perceptual gap exists between the physical writing layer and the digital information layer, making it impossible to implement deep interactive features such as dynamic content annotation and real-time floating display of related information through augmented reality technology.

[0004] Existing technologies have also proposed meeting minutes generation solutions based on speech recognition. However, they have not established a spatiotemporal alignment model for multi-source heterogeneous data, resulting in the inability to accurately match handwritten annotation timestamps with corresponding speech paragraphs. They also lack a dynamic update mechanism based on knowledge graphs, making it impossible to track the evolution of specific topics in different meetings. Moreover, the interactive level is still limited to two-dimensional plane display, and spatial computing technology has not been integrated to achieve three-dimensional enhanced presentation of meeting content, resulting in insufficient utilization of corporate meeting knowledge assets, seriously restricting the continuous improvement of organizational decision-making quality and operational efficiency. Summary of the Invention

[0005] The embodiments of the present application provide a conference memory enhancement method, device, and medium for a conference tablet to solve the above-mentioned technical problems.

[0006] In one aspect, an embodiment of the present application provides a conference memory enhancement method for a conference tablet, comprising:

[0007] Based on a preset data acquisition technology, multimodal data is collected in real time, and the multimodal data is encoded into a spatiotemporal event stream according to the time sequence of the multimodal data; the multimodal data includes touch data, voice data, and sensory data;

[0008] Recognize handwritten text on the conference tablet and, combined with a pre-trained natural language processing model, extract key tags from the speech to build a conference knowledge graph corresponding to the current meeting;

[0009] Comparing the conference knowledge graph with historical conference knowledge graphs to determine conflicting content, and overlaying a virtual information layer on the physical interface of the conference tablet;

[0010] Based on the conference knowledge graph and the spatiotemporal time flow, query association information corresponding to the semantic query request is queried to determine the semantic contradiction, and the historical conference evidence chain of the semantic contradiction is displayed based on the virtual information layer.

[0011] In one implementation of the present application, multimodal data is collected in real time based on a preset data collection technology, specifically including:

[0012] Using preset touch note tracking technology, the writing trajectory, note pressure, and erasure operations on the conference tablet are collected in real time; the touch note tracking technology includes electromagnetic induction or capacitive touch technology;

[0013] The direction of the sound source is detected through beamforming technology, and combined with the coordinate positioning of the infrared sensor, the identity of the speaker to be bound is determined and the speaker's voice characteristics are extracted;

[0014] Comparing the voice feature with a pre-stored voiceprint template of the speaker identity to be bound, and completing the binding of the speaker identity when the similarity is greater than a confidence threshold, and associating and storing the bound voice segment and whiteboard operation record according to the timestamp;

[0015] Through the camera and infrared sensor integrated in the conference tablet, the speaker and the corresponding gesture operations are identified, and the spatial distribution of each participant is determined.

[0016] In one implementation of this application, a pre-trained natural language processing model is combined to extract key tags from the speech to construct a conference knowledge graph corresponding to the current meeting, specifically including:

[0017] Using a pre-trained natural language processing model, the meeting topics and task trigger words in the speech are identified to extract the task items and the corresponding task responsible persons, and the task responsible persons are bound to the task items.

[0018] Perform sentiment analysis on the speech, determine the controversial issues, meeting decisions, and corresponding responsible persons in the speech, and mark the controversial issues as controversial;

[0019] Constructing a conference knowledge graph based on a graph neural network and integrating the conference knowledge graph with the enterprise database;

[0020] The meeting agenda, meeting decisions, responsible persons and related documents of the current meeting are embedded into corresponding nodes in the meeting knowledge graph, a dependency chain is generated, and each node is associated with the corresponding historical task node.

[0021] In one implementation of the present application, the conference knowledge graph is compared with the historical conference knowledge graph to determine the conflicting content, specifically including:

[0022] If the identified meeting topics, meeting decisions, and task items of the current meeting are associated with the historical topics corresponding to the previous meetings, the corresponding related content will be pushed to the sidebar of the conference tablet;

[0023] Perform semantic comparison of key entities of the current meeting decision with those of historical meeting decisions, and calculate the difference between the current meeting decision and the historical meeting decision; the difference includes budget value, time node, and responsible person allocation;

[0024] When the difference exceeds a preset threshold, a conflict node is generated to mark the conflict relationship between the current meeting decision and the historical meeting decision on the conference tablet and display the corresponding conflict content.

[0025] In one implementation of the present application, based on the conference knowledge graph and spatiotemporal time stream, query association information corresponding to the semantic query request is queried to determine semantic contradictions, and the historical conference evidence chain of the semantic contradiction is displayed based on the virtual information layer, specifically including:

[0026] Based on the intelligent search interface on the conference tablet, a semantic query request is received, and based on the conference knowledge graph and spatiotemporal event stream in the enterprise database, the corresponding request-related information is searched; the request-related information includes the associated whiteboard image, voice clips and task status;

[0027] Determine semantic contradictions between the discussion content of the current meeting and historical decisions based on the associated whiteboard screen, the voice clips, and the task status, and display the disputed clips with semantic contradictions as a heat map on the virtual information layer;

[0028] Generate a meeting sequence report based on the similarity of topics between the current meeting and historical meetings to visualize the knowledge evolution path.

[0029] In one implementation of the present application, superimposing a virtual information layer on the physical interface of the conference tablet specifically includes:

[0030] Through the built-in projection or external virtual device of the conference tablet, a virtual information layer is superimposed on the physical interface of the conference tablet, and the relevant content of the historical meeting is highlighted through the virtual information layer;

[0031] Obtaining task status indications in the current meeting in real time and visually displaying the task status indications on the virtual information layer; the task status indications include task completion, approval status, and overdue mark;

[0032] The entire version change history of the whiteboard content is recorded on the virtual information layer, and the entire version change history can be slid back according to the time axis.

[0033] In one implementation of the present application, before comparing the conference knowledge graph with the historical conference knowledge graph to determine conflicting content, the method further includes:

[0034] Map the meeting responsible person to the corresponding employee ID in the employee data table in the enterprise database to perform entity alignment;

[0035] The meeting decision dependencies are associated with the financial budget table and the project schedule, and the current decision is compared with the constraints of the enterprise database to generate corresponding risk path prompts.

[0036] In one implementation of the present application, after constructing the conference knowledge graph corresponding to the current conference, the method further includes:

[0037] Analyze the responsible persons, deadlines, and dependencies of the current meeting's tasks, and associate the tasks with the corresponding whiteboard images and voice clips;

[0038] The task ID of the project management in the associated enterprise database is generated to jump to the task details and a clickable hyperlink is embedded in the meeting minutes document.

[0039] On the other hand, an embodiment of the present application further provides a conference memory enhancement device for a conference tablet, the device comprising:

[0040] at least one processor;

[0041] and, a memory communicatively coupled to the at least one processor;

[0042] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned conference memory enhancement method for a conference tablet.

[0043] On the other hand, an embodiment of the present application further provides a non-volatile computer storage medium storing computer-executable instructions. When the computer-executable instructions are executed, a conference memory enhancement method for a conference tablet as described above is implemented.

[0044] The present application provides a conference memory enhancement method, device, and medium for a conference tablet, which have at least the following beneficial effects:

[0045] By collecting, structuring, and associating multimodal data during meetings, users can easily trace the basis and process of meeting decisions. The integration of historical meeting records, relevant documents, task status, and other information makes every aspect of decision-making traceable, greatly improving the traceability of decisions. By utilizing real-time data analysis and semantic understanding capabilities, it is possible to quickly identify content that conflicts with historical decisions or existing task status during meeting discussions. Once a conflict is detected, it helps users adjust the direction of the discussion in a timely manner to avoid invalid decisions. Relevant information is directly superimposed on the virtual information layer interface of the conference tablet, eliminating the need for users to switch between multiple applications or documents to find information. The intuitive information presentation reduces the user's cognitive load during information retrieval and understanding, allowing users to focus more on the meeting discussion itself. Relevant topics, decisions, and tasks are accurately identified and associated between different meetings, providing users with a coherent knowledge system, improving meeting efficiency and knowledge transfer effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0047] Figure 1 A flowchart of a conference memory enhancement method for a conference tablet provided in an embodiment of the present application;

[0048] Figure 2 A schematic diagram of a system architecture provided in an embodiment of the present application;

[0049] Figure 3 A flowchart of another method for enhancing conference memory for a conference tablet provided in an embodiment of the present application;

[0050] Figure 4A schematic diagram of the internal structure of a conference memory enhancement device for a conference tablet provided in an embodiment of the present application. DETAILED DESCRIPTION

[0051] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0052] The technical solutions provided by the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0053] Figure 1 A flowchart of a conference memory enhancement method for a conference tablet provided in an embodiment of the present application.

[0054] The analysis method involved in the embodiments of the present application can be implemented by a terminal device or a server, and the present application does not impose any special restrictions on this. For ease of understanding and description, the following embodiments are described in detail using a server as an example.

[0055] It should be noted that the server can be a single device or a system composed of multiple devices, that is, a distributed server, and this application does not make any specific restrictions on this.

[0056] like Figure 1 As shown, an embodiment of the present application provides a conference memory enhancement method for a conference tablet, comprising:

[0057] 101. Based on the preset data collection technology, multimodal data is collected in real time, and the multimodal data is encoded into a spatiotemporal event stream according to the time sequence of the multimodal data.

[0058] It should be noted that the multimodal data in the embodiment of the present application includes touch data, voice data and perception data.

[0059] Figure 2 A schematic diagram of a system architecture provided in an embodiment of the present application. Figure 2 As shown, the present application discloses a system that can integrate multimodal conference data in real time, automatically construct a knowledge network of structured conference knowledge graph, and realize dynamic backtracking of conference memory and decision assistance through augmented reality technology, including a multimodal acquisition layer corresponding to the multimodal data acquisition module, an intelligent processing layer corresponding to the intelligent processing center, an AR interaction layer corresponding to the augmented reality interaction module, and a memory enhancement layer corresponding to the memory enhancement service module.

[0060] The multimodal acquisition layer includes a touch note tracking unit, a voice directional acquisition unit and an environmental perception unit; the intelligent processing layer includes a spatiotemporal alignment engine, content structured processing and incremental knowledge management modules; the AR interaction layer includes an AR overlay unit and a dynamic version controller; and the memory enhancement layer includes an intelligent retrieval interface, a decision conflict warning unit and a cross-conference association engine.

[0061] Specifically, in one embodiment of the present application, the multimodal data acquisition module includes a touch handwriting tracking unit, a voice directional acquisition unit, and an environmental perception unit. During the meeting, the touch handwriting tracking unit records the writing trajectory, handwriting pressure, and erasing operation through electromagnetic induction or capacitive touch technology. The voice directional acquisition unit uses a beamforming microphone array to distinguish different speakers and associate them with the whiteboard operation timeline. The environmental perception unit recognizes user gesture operations based on the built-in camera of the conference tablet, such as area circling and tapping reminders. The infrared sensor detects the location distribution of the participants. The voice directional acquisition unit maps and matches the sound source direction detected by beamforming with the position of the person identified by infrared detection. The system can bind the voice signal to a specific speaker.

[0062] In one embodiment, beamforming detected that the direction of the sound source was 45°, that is, right front, with coordinates of (6m, 3m). Infrared detection found that the coordinates corresponded to ID002, which was Li Si. Voiceprint verification extracted voice features and compared them with Li Si's voiceprint template with a confidence level of 92%. The final system output whiteboard operation record timestamp was 10:15. Li Si circled the whiteboard area and spoke, and the content of his speech was "It is recommended to modify the third part."

[0063] In one embodiment, based on the spatiotemporal alignment engine of the intelligent processing center, a unified timestamp is established for each collected touch data, voice data, and perception data, and the handwriting, voice, and gesture operations are encoded into corresponding spatiotemporal event streams in the order of occurrence.

[0064] 102. Recognize handwritten text on the conference tablet and, combined with a pre-trained natural language processing model, extract key tags from the speech to build a conference knowledge graph corresponding to the current meeting.

[0065] Specifically, in one embodiment of the present application, the content structuring processor uses optical character recognition (OCR) technology to recognize handwritten text on a conference tablet. Combined with a natural language processing (NLP) model, it extracts task items, decision points, and controversial issues from speeches. For example, consider the topic node: Product 2025-Q2 feature development priorities. Discussion process: Speech A: "An AI assistant must be added. Controversy: Insufficient resources." Specifically, sentiment analysis is performed on speech A to identify the controversial issue associated with speech A and mark it as controversial. Speech B: "It has been decided to launch the payment module first, with Mr. Wang in charge." Based on speech B, task items are extracted, the corresponding responsible individuals are identified, and the task items are then bound to the responsible individuals. Whiteboard writing: "Conclusion: Payment module priority P0" is used to identify the conclusion node through OCR and determine the meeting decision.

[0066] A meeting knowledge graph is constructed based on a graph neural network. The nodes in the meeting knowledge graph include "topic-decision-responsible person-related documents," with related documents such as the historical resolution "2024 Resource Quota." The meeting knowledge graph is integrated with the enterprise database, embedding the current meeting's topic, decision, responsible person, and related documents into the corresponding nodes in the meeting knowledge graph. This generates a dependency chain and links each node to the corresponding historical task node.

[0067] In one embodiment, after constructing the meeting knowledge graph corresponding to the current meeting, a clickable hyperlink is created, associating the task item with the corresponding whiteboard image and discussion recording. Specifically, the task items for the current meeting are parsed for responsible individuals, deadlines, and dependencies, and the task items are associated with the corresponding whiteboard image and audio clips. Then, the task ID for project management in the enterprise database is linked to generate a clickable hyperlink to the task details, which is then embedded in the meeting minutes document.

[0068] In one embodiment, the meeting knowledge graph is compared with the historical meeting knowledge graph. Before determining the conflicting content, the person responsible for the meeting is mapped to the corresponding employee ID in the employee data table in the enterprise database for entity alignment, and the meeting decision dependencies are associated with the financial budget table and project schedule. The current decision is compared with the constraints of the enterprise database to generate corresponding risk path prompts.

[0069] Figure 3 This is a flow chart of another method for enhancing conference memory for a conference tablet provided in an embodiment of the present application. Figure 3As shown in the figure, during the meeting, touch handwriting, voice data and gesture operations are collected synchronously to generate a raw event stream with a timestamp; task tags (such as "★Urgent") in handwritten content are recognized in real time to trigger to-do reminders.

[0070] At the end of the meeting, a structured meeting summary is automatically generated, integrating the meeting knowledge graph with the enterprise database, including clickable hyperlinks. Clicking on a task item jumps to the corresponding whiteboard screen and discussion recording, updating the global decision network. It should be noted that the meeting knowledge graph in the embodiment of the present application includes the meeting's topics, conclusions, and task dependencies, while the enterprise database includes ERP project progress, CRM customer feedback, and financial budget tables.

[0071] Scenario: New product launch decision

[0072] Topic: "Product launch plan for Q3 2025" → Conclusion: "Need to increase testers, budget +20%."

[0073] Enterprise database status:

[0074] HR table: The current number of people in the testing team is 5, and the maximum number of people is 10.

[0075] Financial Statement: The remaining quarterly budget is 500,000.

[0076] Integration and update:

[0077] Entity alignment: Meeting responsible person "Testing Supervisor Li" ↔ "HR Library Wang".

[0078] Relationship enhancement: Decision: Add testing → Dependency: Budget 200,000.

[0079] Conflict detection: If the financial table shows that the remaining budget is only 100,000, a conflict node will be generated, triggering an alert.

[0080] Global decision network changes:

[0081] Added a new node [Product launch decision Q3] and associated historical node [2025-Q2 decision: postpone launch].

[0082] Conflict path prompt: "Budget gap requires financial approval or cutting other expenses."

[0083] In the subsequent retrospective stage, through voice sentiment analysis, that is, using the sentiment analysis model to identify angry / questioning tones, and through speech conflict detection, that is, detecting the frequency of "but" / "however" in the speech ≥ 5 times / minute → dispute + 0.2, the historical meeting heat map is displayed on the AR interface of the virtual information layer, that is, the controversial areas are highlighted, and voice commands are supported to locate key segments; when a new meeting mentions historical issues, related content is automatically pushed to the tablet sidebar.

[0084] 103. Compare the meeting knowledge graph with the historical meeting knowledge graph to identify conflicting content and overlay a virtual information layer on the physical interface of the conference tablet.

[0085] Specifically, in one embodiment of the present application, the incremental knowledge management module of the intelligent processing center calculates the similarity between the meeting content of the current meeting and the historical meeting, and when it is determined that the meeting topics, meeting decisions and task items of the current meeting are associated with the historical topics corresponding to the historical meetings, the associated content corresponding to the historical meetings is pushed to the sidebar of the conference tablet for reference by participants, so as to facilitate sorting out the relationship between the current meeting and the historical meetings.

[0086] This application also performs a semantic comparison between the key entities of the current meeting decision and the historical meeting decision, and calculates the difference between the current meeting decision and the historical meeting decision. It should be noted that the difference in the embodiment of the present application is specifically reflected in the budget value, time node and responsible person allocation. When the difference between the current meeting and the historical meeting exceeds the preset threshold, a conflict node is generated, and then based on the conflict node, the conflict relationship between the current meeting decision and the historical meeting decision is marked on the conference tablet, such as "This budget is inconsistent with the 2024-Q4 meeting resolution", and the specific conflict content between the current meeting and the historical meeting is displayed based on the virtual information layer for the participants to consider.

[0087] In one embodiment, the AR overlay unit of the augmented reality interaction module overlays a virtual information layer on the physical interface of the conference tablet through the built-in projection in the conference tablet or an external virtual AR device. The relevant content of the historical meeting is displayed on the virtual information layer and highlighted. The task status indication in the current meeting is obtained in real time, and the task status indication is visually displayed on the virtual information layer. It should be noted that the task status indication in the embodiment of the present application includes the task completion degree, approval status and overdue mark, such as "Completed-Pending Approval-Overdue". The dynamic version controller in this application records the full version change history of the whiteboard content through the virtual information layer, and supports participants to trace back to any discussion node by sliding the timeline.

[0088] 104. Based on the conference knowledge graph and spatiotemporal time flow, query the query association information corresponding to the semantic query request to determine the semantic contradiction, and display the historical conference evidence chain of the semantic contradiction based on the virtual information layer.

[0089] Specifically, in one embodiment of the present application, the memory enhancement service module provides an intelligent retrieval interface, which is set on the conference tablet for receiving semantic query requests. After receiving the semantic query request, the associated information associated with the request in historical meetings is queried based on the meeting knowledge graph and spatiotemporal event stream in the enterprise database. It should be noted that the request-related information in this embodiment of the application includes associated whiteboard images, voice clips, and task status.

[0090] Based on the associated whiteboard images, voice clips and task status displayed on the virtual information layer, the semantic contradiction between the current meeting's discussion content and historical decisions is determined, and the controversial clips with semantic contradictions are displayed as a heat map on the virtual information layer.

[0091] In one embodiment, the database content is composed as shown in the following table:

[0092]

[0093] Natural language queries are supported, such as "Show all open-loop risk items," which will search for technical risks, compliance risks, schedule risks, and resource risks. Whiteboard images, audio clips, and task tracking status are associated. Importance is determined using a dynamic weighted scoring model: impact coefficient (calculated based on the number of associated tasks and departments involved) × 0.4 + urgency coefficient (calculated backwards from the deadline) × 0.3 + historical recurrence probability (the number of similar risk recurrences counted from the knowledge graph) × 0.3.

[0094] Query "Display high-risk unclosed-loop technical issues", the system responds:

[0095] a. Semantic analysis: Filter conditions: type, technology; Status: Unclosed loop.

[0096] b. Graph retrieval: Returns nodes (insufficient AI model training data), associated whiteboards (lists of data requirements), and meeting snippets (recordings of engineers' discussions).

[0097] c. Status display: Marked in red, indicating that the last update by the responsible person was "Waiting for data procurement approval".

[0098] d. Task status conflict resolution: The system overrides the status to "at risk" and generates an early warning to notify the project manager.

[0099] Furthermore, the decision conflict warning unit automatically displays a chain of historical meeting evidence when it detects that the current discussion content conflicts with an existing decision. The cross-meeting correlation engine automatically generates a meeting sequence report based on topic similarity, visualizing the discussion evolution path of each topic.

[0100] In one embodiment, during a product development meeting, when a user writes "User Privacy Protection Plan" on the conference tablet, the system automatically overlays the privacy protection red line clauses determined in historical meetings, a summary of the legal department's latest compliance requirements document, and a list of unclosed tasks associated with the current topic calculated through a dynamic timeline on the AR window of the virtual information layer on the right.

[0101] The system uses keyword matching and semantic analysis to extract the privacy protection red line clauses related to the "user privacy protection plan" from historical meeting records, and displays them in the form of a list in the AR window of the virtual information layer. This helps users quickly review the consensus and decisions reached in previous meetings and avoid repeated discussions.

[0102] The Legal Department's Latest Compliance Requirements Document Summary integrates with the company's internal legal department document management system to obtain the latest compliance requirements documents in real time. Using natural language processing technology, it extracts key summary information, including the full and abbreviated name of the regulation, the effective date of the latest version of the clause, the legal jurisdiction, the types of companies subject to the regulations, mandatory technical implementation requirements, penalties, and the range of penalties. This application presents this summary information in concise text within an AR window on the virtual information layer, helping users understand the latest legal and compliance requirements that current privacy protection solutions must meet.

[0103] The dynamic timeline calculates a list of unfinished tasks associated with the current topic. Specifically, a timeline is automatically generated when a task is initiated. During task execution, actual progress is compared daily. A warning is sent to the person in charge two days before the task deadline, requiring three hours of overtime daily to complete the remaining work on time. The system identifies unfinished tasks related to the "User Privacy Protection Plan" and displays them as a list in an AR window on the virtual information layer. Each task item displays key information such as the task name, person in charge, and deadline, helping users track task progress and ensuring that all unfinished tasks are covered in the meeting discussion.

[0104] In one embodiment, when the meeting discussion mentions "project delay risk", the system performs voice recognition to trigger keyword retrieval, displays a split-screen comparison chart of the differences between historical processing solutions and the current discussion content, and automatically generates a draft email of risk escalation suggestions.

[0105] The system uses voice recognition technology to monitor meeting discussions in real time. When it detects the keyword "project delay risk," it automatically triggers a search mechanism. The system searches through multiple sources, including historical meeting records, project management documents, and risk databases, to locate records of response measures to similar risks from three months ago.

[0106] The system calculates text similarity and performs semantic analysis on historical solutions and current discussions, extracting differences between the two. This difference is then displayed on the tablet screen as a split-screen comparison chart, visually demonstrating the similarities and differences between historical solutions and current discussions in terms of risk assessment, response measures, and responsibility allocation, helping users quickly understand the pros and cons of different solutions.

[0107] The system automatically generates a draft email with risk escalation recommendations based on meeting discussion content and historical data, combined with a pre-defined risk assessment knowledge base: historical case studies, failure case studies, efficacy of interventions, industry benchmarks, and the company's traffic light rules. The draft email includes a link to the meeting excerpt, allowing recipients to directly access detailed information from the meeting, improving the accuracy and efficiency of information transmission.

[0108] In one example, the current risk score was detected as 85, exceeding the company's stipulated threshold of 70. Comparing this with industry data revealed that peers encountered risks of this magnitude and 100% escalated the issue. A review of historical records revealed that a similar risk had not been escalated six months earlier, resulting in project losses. Company regulations stipulate that "cross-departmental resource coordination must be approved by the Director." The system automatically generated a recommendation: "According to Rule 8, it is recommended that this matter be immediately escalated to the Project Management Office, requiring coordination with the Procurement and Sales departments."

[0109] The above is an embodiment of the method proposed in this application. Based on the same inventive concept, this application embodiment also provides a conference memory enhancement device for a conference tablet, the structure of which is as follows: Figure 2 shown.

[0110] Figure 2 This is a schematic diagram of the internal structure of a conference memory enhancement device for a conference tablet provided in an embodiment of the present application. Figure 2 As shown, the equipment includes:

[0111] at least one processor;

[0112] and, a memory communicatively coupled to the at least one processor;

[0113] The memory stores instructions that can be executed by at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0114] Based on the preset data collection technology, multimodal data is collected in real time and encoded into a spatiotemporal event stream according to the time sequence of the multimodal data; the multimodal data includes touch data, voice data, and perception data;

[0115] Recognize handwritten text on the conference tablet and, combined with a pre-trained natural language processing model, extract key tags from the speech to build a conference knowledge graph corresponding to the current meeting;

[0116] Compare the meeting knowledge graph with the historical meeting knowledge graph to identify conflicting content and overlay a virtual information layer on the physical interface of the conference tablet;

[0117] Based on the conference knowledge graph and spatiotemporal time flow, the query association information corresponding to the semantic query request is queried to determine the semantic contradiction, and the historical conference evidence chain of the semantic contradiction is displayed based on the virtual information layer.

[0118] The present application also provides a non-volatile computer storage medium storing computer-executable instructions. When the computer-executable instructions are executed, they can:

[0119] Based on the preset data collection technology, multimodal data is collected in real time and encoded into a spatiotemporal event stream according to the time sequence of the multimodal data; the multimodal data includes touch data, voice data, and perception data;

[0120] Recognize handwritten text on the conference tablet and, combined with a pre-trained natural language processing model, extract key tags from the speech to build a conference knowledge graph corresponding to the current meeting;

[0121] Compare the meeting knowledge graph with the historical meeting knowledge graph to identify conflicting content and overlay a virtual information layer on the physical interface of the conference tablet;

[0122] Based on the conference knowledge graph and spatiotemporal time flow, the query association information corresponding to the semantic query request is queried to determine the semantic contradiction, and the historical conference evidence chain of the semantic contradiction is displayed based on the virtual information layer.

[0123] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A conference memory enhancement method for a conference tablet, characterized in that: The method comprises: Based on a preset data acquisition technology, multimodal data is collected in real time, and the multimodal data is encoded into a spatiotemporal event stream according to the time sequence of the multimodal data; the multimodal data includes touch data, voice data, and sensory data; Recognize handwritten text on the conference tablet and, combined with a pre-trained natural language processing model, extract key tags from the speech to build a conference knowledge graph corresponding to the current meeting; Comparing the conference knowledge graph with historical conference knowledge graphs to determine conflicting content, and overlaying a virtual information layer on the physical interface of the conference tablet; Based on the conference knowledge graph and the spatiotemporal time flow, querying query association information corresponding to the semantic query request to determine semantic contradictions, and displaying the historical conference evidence chain of the semantic contradictions based on the virtual information layer; Based on the conference knowledge graph and spatiotemporal time stream, query association information corresponding to the semantic query request is queried to determine semantic contradictions, and the historical conference evidence chain of the semantic contradiction is displayed based on the virtual information layer, specifically including: Based on the intelligent retrieval interface on the conference tablet, a semantic query request is received, and based on the conference knowledge graph and spatiotemporal event stream in the enterprise database, the corresponding query-related information is searched; the query-related information includes the associated whiteboard image, voice clips and task status; Determine semantic contradictions between the discussion content of the current meeting and historical decisions based on the associated whiteboard screen, the voice clips, and the task status, and display the disputed clips with semantic contradictions as a heat map on the virtual information layer; Generate a meeting sequence report based on the similarity of topics between the current meeting and previous meetings to visualize the knowledge evolution path; Compare the conference knowledge graph with the historical conference knowledge graph to determine the conflicting content, including: If the identified meeting topics, meeting decisions, and task items of the current meeting are associated with the historical topics corresponding to the previous meetings, the corresponding related content will be pushed to the sidebar of the conference tablet; Perform semantic comparison of key entities of the current meeting decision with those of historical meeting decisions, and calculate the difference between the current meeting decision and the historical meeting decision; the difference includes budget value, time node, and responsible person allocation; When the difference exceeds a preset threshold, a conflict node is generated to mark the conflict relationship between the current meeting decision and the historical meeting decision on the conference tablet and display the corresponding conflict content.

2. A conference memory enhancement method for a conference tablet according to claim 1, characterized in that: Based on the preset data collection technology, multimodal data is collected in real time, including: Using preset touch note tracking technology, the writing trajectory, note pressure, and erasure operations on the conference tablet are collected in real time; the touch note tracking technology includes electromagnetic induction or capacitive touch technology; The direction of the sound source is detected through beamforming technology, and combined with the coordinate positioning of the infrared sensor, the identity of the speaker to be bound is determined and the speaker's voice characteristics are extracted; Comparing the voice feature with a pre-stored voiceprint template of the speaker identity to be bound, and completing the binding of the speaker identity when the similarity is greater than a confidence threshold, and associating and storing the bound voice segment and whiteboard operation record according to the timestamp; Through the camera and infrared sensor integrated in the conference tablet, the speaker and the corresponding gesture operations are identified, and the spatial distribution of each participant is determined.

3. The conference memory enhancement method for a conference tablet according to claim 1, characterized in that: Combined with a pre-trained natural language processing model, key tags in the speech are extracted to build a meeting knowledge graph corresponding to the current meeting, including: Using a pre-trained natural language processing model, the meeting topics and task trigger words in the speech are identified to extract the task items and the corresponding task responsible persons, and the task responsible persons are bound to the task items. Perform sentiment analysis on the speech, determine the controversial issues, meeting decisions, and corresponding responsible persons in the speech, and mark the controversial issues as controversial; Construct a conference knowledge graph based on a graph neural network and integrate the conference knowledge graph with the enterprise database; The meeting agenda, meeting decisions, responsible persons and related documents of the current meeting are embedded into corresponding nodes in the meeting knowledge graph, a dependency chain is generated, and each node is associated with the corresponding historical task node.

4. The conference memory enhancement method for a conference tablet according to claim 1, characterized in that: Overlaying a virtual information layer on the physical interface of the conference tablet specifically includes: Through the built-in projection or external virtual device of the conference tablet, a virtual information layer is superimposed on the physical interface of the conference tablet, and the relevant content of the historical meeting is highlighted through the virtual information layer; Obtaining task status indications in the current meeting in real time and visually displaying the task status indications on the virtual information layer; the task status indications include task completion, approval status, and overdue mark; The entire version change history of the whiteboard content is recorded on the virtual information layer, and the entire version change history can be slid back according to the time axis.

5. The conference memory enhancement method for a conference tablet according to claim 1, characterized in that: Before comparing the conference knowledge graph with the historical conference knowledge graph to determine conflicting content, the method further includes: Map the meeting responsible person to the corresponding employee ID in the employee data table in the enterprise database to perform entity alignment; The meeting decision dependencies are associated with the financial budget table and the project schedule, and the current decision is compared with the constraints of the enterprise database to generate corresponding risk path prompts.

6. The conference memory enhancement method for a conference tablet according to claim 1, characterized in that: After constructing the conference knowledge graph corresponding to the current conference, the method further includes: Analyze the responsible persons, deadlines, and dependencies of the current meeting's tasks, and associate the tasks with the corresponding whiteboard images and voice clips; The task ID of the project management in the associated enterprise database is generated to jump to the task details and a clickable hyperlink is embedded in the meeting minutes document.

7. A conference memory enhancement device for a conference tablet, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the conference memory enhancement method for a conference tablet as described in any one of claims 1-6.

8. A non-volatile computer storage medium storing computer-executable instructions, characterized in that: When the computer-executable instructions are executed, a conference memory enhancement method for a conference tablet as described in any one of claims 1 to 6 is implemented.

Images