Conference quality evaluation method, system, device, storage medium and program product
Patent Information
- Application Number
- CN202610965386.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-30
- Publication Date
- 2026-09-15
Smart Images
Figure CN122760018A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a method, system, device, storage medium, and program product for assessing meeting quality. Background Technology
[0002] Meetings are a core vehicle for information transmission and collaborative decision-making within organizations. With the expansion of enterprise scale and the widespread adoption of remote work, the number of meetings continues to grow, making meeting efficiency management an increasingly important issue for organizational operations. To improve meeting management efficiency, existing technologies have developed two types of meeting support tools: one is scheduling tools, which provide meeting scheduling, reminders, and participant management functions, used to record basic meeting metadata; the other is audio-to-text tools, which use automatic speech recognition technology to convert meeting audio into text, generating meeting minutes.
[0003] However, existing meeting support tools are primarily focused on recording meeting proceedings and lack the ability to analyze the quality of meeting content. In corporate meeting quality management practices, the common method for assessing meeting quality is still through manual observation or post-meeting interviews. This approach is limited by human resource investment, incurs high costs, has a narrow scope, and cannot provide a systematic evaluation of all meetings held within the company. Summary of the Invention
[0004] The purpose of this application is to provide a meeting quality assessment method, system, device, storage medium, and program product to solve the above-mentioned problems.
[0005] In a first aspect, embodiments of this application provide a meeting quality assessment method, the method comprising: acquiring meeting content data and meeting metadata of a meeting to be assessed; performing semantic analysis on the meeting content data, and outputting meeting content quality indicators for each dimension in a structured format according to preset quality inspection dimensions; determining meeting organization standard indicators based on the meeting metadata; obtaining the meeting quality assessment result of the meeting to be assessed based on the meeting content quality indicators and the meeting organization standard indicators; and pushing the meeting quality assessment result to the organizer of the meeting to be assessed.
[0006] In the implementation of the above solution, by integrating semantic analysis and metadata analysis, automated quality assessment of enterprise meetings is achieved. Compared with traditional manual assessment methods, the above method can reduce assessment costs and eliminate subjective bias. On the other hand, by directly pushing the assessment results to the meeting organizers, quality problems can be identified in a timely manner and feedback can be directed, thereby promoting continuous improvement in meeting quality.
[0007] In one implementation of the first aspect, before obtaining the meeting content data and meeting metadata of the meeting to be evaluated, the method further includes: periodically obtaining the title information and description information of each newly added meeting; performing semantic analysis on the title and description information of each meeting to obtain the meeting type of each meeting; filtering target meetings among the meetings based on preset target meeting types; filtering the meeting to be evaluated among the target meetings based on preset filtering rules; wherein the preset filtering rules include at least one of meeting duration filtering rules, number of participants filtering rules, meeting date filtering rules, test schedule filtering rules, and special scenario exclusion rules.
[0008] In the implementation of the above solution, semantic analysis of meeting text information is performed through a large language model to automatically classify meetings. Combined with preset filtering rules, meetings are screened in multiple layers to accurately locate meetings with evaluation value, avoid interference from non-meeting items and abnormal data in the evaluation results, and improve the accuracy and relevance of meeting quality evaluation.
[0009] In one implementation of the first aspect, the step of performing semantic analysis on the meeting content data and outputting meeting content quality indicators for each dimension in a structured format according to preset quality inspection dimensions includes: determining a corresponding quality inspection template based on the meeting type of the meeting to be evaluated; wherein the preset quality inspection dimensions in the quality inspection templates corresponding to different meeting types are different; the quality inspection template includes role definitions, preset quality inspection dimensions, and output format constraints; the role definitions are used to limit the preset roles when the large language model performs semantic analysis, so as to guide the large language model to perform quality evaluation on the meeting content data according to the evaluation perspective corresponding to the preset roles; filling the meeting content data of the meeting to be evaluated into preset placeholders in the quality inspection template to generate a target quality inspection instruction; based on the target quality inspection instruction, calling the large language model to perform semantic analysis on the meeting content data to obtain the evaluation analysis results of each preset quality inspection dimension; parsing the evaluation analysis results according to a preset structured format and outputting the meeting content quality indicators for each dimension.
[0010] In the implementation of the above solution, semantic analysis of meeting content data is performed by calling a large language model. Combined with preset quality inspection dimensions and structured output format constraints, the open semantic reasoning results are transformed into machine-parsable binary evaluation judgments, automatically achieving meeting content quality assessment without human intervention.
[0011] In one implementation of the first aspect, the output format constraint is used to constrain the large language model to output the evaluation and analysis results of the preset quality inspection dimension according to the preset structured labels; The preset structured tags include thinking tags, evaluation tags, and suggestion tags; wherein, the thinking tags are used to output the semantic analysis process information of the meeting content data by the large language model under the corresponding quality inspection dimension; the evaluation tags are used to output the binary evaluation judgment result of the meeting content data under the corresponding quality inspection dimension; and the suggestion tags are used to output corresponding improvement suggestion information when the binary evaluation judgment result is not passed.
[0012] In the implementation of the above scheme, the semantic analysis process, binary evaluation judgment and improvement suggestion information of the large language model are constrained by the pre-set structured label, so that the evaluation results have a fixed format and machine parsing, thereby supporting subsequent automated parsing and targeted feedback.
[0013] In one implementation of the first aspect, the step of performing semantic analysis on the meeting content data and outputting meeting content quality indicators of each dimension in a structured format according to preset quality inspection dimensions further includes: when calling the large language model fails, adding the corresponding quality inspection task to a retry queue; and performing retry processing on the quality inspection task based on the retry queue, and calling the large language model again to perform semantic analysis on the meeting content data.
[0014] In the implementation of the above scheme, the failure compensation for the quality inspection task that fails to call the large language model is achieved by using a retry queue mechanism, which helps to improve the completion rate of the quality inspection task and enhance the coverage and completeness of the evaluation results.
[0015] In one implementation of the first aspect, the preset quality inspection dimensions include a controllability dimension, a task clarification dimension, and a conclusion clarity dimension, wherein: The control dimension includes the time dimension, the speaking dimension, and the topic dimension. The time dimension is used to assess whether the meeting starts and ends on time, the speaking dimension is used to assess whether the speaking is guided in an orderly manner, and the topic dimension is used to assess whether the topic focuses on the core issues. The to-do list clarification dimension includes the task dimension, the responsible person dimension, and the time node dimension. The task dimension is used to assess whether the specific task content is clear, the responsible person dimension is used to assess whether a responsible person has been designated, and the time node dimension is used to assess whether a deadline has been set. The conclusion clarity dimension includes the core viewpoint dimension, the implementation dimension, and the consensus dimension. The core viewpoint dimension is used to assess whether the conclusion is clear, the implementation dimension is used to assess whether it is operable, and the consensus dimension is used to assess whether the participants have reached a consensus.
[0016] In the implementation of the above scheme, the quality of meeting content is evaluated from multiple levels through three primary dimensions and multiple sub-dimensions: control of the meeting, clarity of tasks, and clarity of conclusions. This covers the key links of meeting execution, task implementation, and decision-making, avoiding quality blind spots caused by a single evaluation dimension, thereby improving the objectivity of the meeting quality evaluation method.
[0017] In one implementation of the first aspect, determining the meeting organization specification indicators based on the meeting metadata includes: Based on the comparison between the timestamp of the meeting invitation and the meeting start time, the indicator for sending the meeting invitation in advance is determined. And / or, based on the non-empty verification results of the meeting description field and the attachment field, determine the material completeness index; And / or, based on the meeting recording status information, determine the criteria for enabling audio and video recording; And / or, based on the deviation rate between the actual meeting duration and the planned meeting duration, determine the pre-actual ratio deviation index; And / or, determine meeting cost metrics based on the number of attendees and the actual meeting duration.
[0018] In the implementation of the above scheme, the meeting organization standard indicators are determined by comparing the timestamp of the meeting invitation with the meeting start time, verifying the non-emptiness of the meeting description field and the attachment field, the meeting recording status information, the deviation rate between the actual meeting duration and the planned meeting duration, and the number of participants and the actual meeting duration, so as to achieve a comprehensive quantitative evaluation of meeting organization behavior.
[0019] In one implementation of the first aspect, obtaining the meeting quality assessment result of the meeting to be evaluated based on the meeting content quality indicators and the meeting organization standard indicators includes: determining the score weights of each of the meeting content quality indicators and each of the meeting organization standard indicators based on assessment configuration information; performing a weighted summation of each of the meeting content quality indicators and each of the meeting organization standard indicators based on the score weights to obtain the meeting quality assessment result of the meeting to be evaluated; wherein the assessment configuration information supports dynamic updates; and when the assessment configuration information changes, historical meetings within a preset date range are reassessed for meeting quality.
[0020] In the implementation of the above solution, the weight of each indicator score is determined by supporting dynamically updated evaluation configuration information, and the quality of historical meetings is re-evaluated when the configuration changes, so that the above method can meet the differentiated needs of different business scenarios and governance stages.
[0021] Secondly, embodiments of this application provide a meeting quality assessment system, comprising a data acquisition layer, an AI quality inspection engine, a meeting organization standard indicator calculation layer, and an assessment aggregation layer connected in sequence, wherein: The data acquisition layer is used to acquire meeting content data and meeting metadata of the meeting to be evaluated; The AI quality inspection engine is used to perform semantic analysis on the meeting content data and output the meeting content quality indicators of each dimension in a structured format according to the preset quality inspection dimensions. The meeting organization standard indicator calculation layer is used to determine the meeting organization standard indicators based on the meeting metadata. The evaluation aggregation layer is used to obtain the meeting quality evaluation result of the meeting to be evaluated based on the meeting content quality indicators and the meeting organization standard indicators.
[0022] Thirdly, embodiments of this application provide an electronic device, including: a processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other through the communication bus; the memory stores computer program instructions that can be executed by the processor, and the computer program instructions are read and executed by the processor to perform the method provided in the first aspect or any possible implementation of the first aspect.
[0023] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions, which, when read and executed by a processor, perform the method provided in the first aspect or any possible implementation thereof.
[0024] Fifthly, embodiments of this application provide a computer program product, the computer program product including a computer program, which, when executed by a processor, implements the method provided by the first aspect or any possible implementation of the first aspect.
[0025] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description, claims and drawings. Attached Figure Description
[0026] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0027] Figure 1 A flowchart illustrating the meeting quality assessment method provided in this application embodiment; Figure 2 This is a schematic diagram of the architecture of the meeting quality assessment system provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0028] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.
[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.
[0030] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.
[0031] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0032] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.
[0033] Meetings are a core collaborative method in the daily operations of enterprises; however, inefficient meetings have become a major drain on organizational productivity. In some companies, managers typically spend more than three hours a day in meetings, with frequent occurrences of fruitless meetings, unresolved discussions, and a lack of follow-up on conclusions, resulting in a significant hidden waste of human resources. Currently, meeting quality assessments mostly rely on manual observation or post-meeting interviews, which are costly, have narrow coverage, and are highly subjective, making it impossible to conduct a systematic evaluation of all meetings held within an enterprise.
[0034] In view of this, this application provides a meeting quality assessment method. This method achieves automated quality assessment of enterprise meetings by integrating semantic analysis and metadata analysis. Compared with traditional manual assessment methods, the above method can reduce assessment costs and eliminate subjective bias. On the other hand, by directly pushing the assessment results to the meeting organizer, quality problems can be identified in a timely manner and feedback can be directed, thereby promoting continuous improvement of meeting quality.
[0035] Please see Figure 1 The diagram illustrates a flowchart of the meeting quality assessment method provided in this application embodiment. The meeting quality assessment method provided in this application embodiment can be applied to electronic devices, which may include physical devices such as servers, PCs, tablets, or smartphones, or virtual devices such as virtual machines or containers. The electronic device can be a single device, a combination of multiple devices, or a cluster of a large number of devices. The above-mentioned meeting quality assessment method may include: Step S110: Obtain the meeting content data and meeting metadata of the meeting to be evaluated.
[0036] The electronic devices used to perform the aforementioned meeting quality assessment methods can interface with enterprise schedule management systems, video conferencing platforms, and audio transcription services to automatically collect data throughout the entire lifecycle of the meetings to be evaluated. The meeting content data includes the transcribed text of the meeting recordings generated by automatic speech recognition, serving as the core input for subsequent semantic analysis. The meeting metadata includes basic schedule information and participation behavior data. Basic schedule information includes the meeting title, description, planned start and end times, number of invited participants, and meeting organizer; participation behavior data includes the timestamp of the invitation sent, actual start and end times, actual participation time for each participant, and meeting recording status. It should be noted that all the above data is collected and aggregated under the premise of ensuring enterprise data security and privacy compliance, thus providing a complete, consistent, and traceable data foundation for subsequent intelligent classification, quality assessment, and governance closed-loop.
[0037] After acquiring the meeting content data and meeting metadata, the data can be preprocessed. For example, noise filtering can be applied to the transcript of the meeting content data to remove redundant spoken words, repetitive sentences, and non-semantic segments, and to correct obvious transcription errors that may occur during automatic speech recognition. The meeting metadata can also be standardized by unifying the timestamp format, participant statistics, and meeting duration calculation logic. These preprocessing operations can be performed based on pre-configured preprocessing rules and support anomaly data marking and completion mechanisms.
[0038] It is understood that electronic devices performing the above-described meeting quality assessment method periodically retrieve all recently added meetings from the enterprise's scheduling management system. However, not all of these meetings require meeting quality assessment; for example, some offline meetings do not. Therefore, this application provides the following solution: Optionally, prior to step S110, the meeting quality assessment method further includes: periodically acquiring the title and description information of each newly added meeting; performing semantic analysis on the title and description information of each meeting to obtain the meeting type of each meeting; wherein, the meeting type includes non-meeting, offline meetings, management meetings, training meetings, debriefing meetings, requirement meetings, routine meetings, communication and discussion meetings, and other types of meetings; selecting target meetings from among the meetings based on preset target meeting types; and selecting meetings to be evaluated from among the target meetings based on preset filtering rules; wherein, the preset filtering rules include at least one of the following: meeting duration filtering rules, number of participants filtering rules, meeting date filtering rules, test schedule filtering rules, and special scenario exclusion rules.
[0039] Electronic devices that perform the above-mentioned meeting quality assessment methods can periodically poll the enterprise schedule management system at preset time intervals (such as every working day or every week) to obtain the latest created or modified meeting schedule entries.
[0040] Meetings can be categorized into the following nine types: (1) Non-meeting: refers to arrangements that occupy the calendar but are not actually meetings, such as personal focus time or business trip records; (2) Offline meetings: include offline activities or meetings held on-site; (3) Management meetings: include meetings with management characteristics such as performance appraisal, promotion review, and one-on-one communication; (4) Training meetings: include internal training meetings, external training meetings, and knowledge sharing meetings; (5) Debriefing meetings: include project debriefing meetings and business debriefing meetings; (6) Requirements meetings: include meetings related to product or development requirements such as Market Requirement Document (MRD) review and functional requirement discussion; (7) Regular meetings: include periodically held regular meetings such as weekly meetings and monthly meetings; (8) Communication and discussion meetings: include non-periodic discussion meetings such as special topic communication and cross-departmental coordination; (9) Other meetings: include meetings that cannot be classified into any of the above categories and are uniformly classified as other meetings. Other types of meetings can be categorized using a combination of fixed-label whitelist verification and output rule verification. If a meeting cannot be classified into any of the above categories, or if it violates formatting rules, or if it has multiple labels, empty values, or other anomalies, it can be classified as "other". Alternatively, the type confidence score from the large language model output and a preset confidence threshold can be used to further determine whether a meeting needs to be classified as "other".
[0041] After obtaining the title and description information for each newly added meeting, a Large Language Model (LLM) is invoked to perform semantic analysis on the titles and descriptions of each meeting to determine their meeting type. The LLM receives a fixed Prompt structure consisting of task instructions, definitions of each meeting type, a keyword library for each meeting type, input content, and a strongly constrained output format. It then combines the meeting title and description for semantic understanding and outputs a unique classification result. The keyword library can be derived from business expert analysis, historical annotated data mining, and iterative analysis of online misclassification cases. The keyword library is maintained hierarchically by meeting type, serving as an auxiliary reference for semantic classification.
[0042] After classifying the meetings, target meetings are filtered from all newly added meetings based on preset target meeting types, such as communication and discussion meetings, routine meetings, and requirement-based meetings. Once target meetings are identified, they can be further filtered based on preset filtering rules. The preset filtering rules include at least one of the following: meeting duration filtering rules, number of participants filtering rules, meeting date filtering rules, test schedule filtering rules, and special scenario exclusion rules. Among them: (1) Meeting duration filtering rules: used to filter meetings that are too short to form an effective agenda or too long to cause data abnormalities, such as filtering out meetings with a duration of less than 10 minutes or more than 8 hours; (2) Number of participants filtering rules: used to identify substantive collective collaborative activities and avoid including one-on-one communication or personal schedules in the scope of meeting quality assessment, such as filtering out meetings with fewer than 4 invited or actual participants; (3) Meeting date filtering rules: used to limit the time range of the assessment, usually prioritizing meetings on weekdays and excluding informal or temporary meetings held during holidays, such as filtering out meetings that occur on holidays and retaining meetings that occur on weekdays; (4) Test schedule filtering rules: used to identify and exclude meetings used for debugging, demonstration, or technical verification. Such meetings usually have specific markings and do not reflect real business collaboration behavior. If they are not filtered, they may contaminate the assessment sample. (5) Special Scenario Exclusion Rules: These rules are used to avoid structural deviations in certain specific business processes. For example, some review meetings accessed in batches may affect the accuracy of the comparison between the actual meeting duration and the planned duration. These rules can be used to filter such deviations. In addition, it is understood that users can customize the preset target meeting types and the rule content of the above-mentioned rules, so that the meeting quality assessment method can be applied to a variety of business scenarios.
[0043] Step S120: Perform semantic analysis on the meeting content data, and output the meeting content quality indicators of each dimension in a structured format according to the preset quality inspection dimensions.
[0044] Optionally, step S120 includes: determining a corresponding quality inspection template based on the meeting type of the meeting to be evaluated; wherein the preset quality inspection dimensions in the quality inspection templates corresponding to different meeting types are different; the quality inspection template includes role definitions, preset quality inspection dimensions, and output format constraints; the role definitions are used to limit the preset roles when the large language model performs semantic analysis, so as to guide the large language model to evaluate the quality of the meeting content data according to the evaluation perspective corresponding to the preset roles; filling the meeting content data of the meeting to be evaluated into the preset placeholders of the quality inspection template to generate target quality inspection instructions; based on the target quality inspection instructions, calling the large language model to perform semantic analysis on the meeting content data to obtain the evaluation analysis results of each preset quality inspection dimension; parsing the evaluation analysis results according to the preset structured format and outputting the meeting content quality indicators of each dimension.
[0045] In the meeting quality assessment process, a corresponding quality inspection template can be determined based on the meeting to be assessed. This template is a pre-built quality assessment strategy carrier for different meeting types. The quality inspection template matching mechanism enables different types of meetings to use different preset quality inspection dimensions during quality assessment, forming a differentiated quality assessment mechanism for different types of meetings.
[0046] The aforementioned quality inspection template (or quality inspection prompt) includes role definition, preset quality inspection dimensions, and output format constraints. The role definition limits the preset role of the large language model when performing semantic analysis to a conference quality assessment expert. This setting can guide the large language model to evaluate the conference quality from a professional assessment perspective. The output format constraints force the model to output the evaluation results according to the preset structured label format, thereby eliminating the format uncertainty caused by open generation.
[0047] When generating the target quality inspection instructions, the meeting content data of the meetings to be evaluated can be filled into the preset placeholders in the quality inspection template, ensuring that the model input includes a complete evaluation context and the text to be analyzed. After generating the target quality inspection instructions, the large language model interface can be called to perform semantic analysis. A lower Temperature parameter can be set when calling the interface, for example, to 0.1, to reduce the randomness of the model output and ensure consistent conclusions for similar meetings in different batches of evaluations. The aforementioned quality inspection template can be stored in the configuration center using Base64 encoding and supports hot updates.
[0048] After the large language model returns text containing evaluation conclusions for each preset quality inspection dimension, these conclusions can be further parsed according to a preset structured format. The parsing process, for example, involves using regular expressions to extract the binary evaluation results and improvement suggestions for each dimension, transforming unstructured natural language into structured meeting content quality indicators. The parsed indicator data is then stored in a table, and the quality inspection completion status is marked, providing a data foundation for subsequent comprehensive score calculations and historical data backtracking.
[0049] In addition, the aforementioned quality inspection template can also be set with example guidance content. By providing high-quality meeting cases and low-quality meeting cases as comparative examples to the large language model, the Few-shot guidance mechanism enables the large language model to more accurately understand the judgment criteria of each quality inspection dimension and improve the reliability of the evaluation conclusions.
[0050] It is understandable that a multi-round collaborative reasoning and self-correction mechanism can be introduced during the semantic analysis process using a large language model. This mechanism can be implemented as follows: after the initial LLM call generates the first round of evaluation results, these results are used as one of the inputs to construct reflective prompts, such as: "Based on the following meeting transcripts and preliminary evaluation conclusions, please check for issues such as over-inference, insufficient evidence, or bias in standard application, and output corrective suggestions." If contradictions or omissions are found, a second round of refinement analysis is triggered, re-performing local semantic parsing only on the disputed dimensions. The final meeting content quality index is generated by fusing the results of multiple rounds of reasoning, and the final judgment is determined using a voting mechanism or a confidence-weighted approach.
[0051] The following is an introduction to the aforementioned pre-defined quality inspection dimensions. Optionally, these pre-defined quality inspection dimensions include the control of the situation dimension, the clarity of the task to be completed dimension, and the clarity of the conclusion dimension, wherein: The control of the meeting includes the time dimension, the speaking dimension, and the topic dimension. The time dimension is used to assess whether the meeting starts and ends on time, the speaking dimension is used to assess whether the speaking is guided in an orderly manner, and the topic dimension is used to assess whether the topic focuses on the core issues. The clarity of tasks includes the task dimension, the responsible person dimension, and the time node dimension. The task dimension is used to assess whether the specific task content is clear, the responsible person dimension is used to assess whether a responsible person has been designated, and the time node dimension is used to assess whether a deadline has been set. The conclusion clarity dimension includes the core viewpoint dimension, the implementation dimension, and the consensus dimension. The core viewpoint dimension is used to assess whether the conclusion is clear, the implementation dimension is used to assess whether it is actionable, and the consensus dimension is used to assess whether the participants have reached a consensus.
[0052] The aforementioned control dimensions are used to assess the management quality of the meeting process. Among them, the time sub-dimensional determines whether the meeting started on time by comparing the actual start time with the planned start time, and determines whether the meeting ended on time by comparing the actual end time with the planned end time; the speech sub-dimensional is used to analyze the coherence of the speech sequence in the meeting recording transcript; and the topic sub-dimensional is used to analyze whether the meeting discussion revolves around the pre-set core topics.
[0053] The aforementioned to-do clarification dimension is used to assess the clarity of action items in meeting outputs. Among them, the task sub-dimension checks whether the meeting content clearly records the specific task content to avoid vague descriptions; the responsible person sub-dimension identifies whether a clear person in charge of the task has been assigned to perform it to prevent unclear attribution of responsibility; and the time node sub-dimension determines whether the task has an executable deadline to ensure that subsequent tracking has a time anchor.
[0054] The aforementioned clarity dimension is used to assess the clarity of the final decisions reached at the meeting. Among them, the core viewpoint sub-dimension determines whether the meeting reached a clear conclusion and identifies situations where discussions are indecisive or the conclusions are ambiguous; the implementation sub-dimension assesses whether the conclusions have a feasible operational path and avoids remaining at the level of principled statements; and the consensus sub-dimension analyzes the content of the participants' speeches to determine whether key decisions have gained the recognition and consensus of relevant parties.
[0055] The three primary dimensions (including the control of the meeting, the clarity of tasks, and the clarity of conclusions) and their secondary sub-dimensions constitute a complete framework for evaluating meeting content quality. In actual evaluation, these dimensions are encoded as evaluation instructions in a structured Prompt, and the large language model performs semantic analysis on the transcribed meeting recordings dimension by dimension. Each sub-dimension outputs a binary evaluation result, namely "yes" or "no," where "yes" indicates that the dimension passes and "no" indicates that the dimension fails. For dimensions that fail, the large language model generates specific improvement measures in the suggestion labels.
[0056] Optionally, the above output format constraints are used to constrain the large language model to output the evaluation and analysis results of the preset quality inspection dimensions according to the preset structured labels; The pre-defined structured tags include reflection tags, evaluation tags, and suggestion tags. Among them, the reflection tags are used to output the semantic analysis process information of the meeting content data by the large language model under the corresponding quality inspection dimension; the evaluation tags are used to output the binary evaluation judgment result of the meeting content data under the corresponding quality inspection dimension; and the suggestion tags are used to output the corresponding improvement suggestion information when the binary evaluation judgment result is not passed.
[0057] The above output format constraints are the boundary conditions for the generation behavior of large language models. Output format constraints can force large language models to output evaluation and analysis results dimension by dimension according to the preset structured label format, avoiding format drift and content redundancy caused by open text generation.
[0058] The aforementioned pre-defined structured labels include three levels: reflection labels, evaluation labels, and suggestion labels. The reflection label displays the semantic analysis process of the large language model on the meeting content data under the corresponding quality inspection dimension. This label makes the model's reasoning chain explicit, providing a traceable explanatory basis for the evaluation conclusions. The evaluation label displays the binary evaluation result of the meeting content data under the corresponding quality inspection dimension. This label is strictly limited to two states: "yes" or "no," where "yes" indicates passing the corresponding quality inspection dimension, and "no" indicates failing. This binary judgment mechanism eliminates the subjective differences in ambiguous scoring ranges, ensuring a unified quantitative standard and horizontal comparability for evaluation results from different meetings and different time batches. The suggestion label is activated when the evaluation label outputs "no," and is used to output corresponding improvement suggestions. For dimensions that fail in the meeting content quality indicators or meeting organization standard indicators, the large language model identifies meeting segments that do not meet expectations, summarizes the reasons for failure based on the evaluation criteria of the corresponding quality inspection dimension, and outputs specific improvement measures. The improvement suggestions are matched one-to-one with the dimensions that failed to meet the requirements, ensuring that each suggestion points to a specific meeting quality issue and provides meeting organizers with actionable optimization directions.
[0059] Optionally, step S120 above further includes: when calling the large language model fails, adding the corresponding quality inspection task to the retry queue; based on the retry queue, performing retry processing on the quality inspection task, and calling the large language model again to perform semantic analysis on the meeting content data.
[0060] It is understandable that if a call fails due to network fluctuations, interface timeouts, or service anomalies during the process of calling a large language model for semantic analysis, the corresponding quality inspection task will not be discarded, but will be recorded in the retry queue.
[0061] Afterwards, failed quality inspection tasks can be retrieved from the retry queue according to a preset strategy, and the call to the large language model can be re-initiated to perform semantic analysis on the meeting content data again. In addition, an execution limit and execution interval can be pre-configured for retry processing. If retry is performed according to the execution interval, and the large language model is still not successfully called when the execution limit is reached, an alarm message will be sent to the administrator.
[0062] Step S130: Determine meeting organization standard indicators based on meeting metadata.
[0063] Optionally, step S130 above includes: Based on the comparison between the timestamp of the meeting invitation and the meeting start time, the indicator for sending the meeting invitation in advance is determined. And / or, based on the non-empty verification results of the meeting description field and the attachment field, determine the material completeness index; And / or, based on the meeting recording status information, determine the criteria for enabling audio and video recording; And / or, based on the deviation rate between the actual meeting duration and the planned meeting duration, determine the pre-actual ratio deviation index; And / or, determine meeting cost metrics based on the number of attendees and the actual meeting duration.
[0064] The aforementioned meeting organization standards indicators are objective and quantitative, complementing the meeting content quality indicators that rely on semantic analysis. All these indicators are output using a binary judgment method, directly deriving evaluation conclusions through numerical comparison or field validation, without relying on manual judgment, thus ensuring the objectivity and reproducibility of the evaluation results.
[0065] The aforementioned early invitation sending metric is calculated by comparing the invitation sending timestamp with the meeting start time to determine the lead time between the invitation sending time and the meeting start time. For example, the lead time threshold can be set to 2 hours. When the early invitation sending time is greater than or equal to this threshold, the early invitation sending metric is determined to be "yes," indicating that the pre-meeting notification meets the requirements; if it is less than this threshold, the early invitation sending metric is determined to be "no."
[0066] The above-mentioned material completeness indicators are determined by checking the description field and attachment field in the basic schedule information. If either the description field or the attachment field contains non-empty content, the material completeness indicator is determined to be "yes", indicating that the meeting organizer provided a topic description or reference materials before the meeting; if both are empty, the material completeness indicator is determined to be "no".
[0067] The aforementioned audio and video recording enable criteria are determined by reading the meeting recording status information. For online meeting scenarios, if the audio and video recording function is detected to be enabled, the audio and video recording enable criterion is determined to be "yes"; if it is not enabled, the audio and video recording enable criterion is determined to be "no".
[0068] The aforementioned deviation ratio between planned and actual meeting duration is calculated by determining the deviation rate between the actual meeting duration and the planned meeting duration. The formula for calculating the deviation rate is: |1 - Actual Meeting Duration / Planned Meeting Duration|, where the actual duration is calculated from the actual start time and the actual end time. For example, a deviation rate threshold can be set at 10%. When the deviation rate does not exceed this threshold, the deviation ratio between planned and actual meeting duration is judged as "yes," indicating that the meeting time is controlled within a reasonable range; if it exceeds this threshold, the deviation ratio between planned and actual meeting duration is judged as "no." Taking a planned duration of 60 minutes as an example, if the actual duration is within the range of 54 to 66 minutes, it is considered to have passed the judgment, and the deviation ratio between planned and actual meeting duration is judged as "yes."
[0069] The above meeting cost indicators calculate meeting resource input in person-hours. The calculation formula is: number of participants × actual meeting duration, and the result represents the total person-hours consumed by the meeting. For example, a cost threshold can be set at 30 person-hours. If the cost does not exceed this threshold, the meeting cost indicator is judged as "yes"; if it exceeds this threshold, the meeting cost indicator is judged as "no".
[0070] Step S140: Based on the meeting content quality indicators and meeting organization standard indicators, obtain the meeting quality assessment results of the meeting to be evaluated.
[0071] Optionally, step S140 includes: determining the weight of each meeting content quality indicator and each meeting organization standard indicator based on the evaluation configuration information; performing a weighted summation of each meeting content quality indicator and each meeting organization standard indicator based on the weight of the scores to obtain the meeting quality evaluation result of the meeting to be evaluated; wherein the evaluation configuration information supports dynamic updates; when the evaluation configuration information changes, the meeting quality of historical meetings within the preset date range is re-evaluated.
[0072] The aforementioned assessment configuration information includes the weighted scores of each meeting content quality indicator and each meeting organization standardization indicator during the aggregate calculation. Before calculating the meeting quality assessment result for the meeting to be evaluated, the currently effective assessment configuration information can be read from the configuration center, and the weighted score of each dimension can be determined according to the weight allocation scheme recorded therein. After determining the weighted scores, the weighted sum of each meeting content quality indicator and each meeting organization standardization indicator can be performed. The binary assessment result of each dimension is converted into a corresponding score according to its weighted score. By accumulating the scores of each dimension, the comprehensive score of the meeting to be evaluated is obtained, generating the meeting quality assessment result. The comprehensive score can use 100 points as the full score benchmark and be presented in the form of a comprehensive quality assessment report, reflecting the overall performance of the meeting to be evaluated across all assessment dimensions.
[0073] The aforementioned assessment configuration information supports dynamic updates. The weights and assessment rules for each dimension can be adjusted online through the configuration center without interrupting the assessment process or redeploying services. When business scenarios or governance strategies change, administrators can directly modify the weight parameters in the configuration center. The adjusted configuration information takes effect immediately, and subsequent meetings entering the assessment process will calculate the overall score based on the updated weights.
[0074] Furthermore, when changes occur in the assessment configuration information, historical meeting data can be batch-recalculated. Specifically, when changes in the assessment configuration information are detected, historical meetings and their stored dimension judgment results within a specified preset date range are extracted. A weighted summation is then re-executed based on the changed score weights to update the meeting quality assessment results for the historical meetings. For example, in the initial phase of the feature's launch, historical data for meetings meeting the filtering criteria within the past two weeks can be refreshed; similarly, when subsequent indicator weight adjustments are made, batch recalculation of the comprehensive score by date range is also supported.
[0075] Step S150: Push the meeting quality assessment results to the organizer of the meeting to be assessed.
[0076] After the meeting quality assessment results are generated, they are sent to the organizers of the meetings to be assessed through preset push channels. The sent content includes the comprehensive quality assessment results, details of the items that failed in each dimension, and corresponding improvement suggestions, so that the organizers can accurately identify the specific quality problems that exist in the meeting.
[0077] The aforementioned push notifications can take the form of instant notifications. Immediately after the quality assessment of a single meeting is completed, a quality inspection report card is sent to the organizer via the enterprise instant messaging tool. This card presents the overall meeting score, the results of each dimension's assessment, and targeted improvement suggestions in a structured format, ensuring immediate delivery of the assessment conclusions and allowing the organizer to promptly understand the quality status of the single meeting. Alternatively, the push notifications can also take the form of periodic reports, summarizing all meetings under the organizer's responsibility at a fixed period (e.g., weekly), identifying various issues, and generating a personal-level meeting effectiveness analysis report, which is then pushed to the organizer in card format via the enterprise instant messaging tool.
[0078] In the above scheme, the evaluation results are transformed from data into actionable feedback information through a mechanism that pushes the evaluation results to the meeting, forming a closed-loop link from post-meeting quality evaluation to the delivery of improvement suggestions. Based on the received push content, organizers make targeted adjustments to meeting organization behavior, enabling timely identification, targeted tracking, and continuous improvement of meeting quality issues, ultimately driving iterative improvement of meeting quality at the organizational level.
[0079] Furthermore, to ensure the continuous evolution of the evaluation system, an iterative evaluation strategy mechanism based on feedback data can be established. This mechanism can be implemented as follows: whenever a meeting organizer receives an evaluation result and responds (e.g., confirming a problem, submitting a rectification explanation, or marking a false alarm), this feedback information, along with the original meeting data and intermediate evaluation states (e.g., the model's thought process), is stored in a training sample library. These high-value samples are then periodically used to incrementally train a lightweight fine-tuning model (e.g., a LoRA adapter), and the updated evaluation strategy is deployed to the A / B testing channel for parallel operation with the baseline model. By comparing the evaluation consistency, organizer adoption rate, and subsequent meeting quality improvement metrics, a decision is made on whether to set the new strategy as the default configuration.
[0080] Understandably, after acquiring the meeting content data of the meeting to be evaluated, in addition to executing the standard semantic analysis process based on preset quality inspection dimensions, a context-aware dynamic prompt generation mechanism can be introduced. This mechanism first performs preliminary semantic slicing of the meeting content data, identifying core issue paragraphs, key decision nodes, task allocation statements, and controversial discussion segments. Subsequently, combining the meeting type and the quality assessment results of similar historical meetings, it dynamically constructs differentiated prompts, guiding the large language model to adopt different analytical granularities and evaluation focuses in different semantic regions. For example, for debriefing meetings, when the model analyzes the "problem attribution" paragraph, it automatically activates the root cause analysis template, strengthening the evaluation of the completeness of the attribution logic and the rationality of the attribution. For requirement review meetings, when the "solution comparison" context is detected, a multi-option advantage / disadvantage analysis instruction is triggered, outputting the support distribution of each solution and the consensus achievement score. The above-mentioned dynamic prompt generation mechanism improves the adaptability and depth of semantic analysis by upgrading the static structured prompt to a context-sensitive dynamic prompt flow.
[0081] like Figure 2 As shown, based on the same inventive concept, this embodiment of the invention also provides a meeting quality assessment system 200, including a data acquisition layer 210, an AI quality inspection engine 220, a meeting organization standard indicator calculation layer 230, and an assessment aggregation layer 240 connected in sequence, wherein: The data acquisition layer 210 is used to acquire the meeting content data and meeting metadata of the meeting to be evaluated; AI Quality Inspection Engine 220 is used to perform semantic analysis on meeting content data and output meeting content quality indicators in a structured format according to preset quality inspection dimensions. The meeting organization standard indicator calculation layer 230 is used to determine the meeting organization standard indicators based on meeting metadata; The evaluation aggregation layer 240 is used to obtain the meeting quality evaluation results of the meeting to be evaluated based on meeting content quality indicators and meeting organization standard indicators. Application layer 250 is used to push the meeting quality assessment results to the organizers of the meetings to be evaluated.
[0082] The aforementioned meeting quality assessment system 200 also includes: The AI labeling layer is used to perform semantic analysis on the titles and descriptions of each meeting to obtain the meeting type. The meeting types include non-meeting, offline meetings, management meetings, training meetings, debriefing meetings, requirement meetings, routine meetings, communication and discussion meetings, and other types of meetings.
[0083] It should be noted that the data acquisition layer 210 can not only acquire meeting content data and meeting metadata of the meetings to be evaluated, but also acquire full lifecycle data of all meetings. The application layer 250 can also set up functional modules such as meeting quality governance dashboards, detailed reports on meeting organizers, and departmental quality rankings, thereby optimizing the user experience.
[0084] The meeting quality assessment system 200 provided in this application embodiment can realize any of the functions of the above-mentioned meeting quality assessment methods. For the method embodiment section, please refer to the method embodiment section for the way each function is realized and the working principle. The system embodiment section will not repeat it here.
[0085] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of this application. (Refer to...) Figure 3 The electronic device 300 includes a processor 310, a memory 320, and a communication interface 330. These components are interconnected and communicate with each other via a communication bus 340 and / or other forms of connection mechanism (not shown).
[0086] The memory 320 includes one or more (only one is shown in the figure), which may be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The processor 310 and other possible components may access the memory 320 to read and / or write data therein.
[0087] Processor 310 includes one or more (only one is shown in the figure), which can be an integrated circuit chip with signal processing capabilities. The processor 310 described above can be a general-purpose processor, including a central processing unit (CPU), a microcontroller unit (MCU), a network processor (NP), or other conventional processors; it can also be a special-purpose processor, including a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.
[0088] The communication interface 330 includes one or more (only one is shown in the figure) and can be used to communicate directly or indirectly with other devices to exchange data. For example, the communication interface 330 can be an Ethernet interface; it can be a mobile communication network interface, such as an interface for 3G, 4G, or 5G networks; or it can be other types of interfaces with data transmission and reception functions.
[0089] One or more computer program instructions may be stored in memory 320, and processor 310 may read and run these computer program instructions to implement the meeting quality assessment method provided in the embodiments of this application and other desired functions.
[0090] Understandable. Figure 3 The structure shown is for illustrative purposes only; the electronic device 300 may also include components that are more advanced than those shown. Figure 3 The more or fewer components shown, or having the same Figure 3 The different configurations shown. Figure 3 The components shown can be implemented using hardware, software, or a combination thereof. For example, electronic device 300 can be a single server (or other device with computing power), a combination of multiple servers, a cluster of a large number of servers, etc., and can be either a physical device or a virtual device.
[0091] This application also provides a computer-readable storage medium storing computer program instructions. These computer program instructions are read and executed by a processor to perform the meeting quality assessment method provided in this application. For example, the computer-readable storage medium can be implemented as follows: Figure 3The memory 320 in the electronic device 300, or a separate storage product (such as a USB flash drive, portable hard drive, etc.).
[0092] This application also provides a computer program product comprising computer program instructions that are read and executed by a processor to perform the meeting quality assessment method provided in this application. For example, these computer program instructions may be stored in... Figure 3 The memory 320 in the electronic device 300 is located inside the memory, or it is stored in a separate storage product (such as a USB flash drive, portable hard drive, etc.).
[0093] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. The system embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.
[0094] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0095] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0096] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.
Claims
1. A method of assessing the quality of a conference, characterized by, The method includes: Obtain meeting content data and meeting metadata for the meeting to be evaluated; Semantic analysis is performed on the meeting content data, and meeting content quality indicators for each dimension are output in a structured format according to preset quality inspection dimensions. Based on the aforementioned meeting metadata, determine the meeting organization standard indicators; Based on the meeting content quality indicators and the meeting organization standard indicators, obtain the meeting quality assessment results of the meeting to be evaluated; The meeting quality assessment results will be sent to the organizer of the meeting to be assessed.
2. The conference quality evaluation method according to claim 1, characterized by, Before acquiring the meeting content data and meeting metadata of the meeting to be evaluated, the method further includes: Regularly obtain the title and description information for each newly added meeting; Semantic analysis is performed on the titles and descriptions of each meeting to obtain the meeting type of each meeting; Based on the preset target meeting type, select the target meeting from among the various meetings; Based on preset filtering rules, meetings to be evaluated are selected from the target meetings; wherein, the preset filtering rules include at least one of the following: meeting duration filtering rules, number of participants filtering rules, meeting date filtering rules, test schedule filtering rules, and special scenario exclusion rules.
3. The conference quality evaluation method according to claim 2, characterized by, The step of performing semantic analysis on the meeting content data and outputting meeting content quality indicators for each dimension in a structured format according to preset quality inspection dimensions includes: Based on the meeting type of the meeting to be evaluated, a corresponding quality inspection template is determined; wherein, the preset quality inspection dimensions in the quality inspection template are different for different meeting types; the quality inspection template includes role definition, preset quality inspection dimensions, and output format constraints; the role definition is used to limit the preset roles when the large language model performs semantic analysis, so as to guide the large language model to evaluate the quality of the meeting content data according to the evaluation perspective corresponding to the preset role; The meeting content data of the meeting to be evaluated is filled into the preset placeholders of the quality inspection template to generate the target quality inspection instruction; Based on the target quality inspection instructions, the large language model is invoked to perform semantic analysis on the meeting content data, and the evaluation and analysis results of each preset quality inspection dimension are obtained. The evaluation and analysis results are parsed according to a preset structured format, and the meeting content quality indicators for each dimension are output.
4. The conference quality evaluation method according to claim 3, characterized by, The output format constraint is used to constrain the large language model to output the evaluation and analysis results of the preset quality inspection dimensions according to the preset structured labels; The preset structured tags include thinking tags, evaluation tags, and suggestion tags; wherein, the thinking tags are used to output the semantic analysis process information of the meeting content data by the large language model under the corresponding quality inspection dimension; the evaluation tags are used to output the binary evaluation judgment result of the meeting content data under the corresponding quality inspection dimension; and the suggestion tags are used to output corresponding improvement suggestion information when the binary evaluation judgment result is not passed.
5. The conference quality evaluation method according to claim 3, characterized by, The step of performing semantic analysis on the meeting content data and outputting meeting content quality indicators for each dimension in a structured format according to preset quality inspection dimensions also includes: When the call to the large language model fails, the corresponding quality inspection task is added to the retry queue; Based on the retry queue, the quality inspection task is retried, and the large language model is called again to perform semantic analysis on the meeting content data.
6. The conference quality assessment method according to any one of claims 1 to 5, characterized by, The preset quality inspection dimensions include the control of the field, the clarity of tasks, and the clarity of conclusions, wherein: The control dimension includes the time dimension, the speaking dimension, and the topic dimension. The time dimension is used to assess whether the meeting starts and ends on time, the speaking dimension is used to assess whether the speaking is guided in an orderly manner, and the topic dimension is used to assess whether the topic focuses on the core issues. The to-do list clarification dimension includes the task dimension, the responsible person dimension, and the time node dimension. The task dimension is used to assess whether the specific task content is clear, the responsible person dimension is used to assess whether a responsible person has been designated, and the time node dimension is used to assess whether a deadline has been set. The conclusion clarity dimension includes the core viewpoint dimension, the implementation dimension, and the consensus dimension. The core viewpoint dimension is used to assess whether the conclusion is clear, the implementation dimension is used to assess whether it is operable, and the consensus dimension is used to assess whether the participants have reached a consensus.
7. The meeting quality assessment method according to any one of claims 1-5, characterized in that, The determination of meeting organization standard indicators based on the meeting metadata includes: Based on the comparison between the timestamp of the meeting invitation and the meeting start time, the indicator for sending the meeting invitation in advance is determined. And / or, based on the non-empty verification results of the meeting description field and the attachment field, determine the material completeness index; And / or, based on the meeting recording status information, determine the criteria for enabling audio and video recording; And / or, based on the deviation rate between the actual meeting duration and the planned meeting duration, determine the pre-actual ratio deviation index; And / or, determine meeting cost metrics based on the number of attendees and the actual meeting duration.
8. The meeting quality assessment method according to any one of claims 1-5, characterized in that, The process of obtaining the meeting quality assessment result of the meeting to be evaluated based on the meeting content quality indicators and the meeting organization standard indicators includes: Based on the evaluation configuration information, the score weights of each of the meeting content quality indicators and each of the meeting organization standard indicators are determined; Based on the aforementioned score weights, the quality indicators of each meeting content and the quality indicators of each meeting organization standard are weighted and summed to obtain the meeting quality assessment result of the meeting to be evaluated. The evaluation configuration information supports dynamic updates; when the evaluation configuration information changes, the meeting quality of historical meetings within a preset date range is re-evaluated.
9. A meeting quality assessment system, characterized in that, It includes a data acquisition layer, an AI quality inspection engine, a meeting organization standard indicator calculation layer, an evaluation aggregation layer, and an application layer, which are connected in sequence. The data acquisition layer is used to acquire meeting content data and meeting metadata of the meeting to be evaluated; The AI quality inspection engine is used to perform semantic analysis on the meeting content data and output the meeting content quality indicators of each dimension in a structured format according to the preset quality inspection dimensions. The meeting organization standard indicator calculation layer is used to determine the meeting organization standard indicators based on the meeting metadata. The evaluation aggregation layer is used to obtain the meeting quality evaluation result of the meeting to be evaluated based on the meeting content quality indicators and the meeting organization standard indicators. The application layer is used to push the meeting quality assessment results to the organizer of the meeting to be assessed.
10. An electronic device, characterized in that, include: A processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; The memory stores program instructions that can be executed by the processor, and the processor can execute the method as described in any one of claims 1 to 8 by calling the program instructions.
11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, cause the computer to perform the method as described in any one of claims 1 to 8.
12. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 8.