Conference information analysis method, electronic equipment and storage medium

Abstract

The invention provides a conference information analysis method, electronic equipment and a storage medium. The method comprises the following steps: acquiring multimedia information in a conference process, wherein the multimedia information comprises audio information and / or video information; based on the multimedia information, conference information is determined, the conference information comprises a conference text and speaking times and speaking duration of each speaker participating in the conference, the conference text is composed of a plurality of text paragraphs arranged according to a time sequence, and each text paragraph of the plurality of text paragraphs corresponds to any speaker participating in the conference; based on the conference information, conference analysis data is determined, and the conference analysis data comprises at least one of the following data: discussion hotness, conference conclusion quality, conference keyword condition, conclusion contribution degree and conference efficiency; determining a conference quality score based on the conference analysis data; a conference summary is generated based on the conference information, the conference analysis data, and the conference quality score. According to the scheme, the conference content can be recorded and analyzed more comprehensively and accurately.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a conference information analysis method, electronic equipment and storage medium. Background Art

[0002] With the advancement of science and technology, natural language processing technology, meeting minutes generation technology and information labeling technology have been widely used in our daily lives. Natural language processing technology enables computers to understand and generate human language, thereby achieving smoother human-computer interaction. Meeting minutes generation technology records the content of meeting discussions for easy subsequent review and reference. Information labeling technology achieves information classification and management by adding tags to information.

[0003] Currently, the processing and analysis of meeting information generally involves converting the meeting content into text form through speech-to-text technology to obtain meeting minutes. Although the meeting minutes generated by this method can help users understand the meeting content to a certain extent, it is limited to recording and transcribing the entire meeting content and lacks in-depth analysis of the meeting information. This limitation restricts users from having a more comprehensive understanding of the meeting content and affects the user experience.

[0004] In view of this, this application is hereby filed. Summary of the invention

[0005] The present application is proposed in consideration of the above-mentioned problems. According to one aspect of the present application, a method for analyzing conference information is provided, comprising: obtaining multimedia information during a conference, the multimedia information including audio information and / or video information; based on the multimedia information, determining conference information, the conference information including conference text and the number of speeches and speech duration of each speaker participating in the conference, the conference text consisting of a plurality of text paragraphs arranged in chronological order, each of the plurality of text paragraphs corresponding to any speaker participating in the conference; based on the conference information, determining conference analysis data, the conference analysis data including at least one of the following data: discussion enthusiasm, conference conclusion quality, conference keyword situation, conclusion contribution, conference efficiency; based on the conference analysis data, determining a conference quality score; based on the conference information, the conference analysis data and the conference quality score, generating meeting minutes.

[0006] Exemplarily, the conference analysis data includes discussion enthusiasm, and the determination of the conference analysis data based on the conference information includes: calculating the ratio of the total number of speeches to the total duration of the meeting to determine the frequency of speeches, wherein the total number of speeches is the sum of the number of speeches of all speakers participating in the meeting; determining the complexity of speech content based on conference keywords in the conference text; analyzing the interaction during the meeting based on the conference text to obtain an analysis result of the interaction; performing sentiment analysis on the conference text to determine the sentiment analysis result; determining the speech density based on the ratio of the total duration of speeches to the total duration of the meeting, wherein the total duration of speeches is the sum of the duration of speeches of each speaker participating in the meeting; The discussion intensity is determined based on the complexity of the speech content, the interaction analysis results, the sentiment analysis results, the speech density and the speech frequency; and / or, the conference analysis data includes conference keyword conditions, and the conference analysis data is determined based on the conference information, including: determining the quantity and quality of conference keywords in the conference text; determining the conference keyword conditions based on the quantity and quality of conference keywords in the conference text; and / or, the conference analysis data includes the quality of conference conclusions, and the conference analysis data is determined based on the conference information, including: identifying the text paragraphs in the conference text that belong to the conclusion part to determine the content of the conclusion; analyzing the conclusion The clarity and effectiveness of the content; based on the clarity and effectiveness of the conclusion content, determine the quality of the conference conclusion; and / or, the conference analysis data includes the conclusion contribution, and the conference analysis data is determined based on the conference information, including: identifying the text paragraphs belonging to the conclusion part in the conference text to determine the conclusion content; analyzing the conclusion content to determine the conclusion theme of the conclusion content and the conclusion keywords in the conclusion content; for each speaker participating in the meeting, analyzing the speech content corresponding to the speaker in the conference text to determine the speech theme and speech keywords of the speaker; calculating the theme consistency between the speaker's speech theme and the conclusion theme; Calculate the keyword consistency between the speaker's speech keywords and the conclusion keywords; determine the speaker's speech contribution based on the topic consistency and the keyword consistency; determine the conclusion contribution based on the speech contribution of each speaker participating in the meeting; and / or, the meeting analysis data includes meeting efficiency, and the meeting analysis data is determined based on the meeting information, including: determining the meeting time efficiency based on the ratio of the planned meeting duration to the total meeting duration; determining the number of completed topics based on the meeting text; determining the topic completion rate based on the ratio of the number of completed topics to the number of scheduled topics; determining the meeting efficiency based on the meeting time efficiency and the topic completion rate.

[0007] Exemplarily, determining the meeting quality score based on the meeting analysis data includes: normalizing each data item in the meeting analysis data to standardize each data item in the meeting analysis data to the range of [0,1]; mapping the normalized meeting analysis data into a score; and calculating the sum of the scores corresponding to each data item in the meeting analysis data to determine the meeting quality score.

[0008] Exemplarily, calculating the sum of the scores corresponding to each item of data in the meeting analysis data to determine the meeting quality score includes: calculating the weighted sum of the scores corresponding to each item of data in the meeting analysis data to determine the meeting quality score; wherein the weight value corresponding to each item of data in the meeting analysis data is determined according to the meeting type.

[0009] Exemplarily, the method further includes: converting conference keywords in the conference text into keyword vectors; and inputting the keyword vectors into a pre-trained classification model to determine the conference type.

[0010] Exemplarily, based on the multimedia information, determining the conference information includes: aligning the audio information with the video information in time so that each video frame in the video information corresponds to each audio segment in the audio information; for each video frame in the video information, determining the speaker corresponding to the video frame based on the video frame and the audio segment corresponding to the video frame; and determining the number of speeches and speech duration of each speaker participating in the meeting based on the speaker corresponding to each video frame in the video information.

[0011] Exemplarily, the audio information is collected by an array microphone, which includes a plurality of microphones; determining the speaker corresponding to the video frame based on the video frames and the audio clips corresponding to the video frames includes: determining the angle range of each speaker relative to the array microphone based on the faces in the video frames; identifying the audio clips corresponding to the video frames to determine the sound source angle of the audio clips; and determining that the speaker corresponding to the angle range to which the sound source angle belongs is the speaker corresponding to the video frame.

[0012] Exemplarily, determining the angular range of each speaker relative to the array microphone based on the faces in the video frames includes: performing face recognition on the video frames to determine the faces included in the video frames; determining the speakers corresponding to the faces included in the video frames; determining the angular range of the speakers corresponding to the faces relative to the array microphone based on the positions of each face included in the video frames; and / or the array microphone includes multiple microphones arranged in a preset arrangement; identifying the audio segment corresponding to the video frames to determine the sound source angle of the audio segment includes: calculating the time difference between the multiple microphones receiving the audio segment; and determining the sound source angle of the audio segment using a sound source localization algorithm based on the positions of the multiple microphones and the time difference between the multiple microphones receiving the audio segment.

[0013] According to another aspect of the present application, an electronic device is provided, including a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to execute the computer program to implement the above method.

[0014] According to another aspect of the present application, a computer-readable storage medium is provided, storing a computer program / instruction, and the computer program / instruction implements the above method when executed by a processor.

[0015] In the above technical solution, after obtaining the meeting information, the meeting analysis data can be obtained by further analyzing the meeting information, and the meeting quality score can be determined based on the meeting analysis data. In this way, the content of the meeting can be sorted out more comprehensively. The in-depth analysis of the meeting content is achieved, so that the generated meeting minutes are not only a simple record of the meeting content, but also a refined processing and valuable summary of the meeting content. In short, this solution can record and analyze the meeting content more comprehensively and accurately, which helps to improve the efficiency and effectiveness of meeting management, and helps users to understand the meeting content more comprehensively and improve the user experience. The application of this solution not only helps to improve the quality of meeting minutes, but also helps to provide stronger support for in-depth analysis and decision-making of meeting content after the meeting.

[0016] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 A schematic flow chart showing a conference information analysis method according to an embodiment of the present application; Figure 2 A schematic block diagram of an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the present application more obvious, the example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application described in the present application, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present application.

[0020] According to the above description, the processing of meeting information in the related art is usually only to convert the meeting content into text form to obtain the meeting minutes, which lacks further analysis of the meeting information. In view of this, the present application provides a meeting information analysis method, an electronic device and a storage medium. The meeting information analysis method can achieve further analysis of the meeting information, thereby assisting users to understand the meeting content more comprehensively, thereby helping to improve the user experience. The meeting information analysis method, electronic device and storage medium are described in detail below.

[0021] According to one aspect of an embodiment of the present application, a conference information analysis method is provided. Figure 1 FIG. 1 is a schematic flow chart of a conference information analysis method according to an embodiment of the present application. Figure 1 As shown, the method may include the following steps S110, S120, S130, S140 and S150.

[0022] In step S110, multimedia information during the conference is acquired, where the multimedia information includes audio information and / or video information.

[0023] The multimedia information during the meeting can be obtained by using any one or more existing or future developed devices that can record or record audio. In some embodiments, a panoramic camera can be used to obtain video information during the meeting, and a microphone device (such as an array microphone described below) can be used to obtain audio information during the meeting.

[0024] In step S120, based on the multimedia information, the conference information is determined. The conference information includes the conference text and the number of speeches and the duration of speeches of each speaker participating in the conference. The conference text consists of multiple text paragraphs arranged in chronological order, and each text paragraph of the multiple text paragraphs corresponds to any speaker participating in the conference.

[0025] After the audio information and / or video information is obtained, the conference information may be determined according to the video information and / or audio information.

[0026] In some embodiments, the multimedia information includes audio information. In this embodiment, the audio information can be converted into text through speech-to-text technology to obtain the conference text. At the same time, the identity of the speaker corresponding to each text paragraph in the conference text can be determined by identifying the voiceprint features in the audio information. After the identification is completed, the number of speeches of each speaker and the length of speech of each speaker (the length of speech can include the length of each speech of the corresponding speaker during the meeting and the total length of speech) can be counted, thereby obtaining the conference information.

[0027] In some other embodiments, the multimedia information includes video information. In this embodiment, the conference information can be determined through the video information, for example, the conference information can be extracted from the video information through a pre-trained neural network model.

[0028] In some other embodiments, the multimedia information includes video information and audio information. In this embodiment, the video information and audio information can be combined to determine the conference information, which will not be described in detail.

[0029] In this example, the meeting information includes the meeting text and the number of speeches and speaking time of each speaker participating in the meeting. It can be understood that the number of speeches and speaking time of each speaker can be used to indicate the degree of participation of the corresponding speaker in the meeting. The meeting text can be used to fully understand the content of the meeting. This information can provide a more accurate and reliable basis for the analysis in the subsequent steps.

[0030] In the solution of this example, the conference text is composed of multiple text paragraphs arranged in chronological order, and each of the multiple text paragraphs corresponds to any speaker participating in the meeting. In other words, the conference text is segmented according to the speaker and chronological order, thereby obtaining multiple text paragraphs. In a specific embodiment, the audio information can be converted into text using ASR technology, and each speaker can be identified using an NLP model to obtain the above-mentioned conference text. Those skilled in the art can understand the specific implementation method of voice-to-text conversion, which will not be repeated here.

[0031] In step S130, based on the meeting information, the meeting analysis data is determined, and the meeting analysis data includes at least one of the following data: discussion enthusiasm, meeting conclusion quality, meeting keyword situation, conclusion contribution, and meeting efficiency.

[0032] After obtaining the meeting information, further analysis can be performed based on the meeting information to obtain meeting analysis data. The meeting analysis data may include at least one of the discussion enthusiasm, meeting conclusion quality, meeting keyword situation, conclusion contribution, and meeting efficiency. It can be understood that the discussion enthusiasm can reflect the activeness of the meeting discussion. The quality of the meeting conclusion and the meeting efficiency can reflect the overall quality and effectiveness of the meeting. The meeting keyword situation can be used to reflect the professionalism of the meeting. The conclusion contribution can be used to characterize the contribution of the speaker's speech to the final conclusion of the meeting, which helps users understand the role and influence of each speaker in the meeting. These meeting analysis data can assist users in understanding the content of the meeting more comprehensively and accurately, thereby helping to improve the efficiency and effectiveness of meeting management.

[0033] In step S140, a conference quality score is determined based on the conference analysis data.

[0034] After obtaining the conference analysis data, the conference quality can be determined based on the obtained conference analysis data. For example, each conference analysis data can be mapped to a score, and the sum of the scores corresponding to each conference analysis data is used as the conference quality score. The mapping method includes but is not limited to normalization, standardization, etc., which will not be described in detail.

[0035] In step S150, a meeting summary is generated based on the meeting information, the meeting analysis data and the meeting quality score.

[0036] After obtaining the meeting quality score, the meeting quality score, meeting analysis data and meeting information can be output as a meeting minutes so that users can have a more comprehensive understanding of the meeting content.

[0037] After obtaining the meeting information, the above technical solution can obtain the meeting analysis data by further analyzing the meeting information, and determine the meeting quality score based on the meeting analysis data. In this way, the content of the meeting can be sorted out more comprehensively. The in-depth analysis of the meeting content is achieved, so that the generated meeting minutes are not only a simple record of the meeting content, but also a refined processing and valuable summary of the meeting content. In short, this solution can record and analyze the meeting content more comprehensively and accurately, which helps to improve the efficiency and effectiveness of meeting management, and helps users to understand the meeting content more comprehensively and improve the user experience. The application of this solution not only helps to improve the quality of meeting minutes, but also helps to provide stronger support for in-depth analysis and decision-making of meeting content after the meeting.

[0038] Exemplarily, the conference analysis data includes discussion enthusiasm. Based on the conference information, the conference analysis data is determined, including: calculating the ratio of the total number of speeches to the total duration of the meeting to determine the frequency of speeches, where the total number of speeches is the sum of the number of speeches of all speakers participating in the meeting; determining the complexity of the speech content based on the conference keywords in the conference text; analyzing the interaction during the meeting based on the conference text to obtain the analysis results of the interaction; performing sentiment analysis on the conference text to determine the sentiment analysis results; determining the speech density based on the ratio of the total duration of speeches to the total duration of the meeting, where the total duration of speeches is the total duration of speeches for each speaker participating in the meeting. The sum of the length of speeches; based on the complexity of speech content, the results of interaction analysis, the results of sentiment analysis, the density of speeches and the frequency of speeches, the intensity of discussions is determined; and / or, the conference analysis data includes the situation of conference keywords. Based on the conference information, the conference analysis data is determined, including: determining the quantity and quality of conference keywords in the conference text; determining the situation of conference keywords based on the quantity and quality of conference keywords in the conference text; and / or, the conference analysis data includes the quality of conference conclusions. Based on the conference information, the conference analysis data is determined, including: identifying the text paragraphs in the conference text that belong to the conclusion part to determine the content of the conclusion; analyzing the content of the conclusion the clarity and effectiveness of the conclusion content; determine the quality of the conference conclusion based on the clarity and effectiveness of the conclusion content; and / or, the conference analysis data includes the conclusion contribution, and the conference analysis data is determined based on the conference information, including: identifying the text paragraphs belonging to the conclusion part in the conference text to determine the conclusion content; analyzing the conclusion content to determine the conclusion theme of the conclusion content and the conclusion keywords in the conclusion content; for each speaker participating in the meeting, analyzing the speech content corresponding to the speaker in the conference text to determine the speech theme and speech keywords of the speaker; calculating the theme consistency between the speaker's speech theme and the conclusion theme; calculating the keyword consistency between the speaker's speech keywords and the conclusion keywords; determining the speaker's speech contribution based on the theme consistency and keyword consistency; determining the conclusion contribution based on the speech contribution of each speaker participating in the meeting; and / or, the conference analysis data includes meeting efficiency, and the conference analysis data is determined based on the conference information, including: determining the meeting time efficiency based on the ratio of the planned meeting time to the total meeting time; determining the number of completed topics based on the conference text; determining the topic completion rate based on the ratio of the number of completed topics to the number of scheduled topics; determining the meeting efficiency based on the meeting time efficiency and the topic completion rate.

[0039] Optionally, the meeting analysis data includes discussion enthusiasm. Step S130 determines the meeting analysis data based on the meeting information, including: calculating the ratio of the total number of speeches to the total meeting duration to determine the speech frequency, wherein the total number of speeches is the sum of the number of speeches of all speakers participating in the meeting; determining the complexity of the speech content based on the meeting keywords in the meeting text; analyzing the interaction during the meeting based on the meeting text to obtain the interaction analysis results; performing sentiment analysis on the meeting text to determine the sentiment analysis results; determining the speech density based on the ratio of the total speech duration to the total meeting duration, wherein the total speech duration is the sum of the speech duration of each speaker participating in the meeting; determining the discussion enthusiasm based on the complexity of the speech content, the interaction analysis results, the sentiment analysis results, the speech density and the speech frequency.

[0040] Conference keywords in conference texts can be extracted by any existing or future developed keyword extraction method, for example, by using a TF-IDF algorithm or a RAKE algorithm.

[0041] In some embodiments, determining the complexity of the speech content based on the conference keywords in the conference text may specifically include the following steps: Calculating the ratio of the number of professional keywords in the conference keywords to the total number of conference keywords as the complexity of the speech content. In this embodiment, the number of professional keywords in multiple conference keywords can be first determined, and then the proportion of professional keywords in all conference keywords can be determined as the complexity of the speech content. The professional keywords in the conference keywords can be determined by template matching, which will not be repeated. It can be understood that professional keywords are keywords with high professionalism and complexity. The more professional keywords, the more complex the discussion content. In this embodiment, the accuracy of the results can be improved by using the proportion of professional keywords in the conference keywords as the complexity of the speech content.

[0042] In some embodiments, based on the conference text, the interaction during the conference is analyzed to obtain the interaction analysis results, which may specifically include the following steps: using NLP tools (such as regular expressions, syntax analyzers, or machine learning models) to identify text paragraphs in the conference text that belong to interactions; based on the number of identified text paragraphs that belong to interactions, counting the number of interactions; using a pre-trained neural network model to analyze the interaction depth of the text paragraphs corresponding to each interaction (for example, the interaction depth can be determined based on the number of viewpoints contained in the text paragraphs corresponding to each interaction); based on the number of interactions and the interaction depth of each interaction, generating the interaction analysis results. In a specific embodiment, the ratio of the number of interactions to the total number of speeches and the average interaction depth of each interaction can be multiplied to obtain the interaction analysis result.

[0043] In some embodiments, sentiment analysis is performed on the conference text to determine the sentiment analysis result, which may specifically include the following steps: sentiment analysis is performed on the conference text to determine the sentiment tendency corresponding to each text paragraph in the conference text; based on the sentiment tendency corresponding to each text paragraph in the conference text, the sentiment analysis result is determined. In this embodiment, sentiment analysis technology (such as VADER, TextBlob, etc.) can be used to analyze the sentiment tendency corresponding to each text paragraph in the conference text. Then, the ratio of text paragraphs with positive sentiment tendency (i.e., positive and intense emotions) to the total number of text paragraphs can be calculated as the sentiment analysis result.

[0044] In this optional solution, the frequency of speech, the complexity of speech content, the results of interaction analysis, the results of sentiment analysis, and the density of speech can be determined first. It can be understood that these indicators are positively correlated with the enthusiasm of the discussion. Therefore, the enthusiasm of the discussion can be determined more accurately through these indicators. For example, the sum (arithmetic sum or weighted sum) of these indicators can be used as the enthusiasm of the discussion. In a specific embodiment, the frequency of speech, the complexity of speech content, the results of interaction analysis, the results of sentiment analysis, and the density of speech can be normalized respectively so that they are all in the interval of [0,1]. Then, the sum (arithmetic sum or weighted sum) of the normalized frequency of speech, the complexity of speech content, the results of interaction analysis, the results of sentiment analysis, and the density of speech can be calculated to obtain the enthusiasm of the discussion.

[0045] Optionally, the conference analysis data includes conference keyword situations. Based on the conference information, the conference analysis data is determined, including: determining the quantity and quality of conference keywords in the conference text; and determining the conference keyword situation based on the quantity and quality of conference keywords in the conference text.

[0046] In some embodiments, the quality of the conference keywords can be represented by the proportion of professional keywords in the conference keywords. The method for determining the proportion of professional keywords in the conference keywords has been described in detail above and will not be repeated here. After determining the quantity and quality of the conference keywords, the two can be multiplied to obtain the conference keyword situation. The conference keyword situation can also be referred to as the conference keyword evaluation result.

[0047] Optionally, the meeting analysis data includes the quality of the meeting conclusions. Based on the meeting information, the meeting analysis data is determined, including: identifying text paragraphs in the meeting text that belong to the conclusion part to determine the content of the conclusion; analyzing the clarity and effectiveness of the conclusion content; and determining the quality of the meeting conclusion based on the clarity and effectiveness of the conclusion content.

[0048] In some embodiments, the step of identifying the text paragraphs belonging to the conclusion part in the conference text can be determined by keyword recognition technology. For example, keywords such as "summary", "conclusion", "decision" and the like used to represent the conclusion can be identified in each text paragraph, and the paragraphs containing the keywords used to represent the conclusion belong to the conclusion part.

[0049] After obtaining the conclusion content, the clarity and effectiveness of the conclusion content can be analyzed using a pre-trained NLP model to obtain a clarity score and an effectiveness score. Then, the quality of the meeting conclusion can be determined based on the clarity score and the effectiveness score. For example, the sum of the clarity score and the effectiveness score can be used as the quality of the meeting conclusion.

[0050] In a specific embodiment, a scoring standard can be established in advance: the clarity score is 5 points, and the effectiveness score is 5 points. Then, the pre-trained NLP model is used to output the clarity score and effectiveness score of the conclusion content. Then, the two scores are added to determine the quality of the meeting conclusion.

[0051] Optionally, the meeting analysis data includes conclusion contribution. Based on the meeting information, the meeting analysis data is determined, including: identifying text paragraphs belonging to the conclusion part in the meeting text to determine the conclusion content; analyzing the conclusion content to determine the conclusion theme of the conclusion content and the conclusion keywords in the conclusion content; for each speaker participating in the meeting, analyzing the speech content corresponding to the speaker in the meeting text to determine the speech theme and speech keywords of the speaker; calculating the theme consistency between the speaker's speech theme and the conclusion theme; calculating the keyword consistency between the speaker's speech keywords and the conclusion keywords; based on the theme consistency and keyword consistency, determining the speaker's speech contribution; based on the speech contribution of each speaker participating in the meeting, determining the conclusion contribution.

[0052] The specific method of identifying the text paragraphs belonging to the conclusion part in the conference text has been described in detail above and will not be repeated here.

[0053] In the scheme of this example, a topic model such as an LDA model can be used to identify the conclusion topic of the conclusion content and the speaker's speech topic, and the TF-IDF or RAKE algorithm can be used to extract the conclusion keywords in the conclusion content and the speaker's speech keywords. Then, the topic consistency between the speech topic and the conclusion topic (this consistency can be represented by the cosine similarity of the text vector, which is not repeated here) and the keyword consistency between the speech keywords and the conclusion keywords (this consistency can be the ratio of the number of overlaps between the conclusion keywords in the conclusion content and the speaker's speech keywords to the total number of conclusion keywords) can be calculated, and the sum (arithmetic sum or weighted sum) of the topic consistency and the keyword consistency is used as the speech contribution of the corresponding speaker.

[0054] After obtaining the speech contribution of each speaker, the conclusion contribution can be determined based on the speech contribution of each speaker participating in the meeting. For example, the average of the speech contribution of each speaker can be used as the conclusion contribution. For another example, the maximum speech contribution can be used as the conclusion contribution.

[0055] Optionally, the meeting analysis data includes meeting efficiency. The meeting analysis data is determined based on the meeting information, including: determining the meeting time efficiency based on the ratio of the planned meeting duration to the total meeting duration; determining the number of completed topics based on the meeting text; determining the topic completion rate based on the ratio of the number of completed topics to the number of scheduled topics; and determining the meeting efficiency based on the meeting time efficiency and the topic completion rate.

[0056] In this example, the planned meeting duration and the number of scheduled topics can be pre-set. After the meeting, the meeting time efficiency can be determined based on the ratio of the planned meeting duration in the meeting information to the total meeting duration. If the meeting time efficiency is greater than or equal to 1, it means that the meeting time efficiency is high, otherwise, the meeting time efficiency is low.

[0057] In addition, the topic completion rate can also be determined based on the ratio of the number of completed topics to the number of scheduled topics. In this embodiment, the conference text can be identified by methods such as keyword extraction, template matching, named entity recognition, etc. to identify the topics included in the conference text, thereby obtaining the number of completed topics.

[0058] After obtaining the meeting time efficiency and the topic completion rate, the meeting efficiency can be further determined. For example, the sum or average of the meeting time efficiency and the topic completion rate can be used as the meeting efficiency.

[0059] The above technical solution can more accurately determine at least one of the following data based on the meeting information, such as discussion enthusiasm, meeting conclusion quality, meeting keywords, conclusion contribution, meeting efficiency, etc. These data can characterize the meeting quality from different dimensions. By determining these data, users can be assisted to understand the meeting content more comprehensively and accurately, thereby helping to improve the user's management efficiency of the meeting content.

[0060] Exemplarily, based on the meeting analysis data, determining the meeting quality score includes: normalizing each data item in the meeting analysis data to standardize each data item in the meeting analysis data to the range of [0,1]; mapping the normalized meeting analysis data into a score; and calculating the sum of the scores corresponding to each data item in the meeting analysis data to determine the meeting quality score.

[0061] It is understandable that different data have different numerical ranges. In the solution of this example, by standardizing each data item to the range of [0,1] and mapping it to a score, it is possible to facilitate the calculation of the meeting quality score in the subsequent process, reduce the calculation complexity, and improve the calculation efficiency.

[0062] The following uses meeting efficiency as an example to illustrate the process of mapping the above-mentioned data into scores. In this embodiment, the meeting efficiency may be first normalized so that its value is in the range of [0,1]. Then, the normalized meeting efficiency may be mapped into a meeting efficiency score. For example, meeting efficiency score = normalized meeting efficiency × 9 + 1. The process of mapping other data in the meeting analysis data into scores is similar to that of meeting efficiency and will not be described in detail.

[0063] After obtaining the score corresponding to each item of data in the conference analysis data, the sum of the scores corresponding to each item of data in the conference analysis data may be calculated to obtain the conference quality score. For example, the scores corresponding to each item of data in the conference analysis data may be directly added to obtain the conference quality score. For another example, the weighted sum of the scores corresponding to each item of data in the conference analysis data may be calculated to obtain the conference quality score.

[0064] The above technical solution can unify the dimensions of different data by normalizing the data and mapping them into scores, thereby facilitating the calculation of the conference quality score.

[0065] Exemplarily, calculating the sum of the scores corresponding to each item of data in the meeting analysis data to determine the meeting quality score includes: calculating the weighted sum of the scores corresponding to each item of data in the meeting analysis data to determine the meeting quality score; wherein the weight value corresponding to each item of data in the meeting analysis data is determined according to the meeting type.

[0066] Optionally, the conference type can be determined based on the user's selection operation. For example, the user can select the conference type according to the actual situation. The step of determining the conference quality score is performed after receiving the conference type selected by the user. Of course, the conference type can also be determined by identifying the conference text, which will not be repeated.

[0067] In a specific embodiment, the conference analysis data includes discussion enthusiasm, conference conclusion quality, conference keyword situation, conclusion contribution, and conference efficiency. Conference types include technical seminars, roundtable meetings, academic forums, and internal company meetings. The weight values ​​corresponding to different conference types are shown in Table 1 below.

[0068] Table 1 Meeting type - weight value relationship table

[0069] When the discussion enthusiasm score is 8, the meeting conclusion quality score is 9, the meeting keyword score is 8, the conclusion contribution score is 9, and the meeting efficiency score is 6, if the meeting type is a technical seminar, then the meeting quality score can be determined to be 8.1 through calculation.

[0070] The above weight values ​​corresponding to different conference types are only examples. The weight values ​​corresponding to different conference types can be set according to actual needs and will not be described in detail.

[0071] In the solution of this example, different weight values ​​can be used to calculate the meeting quality score according to the meeting type. It can be understood that different meetings have different focuses. By determining the weight value corresponding to each data in the meeting analysis data according to the meeting type, it is helpful to calculate the meeting quality score more accurately.

[0072] Exemplarily, the method also includes: converting conference keywords in the conference text into keyword vectors; and inputting the keyword vectors into a pre-trained classification model to determine the type of conference.

[0073] Optionally, the classification model can be trained in the following manner: extracting feature vectors of keywords of the sample text as sample feature vectors; inputting the sample feature vectors into the classification model to input sample classification results; optimizing the classification model based on the difference between the sample classification results and the meeting type of the sample text. The classification model can be constructed using SVM, random forest, naive Bayes, etc., which will not be described in detail.

[0074] The above scheme can determine the conference type more accurately, thereby providing a more accurate basis for calculating the conference quality score in subsequent steps.

[0075] Exemplarily, before determining the conference analysis data based on the conference information, the method further includes: preprocessing the conference text. The preprocessing process includes but is not limited to removing stop words, punctuation marks, etc. By preprocessing the conference text in advance, the subsequent calculation process can be simplified and help improve the accuracy of subsequent analysis and calculation.

[0076] Exemplarily, the method may further include step S160 of identifying key nodes of the meeting based on the meeting information.

[0077] Optionally, step S160, based on the meeting information, identifying the key nodes of the meeting, may specifically include the following steps: inputting the meeting keywords and named entities in the meeting text into the topic model to determine the theme of each text paragraph. For any two adjacent text paragraphs, when the themes of the two text paragraphs are different, the start time of the latter text paragraph is used as the key node of the meeting.

[0078] The method for determining the conference keywords in the conference text has been described in detail above and will not be repeated here. Named entities can be identified through NER technology (for example, through NER tools such as spaCy and Stanford NER). The named entities include but are not limited to names of people, places, organizations, time, etc.

[0079] Optionally, step S160, based on the meeting information, identifying the key nodes of the meeting, may also include the following steps: using event detection technology to identify events in the meeting text, and the start time of the paragraph where the identified event is located is the key node of the meeting. The specific method of event detection can be implemented by template matching or a method based on machine learning. For example, template matching can identify events such as "decision-making", "voting", and "summary" based on predefined event templates.

[0080] The above technical solution can more accurately identify the key nodes of the meeting. It can be understood that the key nodes of the meeting are important topic transition points, decision points or major events during the meeting. By identifying the key nodes of the meeting, users can fully understand the content of the meeting.

[0081] Exemplarily, step S120, determining conference information based on multimedia information, may include the following steps S121, S122, and S123.

[0082] In step S121, the audio information and the video information are aligned in time so that each video frame in the video information corresponds to each audio segment in the audio information.

[0083] In step S122, for each video frame in the video information, based on the video frame and the audio segment corresponding to the video frame, a speaker corresponding to the video frame is determined.

[0084] In step S123, based on the speaker corresponding to each video frame in the video information, the number of speeches and the duration of speeches of each speaker participating in the meeting are determined.

[0085] In this example, a timestamp can be added to each frame of data, and the timestamp alignment technology can be used to ensure that the video information and the audio information accurately correspond to each other on the time axis. It can be understood that after the audio information and the video information are aligned in time, a video frame and an audio clip belonging to the same timestamp can be obtained, and the audio clip corresponds to the video frame.

[0086] In the solution of this example, each video frame and the corresponding audio clip can be combined to determine the speaker corresponding to each video frame. Thus, the speaking time, speaking times, speaking content and other information of each speaker can be determined, which can provide a more accurate basis for the analysis and calculation in the subsequent steps, for example, it can be used to calculate the total number of speeches mentioned above.

[0087] Exemplarily, audio information is collected by an array microphone, which includes multiple microphones; based on the video frames and the audio clips corresponding to the video frames, the speaker corresponding to the video frames is determined, including: based on the faces in the video frames, the angle range of each speaker relative to the array microphone is determined; the audio clips corresponding to the video frames are identified to determine the angle of the sound source of the audio clips; and the speaker corresponding to the angle range to which the sound source angle belongs is determined as the speaker corresponding to the video frames.

[0088] In the solution of this example, the angle range of each face relative to the array microphone, that is, the angle range of each speaker relative to the array microphone, can be determined by the position of each face identified in the current video frame relative to the array microphone. Then, the angle of the sound source can be determined by identifying the audio clip. Finally, the sound source angle can be matched with the angle range of each speaker relative to the array microphone, and the speaker corresponding to the angle range to which the sound source angle belongs is the speaker corresponding to the video frame.

[0089] The above technical solution can more accurately identify the identity of each speaker by combining video information and audio information, thereby providing a more accurate basis for analyzing the content of the meeting and calculating the quality of the meeting in subsequent steps.

[0090] Exemplarily, based on the faces in the video frames, determining the angle range of each speaker relative to the array microphone includes: performing face recognition on the video frames to determine the faces included in the video frames; determining the speakers corresponding to the faces included in the video frames; determining the angle range of the speakers corresponding to the faces relative to the array microphone based on the positions of each face included in the video frames; and / or the array microphone includes multiple microphones arranged in a preset arrangement; identifying the audio segment corresponding to the video frames to determine the sound source angle of the audio segment, including: calculating the time difference between multiple microphones receiving the audio segment; and determining the sound source angle of the audio segment using a sound source localization algorithm based on the positions of the multiple microphones and the time difference between the multiple microphones receiving the audio segment.

[0091] Alternatively, the face in the video frame may be identified by any existing or future developed face recognition technology, for example, the Haar Cascades model, the MTCNN model, the FaceNet model or the DeepFace model may be used to identify the face in the video frame.

[0092] After determining the faces included in the video frame, the spokesperson corresponding to each face may be further determined. For example, the faces included in the video frame may be compared with face data in a preset face library to determine the spokesperson identity corresponding to each face.

[0093] After determining the speaker corresponding to the face included in the video frame, the angle range of each speaker relative to the array microphone can be determined according to the position of each face relative to the array microphone.

[0094] In the solution of this example, the array microphone includes a plurality of microphones arranged in a preset arrangement. In some embodiments, the plurality of microphones may be arranged in a specific geometric shape, and the relative positions between the plurality of microphones are recorded.

[0095] In this example, multiple microphones may be used to receive audio information. For any audio segment, the time difference between multiple microphones receiving the audio segment may be calculated. In some embodiments, the time difference between multiple microphones receiving the audio segment may be calculated using a cross-correlation algorithm.

[0096] After obtaining the time difference when multiple microphones receive the audio segment, the sound source angle of the audio segment can be determined using a sound source localization algorithm based on the positions of the multiple microphones and the time difference when the multiple microphones receive the audio segment. The sound source angle may include the azimuth and elevation of the sound source.

[0097] The above technical solution can more accurately determine the angle range of each speaker relative to the array microphone and / or the sound source angle of the audio clip, thereby providing a more reliable basis for identifying the speaker corresponding to each video frame in subsequent steps.

[0098] Exemplarily, after determining that the speaker corresponding to the angle range to which the sound source angle belongs (referred to as the initial speaker for ease of distinction) is the speaker corresponding to the frame of video frame, step S122, based on the frame of video frame and the audio clip corresponding to the frame of video frame, determines the speaker corresponding to the frame of video frame, and may also include the following steps: based on the frame of video frame and / or the audio clip corresponding to the frame of video frame, judging the accuracy of the determined initial speaker; step S123 is executed when the accuracy of the determined initial speaker is accurate.

[0099] In some embodiments, the accuracy of the determined initial speaker can be determined based only on the audio segment corresponding to the video frame. In this case, based on the video frame and / or the audio segment corresponding to the video frame, the accuracy of the determined initial speaker is determined, including: performing voiceprint recognition on the audio segment corresponding to the video frame to determine a voiceprint recognition result, the voiceprint recognition result is used to indicate the identity of the speaker determined by voiceprint recognition; determining whether the speaker indicated by the voiceprint recognition result is the same as the initial speaker; when the speaker indicated by the voiceprint recognition result is the same as the initial speaker, the accuracy of the determined initial speaker is accurate.

[0100] In a specific embodiment, voiceprint recognition is performed on the audio segment corresponding to the video frame to determine the voiceprint recognition result, including: extracting the actual voiceprint feature of the audio segment corresponding to the video frame; comparing the actual voiceprint feature with the sample voiceprint feature stored in the voiceprint feature library to determine the target voiceprint feature in the voiceprint feature library that has the highest matching degree with the actual voiceprint feature; and determining the speaker corresponding to the target voiceprint feature as the voiceprint recognition result. The actual voiceprint feature can be extracted using MFCC, i-vector, x-vector and other methods. No further details are given.

[0101] In other embodiments, the accuracy of the determined initial speaker can be determined based only on the video frame. In this case, the accuracy of the determined initial speaker is determined based on the video frame and / or the audio clip corresponding to the video frame, including: performing face recognition on the face corresponding to the angle range to which the sound source angle belongs to obtain a face recognition result, and the face recognition result is used to indicate the identity of the speaker corresponding to the face; determining whether the speaker indicated by the face recognition result is the same as the initial speaker; when the speaker indicated by the face recognition result is the same as the initial speaker, the accuracy of the determined initial speaker is accurate.

[0102] In some other embodiments, the accuracy of the determined initial speaker can be determined based on the video frame and the audio segment corresponding to the video frame. In this embodiment, when the speaker indicated by the voiceprint recognition result is the same as the initial speaker and the speaker indicated by the face recognition result is the same as the initial speaker, the accuracy of the determined initial speaker is accurate.

[0103] The above technical solution can improve the accuracy of the result by verifying the identity of the determined initial speaker using the video frame and / or the audio segment corresponding to the video frame, thereby providing a more accurate basis for step S123.

[0104] Exemplarily, when the accuracy of the determined initial speaker is inaccurate, step S122, based on the frame of video and the audio segment corresponding to the frame of video (hereinafter referred to as the audio segment), determines the speaker corresponding to the frame of video, and may also include the following steps: respectively determining the angle matching credibility and the voiceprint recognition credibility; when the angle matching credibility is high, keeping the initial speaker unchanged; when the voiceprint recognition credibility is high, updating the speaker represented by the voiceprint recognition result to the initial speaker.

[0105] It can be understood that the angle matching credibility is the credibility of the speaker determined based on the angle range to which the sound source angle belongs. The angle matching credibility can be evaluated by indicators such as signal quality and the confidence of the algorithm used to determine the time difference of receiving the audio segment by multiple microphones (which can be called algorithm confidence). In some embodiments, the angle matching credibility is calculated in the following way: calculating the signal-to-noise ratio of the audio segment, normalizing the signal-to-noise ratio to the interval [0,1] to obtain the signal quality; calculating the cross-correlation function between the audio signals received by each of the multiple microphones, obtaining the peak value, and normalizing the peak value to the interval [0,1] to obtain the algorithm confidence; determining the angle matching credibility based on the weighted sum of the signal quality and the algorithm confidence.

[0106] In some embodiments, calculating the signal-to-noise ratio of the audio segment may include: calculating the signal-to-noise ratio by the following formula: SNR (dB)=10×log10(Psigna / Pnoise); wherein Psigna represents the power of the speech signal in the audio segment, and Pnoise represents the power of the noise signal in the audio segment.

[0107] Normalize the signal-to-noise ratio to the interval [0,1] to obtain the signal quality, including: calculating the signal quality by the following formula: CSNR = SNR - SNRmin / SNRmax - SNRmin; where SNRmin and SNRmax are the minimum reference threshold and maximum reference threshold when the audio data is standardized, which can be obtained through empirical estimation or data statistics. The specific selection can be flexibly made according to different conference environments, recording equipment performance or historical data analysis results, and this application will not elaborate on this.

[0108] Normalize the peak value to the [0,1] interval to obtain the algorithm confidence, including: Calculate the algorithm confidence by the following formula: CAlg = peak value - peak value min / peak value max - peak value min. Peak value min and peak value max are the reference upper and lower limits determined by normalization, which can be obtained by empirical estimation. They define a mapping range, and linearly scale the actual measured peak value into the [0,1] interval to obtain the algorithm confidence CAlg.

[0109] Based on the weighted sum of the signal quality and the algorithm confidence, the angle matching confidence is determined, including using the following formula to determine the angle matching confidence CAngle: CAngle=αSNR×CSNR+αAlg×CAlg, where αSNR is the weight value of the signal quality and αAlg is the weight value of the algorithm confidence. In some embodiments, αSNR=0.6 and αAlg=0.4.

[0110] In the solution of this example, the voiceprint recognition credibility is used to indicate the credibility of the voiceprint recognition result determined by performing voiceprint recognition on the audio segment. The voiceprint recognition credibility can be determined based on the voice quality of the audio segment and the matching degree between the actual voiceprint feature of the audio segment and the target voiceprint feature in the voiceprint feature library.

[0111] In some embodiments, the credibility of voiceprint recognition can be calculated in the following manner: calculate the similarity between the actual voiceprint features of the audio segment and the target voiceprint features to obtain a matching score; normalize the matching score; determine the voice quality score based on the signal-to-noise ratio, speech clarity, and distortion of the audio segment, and normalize the voice quality score; determine the credibility of voiceprint recognition based on the weighted sum of the normalized matching score and the voice quality score.

[0112] In some embodiments, the matching score is normalized, including normalizing the matching score by the following formula: CMatch = matching score - score min / score max - score min, and the result of the normalization is CMatch. Score max and score min are the upper and lower limits selected for normalization of matching scores (such as voiceprint matching, keyword matching), which can be obtained by empirical preset, statistical analysis of historical data, or dynamic adjustment during the actual operation of the system. They exist to map any calculated matching score to a unified [0,1] range, so as to facilitate weighting and comparison in the subsequent calculation of credibility or comprehensive indicators.

[0113] Based on the weighted sum of the normalized matching score and the voice quality score, determining the voiceprint recognition credibility includes: determining the voiceprint recognition credibility CVoiceprint by the following formula: CVoiceprint=βMatch×CMatch+βQuality×CQuality, where CQuality represents the voice quality score, βMatch is the weight value of the matching score, and βQuality is the weight value of the voice quality score. In some embodiments, βMatch=0.7, βQuality=0.3.

[0114] According to another aspect of the embodiments of the present application, an electronic device is also provided. Figure 2FIG. 1 is a schematic block diagram of an electronic device according to an embodiment of the present application. Figure 2 As shown, the electronic device 200 includes: a processor 210 and a memory 220. The memory 220 stores a computer program, and the processor 210 is used to execute the computer program to implement the above method.

[0115] According to another aspect of the embodiment of the present application, a computer-readable storage medium is also provided. The storage medium stores a computer program / instruction, and the computer program / instruction implements the above method when executed by the processor. The storage medium may include, for example, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0116] A person skilled in the art can easily understand the implementation structure, working principle and beneficial effects of the electronic device and the computer-readable storage medium by reading the above method, and will not be described in detail for the sake of brevity.

[0117] Although example embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above example embodiments are merely exemplary and are not intended to limit the scope of the present application to this. Those of ordinary skill in the art may make various changes and modifications therein without departing from the scope and spirit of the present application. All these changes and modifications are intended to be included within the scope of the present application as required by the appended claims.

[0118] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0119] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0120] In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0121] Similarly, it should be understood that in order to streamline the present application and help understand one or more of the various application aspects, in the description of the exemplary embodiments of the present application, the various features of the present application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the method of the present application should not be interpreted as reflecting the following intention: the claimed application requires more features than the features clearly stated in each claim. More specifically, as reflected in the corresponding claims, the application point is that the corresponding technical problem can be solved with less than all the features of a single disclosed embodiment. Therefore, the claims following the specific embodiment are hereby explicitly incorporated into the specific embodiment, wherein each claim itself serves as a separate embodiment of the present application.

[0122] Those skilled in the art will understand that, except for mutually exclusive features, all features disclosed in this specification (including the accompanying claims, abstract and drawings) and all processes or units of any method or device disclosed in this specification may be combined in any combination. Unless otherwise explicitly stated, each feature disclosed in this specification (including the accompanying claims, abstract and drawings) may be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0123] In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present application and form different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0124] The various component embodiments of the present application may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all functions of some modules in the electronic device according to the embodiment of the present application. The present application may also be implemented as a device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing the present application may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0125] It should be noted that the above embodiments illustrate the present application rather than limit the present application, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets should not be constructed as a limitation to the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The present application may be implemented by means of hardware including several different elements and by means of a suitably programmed computer. In a unit claim that lists several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names.

[0126] The above is only a specific implementation or description of a specific implementation of the present application, and the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. The protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A conference information analysis method, characterized in that: include: Acquiring multimedia information during a conference, wherein the multimedia information includes audio information and / or video information; Determine conference information based on the multimedia information, the conference information including conference text and the number of speeches and the duration of speeches of each speaker participating in the conference, the conference text consisting of a plurality of text paragraphs arranged in chronological order, each of the plurality of text paragraphs corresponding to any speaker participating in the conference; Based on the meeting information, determining meeting analysis data, the meeting analysis data including at least one of the following data: discussion enthusiasm, meeting conclusion quality, meeting keywords, conclusion contribution, and meeting efficiency; Determining a meeting quality score based on the meeting analysis data; A meeting minutes is generated based on the meeting information, the meeting analysis data and the meeting quality score.

2. The conference information analysis method according to claim 1, characterized in that: The conference analysis data includes discussion enthusiasm, and determining the conference analysis data based on the conference information includes: Calculate the ratio of the total number of speeches to the total duration of the meeting to determine the speaking frequency, wherein the total number of speeches is the sum of the number of speeches of all speakers participating in the meeting; Determining the complexity of speech content based on conference keywords in the conference text; Based on the meeting text, analyzing the interaction during the meeting to obtain an interaction analysis result; Performing sentiment analysis on the conference text to determine a sentiment analysis result; Determine the speech density based on the ratio of the total speech duration to the total meeting duration, wherein the total speech duration is the sum of the speech durations of each speaker participating in the meeting; Determining the discussion enthusiasm based on the complexity of the speech content, the interaction analysis result, the sentiment analysis result, the speech density, and the speech frequency; and / or, The conference analysis data includes conference keyword information, and determining the conference analysis data based on the conference information includes: Determining the quantity and quality of conference keywords in the conference text; Determining the conference keyword situation based on the quantity and quality of the conference keywords in the conference text; and / or, The conference analysis data includes the quality of the conference conclusions, and the determining of the conference analysis data based on the conference information includes: Identify the text paragraphs belonging to the conclusion part in the conference text to determine the content of the conclusion; Analyze the clarity and validity of the conclusions; Determine the quality of the meeting conclusions based on the clarity and effectiveness of the conclusions; and / or, The conference analysis data includes a conclusion contribution, and determining the conference analysis data based on the conference information includes: Identify the text paragraphs belonging to the conclusion part in the conference text to determine the content of the conclusion; Analyze the conclusion content to determine the conclusion theme of the conclusion content and the conclusion keywords in the conclusion content; For each speaker participating in the conference, Analyze the speech content corresponding to the speaker in the conference text to determine the speech theme and speech keywords of the speaker; Calculate the consistency between the speaker's speech topic and the conclusion topic; Calculate the keyword consistency between the speaker's speech keywords and the conclusion keywords; Determining the speaker's speech contribution based on the topic consistency and the keyword consistency; Determine the contribution of the conclusion based on the speech contribution of each speaker participating in the meeting; and / or, The conference analysis data includes conference efficiency, and determining the conference analysis data based on the conference information includes: Determine the time efficiency of the meeting based on the ratio of the planned meeting duration to the total meeting duration; Based on the meeting text, determine the number of completed topics; Determining a topic completion rate based on the ratio of the number of completed topics to the number of scheduled topics; The meeting efficiency is determined based on the meeting time efficiency and the topic completion rate.

3. The conference information analysis method according to claim 1, characterized in that: Determining the conference quality score based on the conference analysis data includes: Normalizing each item of data in the conference analysis data to standardize each item of data in the conference analysis data to a range of [0, 1]; Mapping the normalized conference analysis data into scores; The sum of the scores corresponding to each item of data in the conference analysis data is calculated to determine the conference quality score.

4. The conference information analysis method according to claim 3, characterized in that: The calculating the sum of the scores corresponding to each item of data in the conference analysis data to determine the conference quality score includes: Calculating a weighted sum of the scores corresponding to each item of data in the conference analysis data to determine the conference quality score; The weight value corresponding to each item of the conference analysis data is determined according to the conference type.

5. The conference information analysis method according to claim 4, characterized in that: The method further comprises: Converting conference keywords in the conference text into keyword vectors; The keyword vector is input into a pre-trained classification model to determine the meeting type.

6. The conference information analysis method according to any one of claims 1 to 5, characterized in that: Determining conference information based on the multimedia information includes: Aligning the audio information with the video information in time so that each video frame in the video information corresponds to each audio segment in the audio information; For each video frame in the video information, based on the video frame and the audio segment corresponding to the video frame, determine a speaker corresponding to the video frame; Based on the speaker corresponding to each video frame in the video information, the number of speeches and the duration of speeches of each speaker participating in the meeting are determined.

7. The conference information analysis method according to claim 6, characterized in that: The audio information is collected by an array microphone, wherein the array microphone includes a plurality of microphones; and determining the speaker corresponding to the video frame based on the video frame and the audio segment corresponding to the video frame includes: Based on the face in the video frame, determining the angle range of each speaker relative to the array microphone; Identify the audio segment corresponding to the video frame to determine the sound source angle of the audio segment; It is determined that the speaker corresponding to the angle range to which the sound source angle belongs is the speaker corresponding to the frame of video.

8. The conference information analysis method according to claim 7, characterized in that: The determining, based on the face in the video frame, the angle range of each speaker relative to the array microphone comprises: Performing face recognition on the video frame to determine the face included in the video frame; Determine a speaker corresponding to the face included in the video frame; According to the position of each face included in the video frame, determining the angle range of the speaker corresponding to the face relative to the array microphone; and / or, The array microphone includes a plurality of microphones arranged in a preset arrangement; the identifying the audio segment corresponding to the video frame to determine the sound source angle of the audio segment includes: Calculate the time difference when the multiple microphones receive the audio segment; Based on the positions of the multiple microphones and the time difference between the multiple microphones receiving the audio segment, a sound source angle of the audio segment is determined using a sound source localization algorithm.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: A computer program / instruction is stored, and when the computer program / instruction is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images