Conference quality audit method, device, equipment, and medium based on artificial intelligence

Abstract

The present application discloses a method, device, equipment, and medium for conference quality review based on artificial intelligence, the method comprising: selecting a recording set from a recording pool in sequence, the community identifiers corresponding to the conference recordings in each selected recording set having multiple types, and the total number of types of community identifiers selected within a preset time period being a first number; inputting the recording set into an AI model in sequence to obtain a first score and sending it to a terminal, receiving feedback information; determining an abnormal score based on the first score and the corresponding feedback information, adjusting the abnormal score to obtain a second score; and retraining the AI ​​model based on the second score and the corresponding conference recording. The present application periodically and randomly extracts conference recordings from different property communities multiple times, scores them through an AI model, and implements abnormal monitoring of the output results of the AI ​​model based on feedback from terminals of different property communities and dynamic identification of the scores, so as to dynamically update the AI ​​model, improve the efficiency of conference quality review, and ensure the accuracy of the review.

Description

Technical Field

[0001] The present application relates to, but is not limited to, the field of computer technology, and in particular to an artificial intelligence-based conference quality review method, device, equipment, and medium. Background Art

[0002] Property management meetings are an important platform for decision-making, communication and coordination in property management. Meeting quality audits can encourage property management teams to pay more attention to the quality and effectiveness of meetings, thereby improving the overall management level. During the audit process, in-depth discussions and analyses can be conducted on key and difficult issues in property management, and practical solutions can be proposed to promote the continuous improvement and optimization of property management. Existing meeting audit methods rely on manual supervision and record keeping, while property companies have established property service centers in multiple communities in different cities. Each property service center has meeting management and meeting audits. Manual audits will consume a lot of labor costs, and the audit efficiency is low. Manual audits are easily affected by personal emotions and subjective factors, resulting in poor accuracy of audit results, which cannot meet the property company's needs for efficient management and accuracy of meeting audits of all property service centers. Summary of the invention

[0003] The embodiments of the present application provide a conference quality review method, device, equipment, and medium based on artificial intelligence, which can effectively improve the review efficiency and accuracy of conference quality review.

[0004] In a first aspect, an embodiment of the present application provides a conference quality review method based on artificial intelligence, which is applied to a property system, wherein the property system is communicatively connected to a first number of terminals, and different terminals correspond to different community identifiers, and the method includes:

[0005] Receiving conference recordings sent by each of the terminals, associating the community identifier with the corresponding conference recording, and saving the conference recording carrying the community identifier to a recording pool, wherein the conference recordings sent from the same terminal share the same community identifier;

[0006] Based on a preset time interval and a preset time period, a recording set is randomly selected from the recording pool in sequence, wherein the community identifiers corresponding to the conference recordings in the recording set selected each time have multiple types, and within the preset time period, the total number of types of the community identifiers selected cumulatively is the first number;

[0007] At the preset time intervals, sequentially inputting the recording set into the AI ​​model for conference quality review to obtain a first score, sending the first score to a corresponding terminal, and receiving feedback information from the corresponding terminal;

[0008] Determine an abnormal score based on all the first scores and the corresponding feedback information within the preset time period, and adjust the abnormal score based on the conference recording corresponding to the abnormal score and the feedback information to obtain a second score;

[0009] Retrain the AI ​​model based on the second score and the corresponding meeting recording.

[0010] In some embodiments, within the preset time period, the recording set includes at least a first set, a second set, and a third set, a first selection time of the first set is earlier than a second selection time of the second set, and the second selection time is earlier than a third selection time of the third set, and the recording sets are randomly selected from the recording pool in sequence, including:

[0011] At the first selection time, randomly selecting the first set from the recording pool, and recording the first community identifier corresponding to the first set, wherein the number of types of the first community identifiers is the second number;

[0012] At the second selection time, randomly selecting the second set from the recording pool, and recording the second community identifier corresponding to the second set, wherein the number of types of the second community identifiers is the second number, and the second community identifier is different from the first community identifier;

[0013] At the third selection time, the third set is selected from the recording pool, wherein the third set includes a third number of conference recordings, a third community identifier corresponding to the third set is different from the first community identifier and the second community identifier, and the third number is the number of types of community identifiers that have not been selected in the recording set.

[0014] In some embodiments, the AI ​​model includes a text conversion module, a content analysis module, and multiple target scoring modules, and different target scoring modules correspond to different target scoring item identifiers. The recording set is sequentially input into the AI ​​model for conference quality review to obtain a first score, including:

[0015] Inputting the conference recording in the recording set into the text conversion module, and outputting a corresponding first text;

[0016] Calling the content parsing module to segment the first text to obtain a plurality of second texts, and associating each of the target rating item identifiers with each of the second texts, wherein each of the target rating item identifiers corresponding to the same first text is different from each other;

[0017] Each of the target scoring modules is called to calculate a sub-score of each of the second texts, and all of the sub-scores are summed to obtain the first score.

[0018] In some embodiments, the AI ​​model includes multiple candidate scoring modules, the target scoring module belongs to the candidate scoring modules, and the method further includes:

[0019] Displaying a first interface, wherein the first interface includes a plurality of candidate fields, and different candidate fields are used to indicate different candidate rating item identifiers;

[0020] A plurality of target fields selected by a user from the plurality of candidate fields are obtained, and the candidate rating item identifiers corresponding to the target fields are determined as the target rating item identifiers.

[0021] In some embodiments, the abnormality score includes a first abnormality score, a second abnormality score, and a third abnormality score, and the feedback information determines the abnormality score based on all the first scores and the corresponding feedback information within the preset time period, including:

[0022] Selecting a score whose value exceeds a high score threshold from all the first scores and determining it as the first abnormality score;

[0023] Selecting a score whose value is lower than a low score threshold from all the first scores to determine as the second abnormal score;

[0024] A first score whose value is between the high score threshold and the low score threshold is determined as a reference score, and when feedback information corresponding to the reference score indicates that the user does not agree with the reference score, the reference score is determined as the third abnormal score.

[0025] In some embodiments, the first score includes a plurality of sub-scores, each of the sub-scores corresponds to a different scoring item, the terminal displays a second interface, sends the first score to a corresponding terminal, and receives feedback information from the corresponding terminal, including:

[0026] The property system divides the conference recording corresponding to the first score to obtain a plurality of sub-audios, wherein different sub-audios correspond to different sub-scores;

[0027] The property system sends the first score and each of the sub-audios associated with each of the sub-scores to the terminal, and the terminal displays the first score and each of the sub-audios associated with each of the sub-scores in the second interface according to the first display template, and displays an evaluation window in the second interface, wherein the evaluation window includes a thumbs-up button, a thumbs-down button, and an evaluation input box;

[0028] The terminal generates the feedback information in response to a user clicking the like button or the dislike button, and / or obtains the user's input content in the evaluation input box, and sends the feedback information to the property system.

[0029] In some embodiments, there are multiple abnormal scores, each of which corresponds to a second score, and the abnormal score is adjusted based on the conference recording and the feedback information corresponding to the abnormal score to obtain the second score, including:

[0030] Generate each abnormal record based on the conference recording and the feedback information corresponding to each abnormal score;

[0031] Displaying each of the abnormal records in the first interface according to the second display template, each of the abnormal records being associated with a score input box;

[0032] The scores input by the user in the score input boxes are determined as the second scores.

[0033] In a second aspect, an embodiment of the present application provides a control device, comprising at least one control processor and a memory for communicating with the at least one control processor; the memory stores instructions executable by the at least one control processor, and the instructions are executed by the at least one control processor so that the at least one control processor can execute the artificial intelligence-based conference quality review method as described in the first aspect.

[0034] In a third aspect, an embodiment of the present application further provides an electronic device, comprising the control device of the second aspect.

[0035] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the artificial intelligence-based conference quality review method as described in the first aspect.

[0036] The embodiment of the present application provides a conference quality review method, device, equipment, and medium based on artificial intelligence, the method comprising: receiving conference recordings sent by each of the terminals, associating the community identifier with the corresponding conference recording, and saving the conference recording carrying the community identifier to a recording pool, wherein the conference recordings sent from the same terminal share the same community identifier; based on a preset time interval and a preset time period, randomly selecting a recording set from the recording pool in sequence, wherein the community identifier corresponding to the conference recording in each selected recording set has multiple types, and within the preset time period, the total number of types of the community identifiers selected cumulatively is the first number; according to the preset time interval, inputting the recording set into an AI model in sequence for conference quality review to obtain a first score, sending the first score to the corresponding terminal, and receiving feedback information from the corresponding terminal; determining an abnormal score based on all the first scores and the corresponding feedback information within the preset time period, adjusting the abnormal score based on the conference recording corresponding to the abnormal score and the feedback information to obtain a second score; and retraining the AI ​​model based on the second score and the corresponding conference recording. According to the solution provided in the embodiment of the present application, meeting recordings from different property communities are extracted multiple times at regular intervals, and the meeting quality is reviewed through an AI model. Based on the feedback from different property community terminals and dynamic recognition of scores, abnormal monitoring of the AI ​​model output results is implemented to dynamically update the AI ​​model, thereby effectively improving the review efficiency of the meeting quality review while ensuring the accuracy of the quality review. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a flowchart of the steps of a conference quality review method based on artificial intelligence provided by an embodiment of the present application;

[0038] Figure 2 It is a structural diagram of a control device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0040] It is understood that although the functional modules are divided in the device schematic diagram and the logical order is shown in the flow chart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flow chart. The terms "first", "second", etc. in the specification, claims or the above drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0041] Property management meetings are an important platform for decision-making, communication and coordination in property management. Meeting quality audits can encourage property management teams to pay more attention to the quality and effectiveness of meetings, thereby improving the overall management level. During the audit process, in-depth discussions and analyses can be conducted on key and difficult issues in property management, and practical solutions can be proposed to promote the continuous improvement and optimization of property management. Existing meeting audit methods rely on manual supervision and record keeping, while property companies have established property service centers in multiple communities in different cities. Each property service center has meeting management and meeting audits. Manual audits will consume a lot of labor costs, and the audit efficiency is low. Manual audits are easily affected by personal emotions and subjective factors, resulting in poor accuracy of audit results, which cannot meet the property company's needs for efficient management and accuracy of meeting audits of all property service centers.

[0042] To solve the above-mentioned problems, the embodiments of the present application provide a method, device, equipment, and medium for conference quality review based on artificial intelligence, the method comprising: receiving conference recordings sent by each of the terminals, associating the community identifier with the corresponding conference recording, and saving the conference recording carrying the community identifier to a recording pool, wherein the conference recordings sent from the same terminal share the same community identifier; based on a preset time interval and a preset time period, randomly selecting a recording set from the recording pool in sequence, wherein the community identifier corresponding to the conference recording in each selected recording set has multiple types, and within the preset time period, the total number of types of the community identifiers selected cumulatively is the first number; according to the preset time interval, the recording set is input into the AI ​​model in sequence for conference quality review to obtain a first score, send the first score to the corresponding terminal, and receive feedback information from the corresponding terminal; determine an abnormal score based on all the first scores and the corresponding feedback information within the preset time period, adjust the abnormal score based on the conference recording corresponding to the abnormal score and the feedback information to obtain a second score; retrain the AI ​​model based on the second score and the corresponding conference recording. According to the solution provided in the embodiment of the present application, meeting recordings from different property communities are randomly sampled multiple times at regular intervals, and meeting quality is reviewed through an AI model. Based on feedback from terminals in different property communities and dynamic recognition of scores, abnormal monitoring of the AI ​​model output results is implemented to dynamically update the AI ​​model, thereby effectively improving the review efficiency of the meeting quality review while ensuring the accuracy of the quality review.

[0043] The embodiments of the present application are further described below in conjunction with the accompanying drawings.

[0044] refer to Figure 1 , Figure 1 It is a flowchart of the steps of a conference quality review method based on artificial intelligence provided by an embodiment of the present application. The embodiment of the present application provides a conference quality review method based on artificial intelligence. The method is applied to a property system, the property system is connected to a first number of terminals for communication, and different terminals correspond to different community identifiers. The method includes but is not limited to the following steps:

[0045] Step S10, receiving conference recordings sent by various terminals, associating the community identifier with the corresponding conference recording, and saving the conference recordings carrying the community identifier into a recording pool, wherein conference recordings sent from the same terminal share the same community identifier.

[0046] Specifically, the community identifier of this embodiment can uniquely identify the property service centers in different areas of different cities. The number of community identifiers and terminals corresponds to the number of property service centers corresponding to the property company (the number of terminals in this embodiment is the first number, and the specific value of the first number depends on the operating conditions of the property company). The property system is communicated with each terminal and can achieve centralized management of meeting audits for each property service center through information communication between the property system server and each terminal.

[0047] Specifically, the property system receives the meeting recording sent from each terminal in real time. After the morning meeting is over, the property community corresponding to each terminal sends the meeting recording of the morning meeting to the property system through the terminal.

[0048] Specifically, since the present embodiment receives the conference recordings sent by each terminal in real time, a storage space (recording pool) is allocated in the property system for storing these conference recordings, thereby providing effective support for subsequent conference quality review operations.

[0049] Specifically, the conference quality review operation of this embodiment is implemented by extracting different conference recordings from the recording pool at regular intervals. The quantity here refers to the number of types of community identifiers, and the timing refers to the preset time interval, that is, the object of each conference quality review operation is a different property service center. Within the preset time period, the conference recordings of multiple property service centers are quantitatively extracted each time according to the preset time interval for conference quality review. After the preset time period has passed, the conference quality review of all property service centers corresponding to all community identifiers corresponding to the property company is completed. In this process, the community identifier is to provide a data basis for the extraction operation. For example, the first time 2 conference recordings corresponding to community identifier 1 and 5 conference recordings corresponding to community identifier 2 are extracted, and community identifier 1 and community identifier 2 are recorded as extracted. The next extraction can avoid community identifier 1 and community identifier 2 to extract the conference recordings corresponding to community identifiers that have not been extracted.

[0050] Step S20, based on a preset time interval and a preset time period, randomly select a recording set from the recording pool in sequence, wherein the community identifiers corresponding to the conference recordings in each selected recording set have multiple types, and within the preset time period, the total number of types of community identifiers selected cumulatively is a first number.

[0051] Specifically, in this embodiment, the number of types of community identifiers corresponding to conference recordings selected from the recording pool at preset time intervals can be the same or different, so that multiple acquisitions can be performed at regular intervals.

[0052] Specifically, in one embodiment, within a preset time period, the recording set includes at least a first set, a second set, and a third set, a first selection time of the first set is earlier than a second selection time of the second set, and the second selection time is earlier than a third selection time of the third set. Figure 1 Step S20 includes but is not limited to the following steps:

[0053] Step S21, randomly selecting a first set from the recording pool at a first selection time, and recording a first community identifier corresponding to the first set, wherein the number of types of the first community identifier is a second number;

[0054] Step S22, randomly selecting a second set from the recording pool at a second selection time, and recording a second community identifier corresponding to the second set, wherein the number of types of the second community identifiers is a second number, and the second community identifier is different from the first community identifier;

[0055] Step S23, selecting a third set from the recording pool at a third selection time, wherein the third set includes a third number of conference recordings, a third community identifier corresponding to the third set is different from the first community identifier and the second community identifier, and the third number is the number of types of community identifiers that have not been selected in the recording set.

[0056] Specifically, referring to the description of the above-mentioned embodiment, the conference quality review operation of this embodiment is to select conference recordings corresponding to multiple different types of community identifiers from the recording pool according to preset time intervals within a preset time period, and randomly select conference recordings corresponding to multiple different types of community identifiers from the recording pool each time. The community identifier selected each time has not been selected in the previous recording selection operation, until full coverage selection of all types of community identifiers in the recording pool is completed within the preset time period. The following two examples are used to illustrate the recording selection operation in the recording pool. The recording pool saves meeting recordings sent from terminals. Example 1 takes 7 days as the preset time period and 3 days as the preset time interval as an example. The property system is connected to 100 terminals in communication. The first number is 100 and the second number is 40, accounting for 40% of the first number. That is to say, the meeting recordings sent by each terminal are saved in real time in the recording pool. The recording pool saves multiple meeting recordings corresponding to 100 different community identifiers. On the first day, 40 meeting recordings corresponding to different types of community identifiers are randomly selected from the recording pool as the first set. After an interval of 3 days, that is, on the fourth day, 40 meeting recordings corresponding to different types of community identifiers are randomly selected from the recording pool as the second set. The community identifiers corresponding to the second set are different from those corresponding to the first set. After an interval of 3 days, that is, on the seventh day, the meeting recordings corresponding to the remaining 20 types of community identifiers are selected as the third set, and the AI ​​model meeting quality review operation is performed on each selected recording set; Example 2 takes 7 days as the preset time period and 1 day as the preset time interval. Taking the time interval as an example, the property system is connected to 100 terminals, the first number is 100, and the second number is 15, accounting for 15% of the first number. That is to say, the meeting recordings sent by each terminal are saved in real time in the recording pool. The recording pool saves multiple meeting recordings corresponding to 100 different community identifiers. Every day, 15 meeting recordings corresponding to different types of community identifiers are randomly selected from the recording pool as a recording set, until the 7th day, the meeting recordings corresponding to the remaining 10 types of community identifiers are selected as a recording set, and a meeting quality review is performed on each selected recording set. In this way, meeting recordings from different property communities are randomly selected at a regular and quantitative manner, and meeting quality reviews covering all property communities are carried out in batches, ensuring the randomness of spot checks and the comprehensiveness of coverage, and reducing the sample size of quality reviews. Compared with the solution of comprehensively covering all meeting recordings from all terminals at one time, the efficiency of meeting quality reviews is improved, while reducing the computing cost of the property system and ensuring the fairness of the review and the accuracy of the review results.

[0057] Step S30: Input the recording set into the AI ​​model for conference quality review at preset time intervals, obtain a first score, send the first score to the corresponding terminal, and receive feedback information from the corresponding terminal.

[0058] Specifically, the conference quality review of this embodiment includes operations of voice-to-text conversion of conference recordings, content understanding, and scoring the converted text content.

[0059] In one embodiment, the AI ​​model includes a text conversion module, a content analysis module, and multiple target scoring modules. Different target scoring modules correspond to different target scoring item identifiers. Figure 1 Step S30 sequentially inputs the recording set into the AI ​​model at preset time intervals to review the conference quality and obtain a first score, including but not limited to the following steps:

[0060] Step S31, inputting the conference recording in the recording set into a text conversion module, and outputting a corresponding first text;

[0061] Step S32, calling the content analysis module to segment the first text to obtain a plurality of second texts, and associating each target rating item identifier with each second text, wherein each target rating item identifier corresponding to the same first text is different from each other;

[0062] Step S33, calling each target scoring module to calculate the sub-score of each second text, and summing up all the sub-scores to obtain a first score.

[0063] In addition, in some embodiments, the AI ​​model includes multiple candidate scoring modules, the target scoring module belongs to the candidate scoring module, and the target scoring item identifier is determined according to the following steps:

[0064] Step S61, displaying a first interface, wherein the first interface includes a plurality of candidate fields, and different candidate fields are used to indicate different candidate rating item identifiers;

[0065] Step S62: acquiring multiple target fields selected by the user from multiple candidate fields, and determining the candidate rating item identifier corresponding to the target field as the target rating item identifier.

[0066] Specifically, this embodiment does not limit specific candidate fields, and may include different scoring items such as corporate culture, must-read before work, review and analysis of property fee collection, review and analysis of reporting and complaint handling, work content arrangement (related to fees), work content arrangement (except fees), and sharing of excellent cases.

[0067] It can be understood that the meeting quality review corresponds to a target review rule, and each target scoring module corresponding to the AI ​​model corresponds to a different scoring item identifier, that is, different target scoring modules correspond to different sub-scoring rules, and each sub-scoring rule is combined to obtain a target review rule, and the target review rule can be freely combined according to actual review requirements. Specifically, the system administrator enters the first interface by logging into the property system server. The first interface of this embodiment includes multiple candidate fields, and different candidate fields are used to indicate different candidate scoring item identifiers. Each different candidate scoring item identifier corresponds to a different candidate scoring module in the AI ​​model. The system administrator selects multiple target fields from the multiple candidate fields according to the actual meeting quality review requirements. In response to this operation, the property system determines the candidate scoring item identifier corresponding to the target field as the target scoring item identifier, thereby flexibly adjusting the meeting quality review logic of the AI ​​model as needed, improving the versatility of the AI ​​model, and only one AI model needs to be maintained later. Compared with the existing AI model corresponding to the fixed operation logic solution, that is, one AI model can only realize the scoring calculation of a single scoring item. When the meeting quality review involves a combination of multiple scoring items, it is necessary to combine multiple AI models to achieve it. This implementation can reduce the maintenance cost of the AI ​​model and improve the efficiency of the meeting quality review.

[0068] In addition, in some embodiments, the first score includes multiple sub-scores, each sub-score corresponds to a different scoring item, and the terminal displays a second interface. Figure 1 Step S30 of sending the first score to the corresponding terminal and receiving feedback information from the corresponding terminal includes but is not limited to the following steps:

[0069] Step S34, the property system divides the conference recording corresponding to the first score to obtain a plurality of sub-audios, wherein different sub-audios correspond to different sub-scores;

[0070] Step S35, the property system sends the first score and each sub-audio associated with each sub-score to the terminal, and the terminal displays the first score and each sub-audio associated with each sub-score in the second interface according to the first display template, and displays an evaluation window in the second interface, wherein the evaluation window includes a thumbs-up button, a thumbs-down button, and an evaluation input box;

[0071] Step S36, the terminal generates feedback information in response to the user's click operation on the like button or the dislike button, and / or obtains the user's input content in the evaluation input box, and sends the feedback information to the property system.

[0072] It can be understood that the first score of this embodiment includes the total audit score and sub-scores corresponding to multiple different scoring items. In order to ensure the accuracy of the output results of the AI ​​model, each output score of the AI ​​model is sent to the corresponding terminal, and feedback information sent by the terminal is received. The feedback information is used as a reference factor for subsequent corrections and updates of the AI ​​model.

[0073] Specifically, this embodiment divides each conference recording according to different scoring items, obtains the sub-audio corresponding to each scoring item, and sends the sub-audio and the corresponding scoring item and sub-score to the terminal, so that the terminal displays the first score and each sub-audio associated with each sub-score in the second interface according to the first display template, and displays an evaluation window in the second interface, wherein the evaluation window includes a thumbs-up button, a thumbs-down button and an evaluation input box, and the terminal responds to the user's click operation on the thumbs-up button or the thumbs-down button, and / or obtains the user's input content in the evaluation input box, generates feedback information, and sends the feedback information to the property system. That is to say, it is possible to receive different types of feedback from the staff of the property service center corresponding to the terminal on the score calculated by the AI ​​model, providing an effective data basis for the subsequent improvement of the accuracy of the AI ​​model.

[0074] Step S40, determining an abnormal score based on all first scores and corresponding feedback information within a preset time period, adjusting the abnormal score based on the conference recording and feedback information corresponding to the abnormal score, and obtaining a second score.

[0075] It can be understood that this embodiment also sets up an anomaly detection mechanism. When the conference quality review operation on all terminals is completed within a preset time period, this embodiment also determines the abnormal score based on all the first scores and corresponding feedback information within the preset time period, and adjusts the abnormal score, so as to provide effective support for subsequent retraining and updating of the AI ​​model and improving the accuracy of the AI ​​model.

[0076] Specifically, in some embodiments, the anomaly score includes a first anomaly score, a second anomaly score, and a third anomaly score. Figure 1 Step S40 of determining the abnormality score based on all the first scores and the corresponding feedback information within the preset time period includes but is not limited to the following steps:

[0077] Step S41, selecting a score whose value exceeds a high score threshold from all first scores and determining it as a first abnormal score;

[0078] Step S42, selecting a score whose value is lower than a low score threshold from all first scores and determining it as a second abnormal score;

[0079] Step S43: determine the first score whose value is between the high score threshold and the low score threshold as the reference score; when the feedback information corresponding to the reference score indicates that the user does not agree with the reference score, determine the reference score as a third abnormal score.

[0080] It can be understood that after a preset time period has passed and a full-coverage meeting quality audit is completed for all property service centers corresponding to community identifiers, the specific steps of performing anomaly detection on the meeting quality audit results, i.e., all first scores, are as follows: selecting a score whose value exceeds the high score threshold from all first scores to be determined as a first abnormal score; selecting a score whose value is lower than the low score threshold from all first scores to be determined as a second abnormal score; determining the first score whose value is between the high score threshold and the low score threshold as a reference score, and when the feedback information corresponding to the reference score indicates that the user does not agree with the reference score, determining the reference score as a third abnormal score, that is, considering the feedback information, considering three situations: selecting the highest score, the lowest score, and the score that is greatly different from the feedback information in the first score as abnormal scores, re-reviewing these abnormal scores, and combining the re-reviewed audit results (i.e., the second score) as the data basis for subsequent retraining and updating of the AI ​​model, thereby ensuring the accuracy of the AI ​​model and the accuracy of the audit results of the meeting audit quality.

[0081] In addition, in some embodiments, there are multiple anomaly scores, each of which corresponds to a second score. Figure 1 Step S40 of adjusting the abnormality score based on the conference recording and feedback information corresponding to the abnormality score to obtain a second score includes but is not limited to the following steps:

[0082] Step S44, generating each abnormal record based on the conference recording and feedback information corresponding to each abnormal score;

[0083] Step S45, displaying each abnormal record in the first interface according to the second display template, each abnormal record being associated with a score input box;

[0084] Step S46: Determine the scores input by the user in the score input boxes as the second scores.

[0085] Specifically, this embodiment does not limit the specific format of the second display template. It can be a table that displays each abnormal record in chronological order from top to bottom according to the generation time of the conference recording. The display color of the score depends on the specific situation. For example, if the total score is 100 points, an abnormal score of more than 90 points will be displayed in yellow, and an abnormal score of less than 60 points will be displayed in red.

[0086] It is understandable that the method of adjusting the abnormality score in this embodiment can be manual review. Specifically, each abnormality record is generated based on the conference recording and feedback information corresponding to each abnormality score, and each abnormality record is displayed in the first interface according to the second display template. Each abnormality record is associated with a score input box for the system administrator to enter a new score, and the score entered by the user in each score input box is determined as each second score. That is to say, through the dual verification of the AI ​​model combined with manual review, the accuracy of the AI ​​model can be further guaranteed, and the accuracy of the review results of the meeting review quality can be guaranteed.

[0087] Step S50, retraining the AI ​​model based on the second score and the corresponding meeting recording.

[0088] In summary, this embodiment randomly extracts meeting recordings from different property communities at a regular and quantitative basis, conducts meeting quality audits through an AI model, and implements abnormal monitoring of the AI ​​model output results based on feedback from different property community terminals and dynamic identification of scores, so as to dynamically update the AI ​​model, thereby effectively improving the audit efficiency of the meeting quality audit while ensuring the accuracy of the quality audit.

[0089] like Figure 2 As shown, Figure 2 2 is a structural diagram of a control device provided in an embodiment of the present application. The present invention also provides a control device 200, including:

[0090] The processor 210 may be implemented by a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0091] The memory 220 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 220 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 220, and the processor 210 calls and executes the conference quality review method based on artificial intelligence in the embodiment of this application;

[0092] Input / output interface 230, used to implement information input and output;

[0093] Communication interface 240, used to realize communication interaction between the apparatus and other devices, which can be realized by wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.);

[0094] bus 250 , which transmits information between the various components of the device (e.g., processor 210 , memory 220 , input / output interface 230 , and communication interface 240 );

[0095] The processor 210 , the memory 220 , the input / output interface 230 , and the communication interface 240 are connected to each other in communication within the device via the bus 250 .

[0096] In addition, an embodiment of the present application further provides an electronic device, including the control device 200 of the above embodiment.

[0097] In addition, an embodiment of the present application further provides a storage medium, which is a computer-readable storage medium and stores a computer program. When the computer program is executed by a processor, the above-mentioned artificial intelligence-based conference quality review method is implemented.

[0098] As a non-transient computer-readable storage medium, the memory can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and are implemented to be located in one place, or may also be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0099] It will be appreciated by those skilled in the art that all or some of the steps and systems in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically include computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0100] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above-mentioned implementation mode. Technical personnel familiar with the field can also make various equivalent deformations or substitutions under the shared conditions without violating the spirit of the present invention. These equivalent deformations or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A conference quality review method based on artificial intelligence, characterized in that: Applied to a property system, the property system is in communication connection with a first number of terminals, different terminals correspond to different community identifiers, and the method includes: Receiving conference recordings sent by each of the terminals, associating the community identifier with the corresponding conference recording, and saving the conference recording carrying the community identifier to a recording pool, wherein the conference recordings sent from the same terminal share the same community identifier; Based on a preset time interval and a preset time period, a recording set is randomly selected from the recording pool in sequence, wherein the community identifiers corresponding to the conference recordings in the recording set selected each time have multiple types, and within the preset time period, the total number of types of the community identifiers selected cumulatively is the first number; At the preset time intervals, sequentially inputting the recording set into the AI ​​model for conference quality review to obtain a first score, sending the first score to a corresponding terminal, and receiving feedback information from the corresponding terminal; Determine an abnormal score based on all the first scores and the corresponding feedback information within the preset time period, and adjust the abnormal score based on the conference recording corresponding to the abnormal score and the feedback information to obtain a second score; retraining the AI ​​model based on the second score and the corresponding conference recording; The AI ​​model includes a text conversion module, a content analysis module and a plurality of target scoring modules. Different target scoring modules correspond to different target scoring item identifiers. The recording set is sequentially input into the AI ​​model for conference quality review to obtain a first score, including: Inputting the conference recording in the recording set into the text conversion module, and outputting a corresponding first text; Calling the content parsing module to segment the first text to obtain a plurality of second texts, and associating each of the target rating item identifiers with each of the second texts, wherein each of the target rating item identifiers corresponding to the same first text is different from each other; Calling each of the target scoring modules to calculate the sub-scores of each of the second texts, and summing all of the sub-scores to obtain the first score; The AI ​​model includes a plurality of candidate scoring modules, the target scoring module belongs to the candidate scoring module, and the method further includes: Displaying a first interface, wherein the first interface includes a plurality of candidate fields, and different candidate fields are used to indicate different candidate rating item identifiers; Acquire multiple target fields selected by the user from the multiple candidate fields, and determine the candidate scoring item identifier corresponding to the target field as the target scoring item identifier to adjust the conference quality review logic of the AI ​​model.

2. The conference quality review method based on artificial intelligence according to claim 1 is characterized in that: Within the preset time period, the recording set includes at least a first set, a second set and a third set, a first selection time of the first set is earlier than a second selection time of the second set, and the second selection time is earlier than a third selection time of the third set, and recording sets are randomly selected from the recording pool in sequence, including: At the first selection time, randomly selecting the first set from the recording pool, and recording the first community identifier corresponding to the first set, wherein the number of types of the first community identifiers is the second number; At the second selection time, randomly selecting the second set from the recording pool, and recording the second community identifier corresponding to the second set, wherein the number of types of the second community identifiers is the second number, and the second community identifier is different from the first community identifier; At the third selection time, the third set is selected from the recording pool, wherein the third set includes a third number of conference recordings, a third community identifier corresponding to the third set is different from the first community identifier and the second community identifier, and the third number is the number of types of community identifiers that have not been selected in the recording set.

3. The conference quality review method based on artificial intelligence according to claim 1 is characterized in that: The abnormality score includes a first abnormality score, a second abnormality score and a third abnormality score, and the abnormality score is determined based on all the first scores and the corresponding feedback information within the preset time period, including: Selecting a score whose value exceeds a high score threshold from all the first scores and determining it as the first abnormality score; Selecting a score whose value is lower than a low score threshold from all the first scores to determine as the second abnormal score; A first score whose value is between the high score threshold and the low score threshold is determined as a reference score, and when feedback information corresponding to the reference score indicates that the user does not agree with the reference score, the reference score is determined as the third abnormal score.

4. The conference quality review method based on artificial intelligence according to claim 1 is characterized in that: The first score includes a plurality of sub-scores, each of which corresponds to a different score item, the terminal displays a second interface, sends the first score to a corresponding terminal, and receives feedback information from the corresponding terminal, including: The property system divides the conference recording corresponding to the first score to obtain a plurality of sub-audios, wherein different sub-audios correspond to different sub-scores; The property system sends the first score and each of the sub-audios associated with each of the sub-scores to the terminal, and the terminal displays the first score and each of the sub-audios associated with each of the sub-scores in the second interface according to the first display template, and displays an evaluation window in the second interface, wherein the evaluation window includes a thumbs-up button, a thumbs-down button, and an evaluation input box; The terminal generates the feedback information in response to a user clicking the like button or the dislike button, and / or obtains the user's input content in the evaluation input box, and sends the feedback information to the property system.

5. The conference quality review method based on artificial intelligence according to claim 1 is characterized in that: There are multiple abnormal scores, each of which corresponds to a second score, and the abnormal score is adjusted based on the conference recording and the feedback information corresponding to the abnormal score to obtain the second score, including: Generate each abnormal record based on the conference recording and the feedback information corresponding to each abnormal score; Displaying each of the abnormal records in the first interface according to the second display template, each of the abnormal records being associated with a score input box; The scores input by the user in the score input boxes are determined as the second scores.

6. A control device, characterized in that: It includes at least one control processor and a memory for communicating with the at least one control processor; the memory stores instructions that can be executed by the at least one control processor, and the instructions are executed by the at least one control processor to enable the at least one control processor to execute the artificial intelligence-based conference quality review method as described in any one of claims 1 to 5.

7. An electronic device, characterized in that: Includes the control device as claimed in claim 6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the artificial intelligence-based conference quality review method as described in any one of claims 1 to 5.

Images