Recording quality inspection method and device, equipment, storage medium and program product

Abstract

The invention relates to a recording quality inspection method and device, equipment, a storage medium and a program product. The recording quality inspection method comprises the following steps: determining an abnormal level of a recording file, and seat recording features and user recording features in the recording file; determining an abnormal quantized value of the sound recording file according to the abnormal level of the sound recording file, the seat sound recording feature and the user sound recording feature; and determining a quality inspection result of the sound recording file based on the abnormal quantized value of the sound recording file. By adopting the method, the timeliness of recording quality inspection is better, and the accuracy is higher.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a recording quality inspection method, device, equipment, storage medium and program product. Background Art

[0002] With the development of the service industry, the requirements for customer service agents are also constantly increasing. For example, some call center agents may fail to meet the requirements in voice communication with customers due to inadequate training or improper communication. Therefore, it is necessary to conduct quality inspection on the recordings of agents and customers, appease customers and reward or punish agents based on the inspection results. Summary of the invention

[0003] Based on this, it is necessary to provide a recording quality inspection method, device, equipment, storage medium and program product with good timeliness and high accuracy to address the above technical problems.

[0004] In a first aspect, the present application provides a recording quality inspection method, comprising:

[0005] Determine the abnormality level of the recording file, as well as the agent recording features and user recording features in the recording file;

[0006] Determine the abnormality quantification value of the recording file based on the abnormality level of the recording file, the agent recording characteristics, and the user recording characteristics;

[0007] Based on the abnormal quantitative value of the recording file, the quality inspection result of the recording file is determined.

[0008] In a second aspect, the present application also provides a recording quality inspection device, including:

[0009] A determination module is used to determine the abnormal level of the recording file, as well as the agent recording features and user recording features in the recording file;

[0010] A processing module, used to determine an abnormality quantification value of the recording file according to the abnormality level of the recording file, the agent recording characteristics and the user recording characteristics;

[0011] The quality inspection module is used to determine the quality inspection results of the recording file based on the abnormal quantitative value of the recording file.

[0012] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0013] Determine the abnormality level of the recording file, as well as the agent recording features and user recording features in the recording file;

[0014] Determine the abnormality quantification value of the recording file based on the abnormality level of the recording file, the agent recording characteristics, and the user recording characteristics;

[0015] Based on the abnormal quantification value of the recording file, the quality inspection result of the recording file is determined. In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which implements the following steps when the computer program is executed by a processor:

[0016] Determine the abnormality level of the recording file, as well as the agent recording features and user recording features in the recording file;

[0017] Determine the abnormality quantification value of the recording file based on the abnormality level of the recording file, the agent recording characteristics, and the user recording characteristics;

[0018] Based on the abnormal quantification value of the recording file, the quality inspection result of the recording file is determined. In a fifth aspect, the present application also provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0019] Determine the abnormality level of the recording file, as well as the agent recording features and user recording features in the recording file;

[0020] Determine the abnormality quantification value of the recording file based on the abnormality level of the recording file, the agent recording characteristics, and the user recording characteristics;

[0021] Based on the abnormal quantitative value of the recording file, the quality inspection result of the recording file is determined.

[0022] The above-mentioned recording quality inspection method, device, computer equipment, storage medium and computer program product can complete the comprehensive quality inspection of all recordings that need to be inspected on the day at one time during the whole process, without missing any important recordings, and can realize the rapid processing of large-scale recordings, saving a lot of manpower and time resources. Furthermore, in combination with the abnormal level of each recording file, the agent recording characteristics and the user recording characteristics, the abnormal situation of each recording file can be quantified, and then based on the quantified abnormal quantification value of each recording, the quality inspection results of multiple recording files can be obtained, so as to intuitively and quickly inspect high-risk recordings with high abnormality from multiple recordings, so that high-risk recordings in all recording files that need to be inspected on the day can be inspected in time, and high-risk recordings will not be missed due to delays or missed inspections, thereby making the timeliness of recording quality inspection better and the quality inspection results more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A diagram showing an application environment of a recording quality inspection method in an embodiment;

[0025] Figure 2 Schematic diagram of a flow chart of a recording quality inspection method in one embodiment;

[0026] Figure 3 It is a schematic diagram of a recording quality inspection process using recording quality inspection rules in a specific application example;

[0027] Figure 4 It is a flowchart of a recording quality inspection method in another embodiment;

[0028] Figure 5 A schematic diagram of allocating seats to callers in a specific application example;

[0029] Figure 6 It is a flowchart of a recording quality inspection method in a specific application example;

[0030] Figure 7 It is a structural block diagram of a recording quality inspection device in one embodiment;

[0031] Figure 8 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0032] 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.

[0033] At present, taking call centers as an example, some call center agents may encounter communication problems with users due to inadequate training or improper communication. A call center is a centralized place composed of a group of people who make or answer customer calls to provide corresponding services; agents are the general term for people who make or answer customer calls; users are the opposite of agents, who communicate with agents.

[0034] In order to reduce the occurrence of this situation, it is necessary to detect the voice communication process as early as possible. Specifically, the call center can store the voice of the communication as a recording file after each voice communication, and push the recording file of the day from the call center offline to the quality inspection system in the early morning of the next day. The quality inspection personnel of the quality inspection system will conduct quality inspection on the recording file of the day, obtain the non-compliant recordings in all the recording files of the day, and appease the customers corresponding to the non-compliant recordings and deduct the seats corresponding to the non-compliant recordings according to the non-compliant situation. Among them, the non-compliant situation can be the violation level, that is, take appeasement measures corresponding to the violation level according to the violation level.

[0035] However, the above-mentioned recording quality inspection method will bring many technical problems. The details are as follows:

[0036] 1. Time-consuming and labor-intensive: Manual quality inspection requires a lot of time and human resources, and is inefficient, especially for large-scale recording data.

[0037] 2. Strong subjectivity: The quality inspection results are affected by individual subjective opinions and emotions, and are not objective and unified enough.

[0038] 3. Limited coverage: Due to human resource and time constraints, it is not possible to conduct a comprehensive quality check on all recordings, which may lead to omissions.

[0039] 4. Difficult to standardize: Since different quality inspectors may have different standards and styles, it is difficult to establish unified quality inspection standards.

[0040] 5. Slow detection of abnormal data: If certain non-compliant recordings are quality-checked in sequence, they may be delayed in sequence during data peak periods or when quality-check capacity is insufficient.

[0041] However, the time-consuming and laborious process, strong subjectivity, limited coverage, difficulty in standardization, and slow detection of abnormal data result in low timeliness and accuracy in quality inspection of recordings. To solve the above problems, the present application provides a recording quality inspection method that can improve the timeliness of recording quality inspection and can accurately inspect recordings.

[0042] During the entire process, all recordings that need to be inspected on the day can be inspected at one time without missing any important recordings, which can achieve rapid processing of large-scale recordings and save a lot of manpower and time resources. Furthermore, by combining the abnormality level of each recording file, the corresponding seat recording characteristics and the user's recording characteristics, the abnormality of each recording file can be quantified, and then based on the quantified abnormality quantification value of each recording, the quality inspection results of multiple recording files can be obtained, so that high-risk recordings with high abnormality levels can be inspected intuitively and quickly from multiple recordings, so that high-risk recordings in all recording files that need to be inspected on the day can be inspected in time, and high-risk recordings will not be missed due to delays or missed inspections, thereby making the recording quality inspection more timely and accurate.

[0043] The recording quality inspection method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the agent terminal 102 held by the agent communicates with the agent server 104 through the network, the agent server 104 communicates with the recording quality inspection server 106, and the recording quality inspection server 106 communicates with the quality inspection terminal 108 held by the recording quality inspector. The data storage system can store the data that the recording quality inspection server 106 needs to process. The data storage system can be integrated on the recording quality inspection server 106, or it can be placed on the cloud or other network servers.

[0044] The agent uses the agent terminal 102 to talk with the user, and after the call is over, the agent obtains a recording of the communication with the user. Furthermore, the agent server 104 stores multiple recording files obtained by the agent terminal 102 in the current time period in the cloud storage system to ensure the data security and reliability of the recording. The agent server 104 obtains the recording addresses corresponding to the stored multiple recording files and uploads the multiple recording addresses to the recording quality inspection server 106.

[0045] When the quality inspector needs to conduct quality inspection on the recording, he / she initiates a recording quality inspection request to the recording quality inspection server 106 through the quality inspection terminal 106. The recording quality inspection request instructs the recording quality inspection server 106 to obtain the recording files corresponding to the multiple recording addresses from the cloud storage system in real time based on the multiple recording addresses uploaded by the agent server 104, and to convert the recording files into text to obtain the recording text corresponding to the recording files. The recording quality inspection server 106 conducts quality inspection on the recording text, and determines the abnormality level of the recording file, as well as the agent recording features and user recording features in the recording file based on the abnormality level of the recording text; determines the abnormality quantification value of the recording file based on the abnormality level of the recording file, the agent recording features and the user recording features; and determines the quality inspection result of the recording file based on the abnormality quantification value of the recording file. Furthermore, the quality inspection result of the recording file can be pushed to the quality inspection terminal 106 held by the quality inspector.

[0046] Among them, the seat terminal 102 and the quality inspection terminal 106 can be, but are not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The seat server 104 and the recording quality inspection server 106 can be implemented using independent servers or server clusters consisting of multiple servers.

[0047] In other embodiments, the recording quality inspection method provided in the embodiments of the present application can also be applied separately to the seat terminal and the quality inspection terminal. The seat has a conversation with the user through the seat terminal held by the seat terminal, and obtains the recording file with the user after the call. The multiple recording files obtained in the current time period are stored in the cloud storage system, and the recording addresses corresponding to the stored multiple recording files are obtained, and the multiple recording addresses are sent to the quality inspection terminal; the quality inspection terminal obtains the multiple recording addresses sent by the seat terminal, obtains the recording files corresponding to the multiple recording addresses from the cloud storage system, and converts the recording files into text to obtain the recording text corresponding to the recording files. The quality inspection terminal uses the above-mentioned recording quality inspection method to perform recording quality inspection on the recording text to determine the quality inspection result of the recording file corresponding to the recording text. Its specific data processing process is similar to the above and will not be repeated here.

[0048] In other embodiments, the recording quality inspection method provided in the embodiments of the present application can also be applied separately to the seat terminal and the recording quality inspection server. The seat has a conversation with the user through the seat terminal held by the seat terminal, and obtains the recording file with the user after the call. And the multiple recording files obtained in the current time period are stored in the cloud storage system, the recording addresses corresponding to the stored multiple recording files are obtained, and the multiple recording addresses are sent to the recording quality inspection server; the recording quality inspection server obtains the multiple recording addresses sent by the seat terminal, obtains the recording files corresponding to the multiple recording addresses from the cloud storage system, and converts the recording files into text to obtain the recording text corresponding to the recording files. The recording quality inspection server adopts the above-mentioned recording quality inspection method to perform recording quality inspection on the recording text to determine the quality inspection result of the recording file corresponding to the recording text. Its specific data processing process is similar to the above and will not be repeated here.

[0049] In an exemplary embodiment, Figure 2 As shown, a recording quality inspection method is provided, which is applied to Figure 1 The recording quality inspection server 104 in the example is used for explanation.

[0050] S200, determining the abnormality level of the recording file, and the agent recording features and user recording features in the recording file.

[0051] Among them, recording is the process of recording sound signals on a medium. The recording file is a recording file generated after the agent and the user have a conversation based on the conversation process. In other words, the agent and the user are the two parties of the recording corresponding to the recording file, and the user can be the outgoing call object or the incoming call object, that is, the recording can be a recording generated when the agent calls the user, or it can be a recording generated when the user calls the agent. The abnormality level of the recording file is the abnormality level of the recording corresponding to the recording file, which can be used to characterize the degree of abnormality of the recording. The agent recording feature is the feature corresponding to several recordings associated with the agent, for example, the agent recording feature includes the abnormality level of the agent's historical recordings within a preset historical time period; the user recording feature is the feature of several recordings associated with the user, for example, the user recording feature includes the abnormality level of the user's historical recordings within a preset historical time period.

[0052] Specifically, the agent makes or receives a user's call through the agent terminal he holds. After the call ends, the agent terminal can automatically obtain the recording of the communication with the user. Furthermore, the agent server can store multiple recording files generated by the agent terminal in the current time period in the cloud storage system to ensure the data security and reliability of the recording. The agent server obtains the recording addresses corresponding to the stored multiple recording files and uploads the multiple recording addresses to the recording quality inspection server.

[0053] When the recording quality inspection is required, the recording quality inspection personnel initiate a recording quality inspection request to the recording quality inspection server through the quality inspection terminal. The recording quality inspection request instructs the recording quality inspection server to obtain the recording files corresponding to multiple recording addresses from the cloud storage system in real time based on the multiple recording addresses uploaded by the seat server. The recording quality inspection server converts the recording file into text and obtains the recording text corresponding to the recording file. The recording quality inspection server performs quality inspection on the recording text based on the abnormal level of the recording text, as well as the seat recording features and user recording features in the recording file.

[0054] In one embodiment, the agent server stores multiple recording files generated by the agent terminal in the current time period in the cloud storage system, including: the agent server performs format unification processing on the multiple recording files generated by the agent terminal in the current time period, and uploads the multiple recording files after the format unification processing to the cloud storage system. For example, if the unified storage format is A, when the initial format of a recording file is B, the recording file needs to be format converted, and the recording file is converted from the initial format B to the unified storage format A.

[0055] In one embodiment, the seat server can send the recording address to the recording quality inspection server through the message bus. In addition, the object identifiers of both parties of the recording file, that is, the seat identifier and the user identifier, can also be sent to the recording quality inspection server together with the recording address through the recording bus. Among them, the message bus has the ability to transfer and interact with data, so that the parties do not communicate directly, but use the bus as a standard communication interface to transfer and interact data. And in this application, the message bus transmits the recording address rather than the recording file, which can achieve the decoupling of recording acquisition and recording quality inspection without affecting each other, that is, the recording acquisition process and the recording quality inspection process are divided into two parts that do not affect each other through the message bus.

[0056] The recording quality inspection server can determine several historical recordings within a preset historical time period associated with the seat identification and several recordings within a preset historical time period associated with the user identification based on the seat identification and the user identification, and determine the seat recording features in the recording file based on the several historical recordings within the preset historical time period associated with the seat identification; and determine the user recording features in the recording file based on the several recordings within the preset historical time period associated with the user identification. In other words, the recording quality inspection server can not only perform quality inspection on the current recording file to determine the abnormality level of the recording file; it can also perform recording quality inspection on several historical recordings associated with the seat identification to determine the abnormality level of the seat historical recordings within the preset historical time period; it can also perform recording quality inspection on several historical recordings associated with the user identification to determine the abnormality level of the user historical recordings within the preset historical time period.

[0057] In one embodiment, the method of converting the recording file into text includes but is not limited to: converting the recording file into recording text using ASR (Automatic Speech Recognition, speech recognition technology).

[0058] S400, determining an abnormality quantification value of the recording file according to the abnormality level of the recording file, the agent recording feature and the user recording feature.

[0059] Among them, quantification refers to a way of expressing evaluation in quantitative form. The abnormality quantification value refers to an evaluation value of the abnormality degree of the recording file. The abnormality quantification value and the abnormality level need to be distinguished. The abnormality level is the abnormality level of the recording file itself, while the abnormality quantification value is the quantitative value of the abnormality degree determined from the three dimensions of the abnormality level of the recording file, the agent recording characteristics, and the user recording characteristics. In other words, the abnormality level is the initial detection result of the call recording file itself, and the abnormality quantification value is the final detection result obtained after a deeper evaluation of the call recording file.

[0060] Specifically, the abnormality quantification value of the recording file is determined based on the abnormality level of the recording file, the characteristics of the seat recording, and the characteristics of the user recording. In other words, the abnormality quantification value of the recording file is determined based on the abnormality level of the recording file, the abnormality level of the seat historical recording within a preset historical time period, and the abnormality level of the user historical recording within a preset historical time period, so that the recording quality inspection process not only focuses on the abnormality level of the recording file itself, but also combines the abnormalities of the historical recordings of both parties of the recording file, that is, the seat and the user, to conduct a deeper quality inspection of the recording file.

[0061] S600, determining a quality inspection result of the recording file based on the abnormal quantification value of the recording file.

[0062] Among them, the quality inspection result of the recording file refers to whether the recording file is a high-risk recording.

[0063] Specifically, the recording files are sorted according to the abnormal quantization values ​​of the recording files to obtain an abnormal sequence of the recording files. The sorting method includes but is not limited to: sorting the recording files according to the abnormal quantization values ​​from high to low, or sorting the recording files according to the abnormal quantization values ​​from low to high.

[0064] Based on the abnormal sequence of recording files, high-risk recordings in the abnormal sequence of recording files are determined. On the basis that the abnormal sequence of recording files is to sort the recording files from high to low according to the abnormal quantification value, the high-risk recordings in the abnormal sequence of recording files can be the N recording files that are ranked first in the abnormal sequence of recording files; on the basis that the abnormal sequence of recording files is to sort the recording files from low to high according to the abnormal quantification value, the high-risk recordings in the abnormal sequence of recording files can be the N recording files that are ranked last in the abnormal sequence of recording files, and N is a positive integer. In actual applications, N can be set by the user. At this time, the purpose of sorting the recording files is to improve the accuracy of determining high-risk recordings from the abnormal sequence of recording files.

[0065] In addition, on the basis that the recording file abnormal sequence is to sort the recording files from high to low according to the abnormal quantization value, the high-risk recordings in the recording file abnormal sequence can also be K recording files in the recording file abnormal sequence whose abnormal quantization values ​​exceed the preset threshold, where K is a positive integer; on the basis that the recording file abnormal sequence is to sort the recording files from low to high according to the abnormal quantization value, the high-risk recordings in the recording file abnormal sequence are K recording files in the recording file abnormal sequence whose abnormal quantization values ​​exceed the preset threshold. At this time, the purpose of sorting the recording files is to improve the efficiency of determining the K recording files in the recording file abnormal sequence whose abnormal quantization values ​​exceed the preset threshold.

[0066] In the above recording quality inspection method, during the whole process, all recordings that need to be inspected on the day can be inspected at one time without missing any important recordings, and large-scale recordings can be quickly processed, saving a lot of manpower and time resources. Furthermore, by combining the abnormality level of each recording file, the agent recording characteristics and the user recording characteristics, the abnormality of each recording file can be quantified, and then based on the quantified abnormality quantification value of each recording, the quality inspection results of multiple recording files can be obtained, so as to intuitively and quickly inspect high-risk recordings with high abnormality from multiple recordings, so that high-risk recordings in all recording files that need to be inspected on the day can be inspected in time, and high-risk recordings will not be missed due to delays or missed inspections, thereby making the timeliness of recording quality inspection better.

[0067] In one embodiment, determining the abnormality level of the recording file includes:

[0068] Convert the recording file into recording text; perform semantic analysis on the recording text to obtain the semantic information results of the recording file; perform quality inspection and analysis on the recording text and semantic information results of the recording file to obtain the abnormality level of the recording file.

[0069] The semantic information result is the semantic information in the recording text, including but not limited to at least one of the sentence meaning, keywords, emotions, speaking speed, interruption, volume, etc. of the recording text. The sentence meaning refers to the meaning of the sentence in the recording text, such as threats and intimidation; keywords are specific words in the recording text; emotions include but are not limited to negative or positive; speaking speed includes but is not limited to fast, medium, slow, etc.; interruption refers to whether the seat interrupts in the recording text, including but not limited to interruption, no interruption, etc.; volume refers to whether the seat volume in the recording text exceeds the preset volume threshold, including but not limited to the volume exceeding the threshold, the volume does not exceed the threshold, etc.

[0070] Specifically, the recording file is converted into recording text, and the recording text is semantically parsed to obtain the semantic information results of the recording file. The semantic parsing of the recording text includes but is not limited to: performing NLP (Natural Language Processing) semantic parsing on the recording text to obtain at least one of the semantic information results such as sentence meaning, keywords, emotions, speaking speed, interruption, volume, etc. of the recording text. Among them, NLP is a branch discipline in the field of artificial intelligence and linguistics, that is, the recording text is processed by combining artificial intelligence and linguistics, which can generate results related to semantic information and generalize the text of some scenes.

[0071] In another embodiment, semantic analysis processing of the recorded text includes but is not limited to: performing NLP semantic analysis processing on each sentence in the recorded text to obtain at least one of the semantic information results of the text, such as sentence meaning, keywords, emotions, speaking speed, interruptions, volume, etc., and summarizing the semantic information results of the recorded text based on the semantic information results of each sentence in the recorded text.

[0072] The recording text and semantic information of the recording file are quality checked and analyzed to obtain the abnormality level of the recording file. In other words, the recording text is quality checked and analyzed from multiple dimensions such as sentence meaning, keywords, emotions, speaking speed, interruptions, volume, etc. to obtain the abnormality level of the recording text, and the abnormality level of the recording text is used as the abnormality level of the recording file.

[0073] In one embodiment, after obtaining the abnormality level of the recording file, the abnormality level of the recording file can also be manually reviewed. When the result determined by the manual review is the same as the abnormality level of the recording file, it is considered that the abnormality level of the recording file is accurate; when the result determined by the manual review is different from the abnormality level of the recording file, it means that the recording file needs to be re-checked.

[0074] In this embodiment, by performing semantic analysis on the recorded text corresponding to the recorded file, semantic information results of multiple dimensions of the recorded text can be obtained, and based on the semantic information results of multiple dimensions corresponding to the recorded text, the recorded text can be accurately and comprehensively quality inspected to obtain the accurate abnormality level of the recorded file.

[0075] In one embodiment, the recording text and semantic information of the recording file are quality checked and analyzed to obtain the abnormality level of the recording file, including:

[0076] Adopt M recording quality inspection rules to detect the recording text and semantic information results of the recording file, and obtain the detection result corresponding to the recording file under each recording quality inspection rule; According to the detection result corresponding to the recording file under the M recording quality inspection rules, determine the abnormal level of the recording file, where M is a positive integer greater than 1.

[0077] Among them, the recording quality inspection rules are necessary elements for quality inspection of recordings. The recording quality inspection rules include at least one of sentence meaning quality inspection rules, keyword quality inspection rules, emotion quality inspection rules, speaking speed quality inspection rules, interruption quality inspection rules, and volume quality inspection rules.

[0078] Specifically, M kinds of recording quality inspection rules are adopted to detect the recording text and semantic information results of the recording file respectively. Each recording quality inspection rule can obtain a detection result, and the detection results corresponding to the recording file under the M kinds of recording quality inspection rules are obtained. In other words, the detection results corresponding to the recording file under the M kinds of recording quality inspection rules include: the detection results corresponding to the sentence meaning quality inspection rules, the detection results corresponding to the keyword quality inspection rules, the detection results corresponding to the emotion quality inspection rules, the detection results corresponding to the speech speed quality inspection rules, the detection results corresponding to the interruption quality inspection rules, and at least one of the detection results corresponding to the volume quality inspection rules. For example, when the keyword quality inspection rules are adopted to detect the recording text and semantic information results of the recording file, it can be that the keywords in the recording text must meet the preset keywords, or the keywords in the recording text must be sorted according to the preset keyword sequence.

[0079] The above-mentioned multiple detection results are summarized and analyzed, and based on the multiple detection results, a final detection result of the recording file is determined, that is, the abnormality level of the recording file is determined.

[0080] In an exemplary embodiment, sentence meaning quality inspection rules are used to detect the recorded text and semantic information results of the recorded file, and the detection results corresponding to the recorded file under each recorded quality inspection rule are obtained, including: using sentence meaning quality inspection rules to detect sentence meanings on the recorded text and semantic information results of the recorded file, and when it is detected that the recorded text or semantic information results of the recorded file represent the sentence meaning as threat or intimidation, the detection result of the recorded file under the sentence meaning quality inspection rule is determined to be abnormal; when it is not detected that the recorded text or semantic information results of the recorded file represent the sentence meaning as threat or intimidation, the detection result of the recorded file under the sentence meaning quality inspection rule is determined to be normal. Furthermore, when it is detected that the recorded text or semantic information results of the recorded file represent the sentence meaning as threat or intimidation, the initial abnormal level of the recorded file under the sentence meaning quality inspection rule can also be determined based on the degree of threat or intimidation.

[0081] In an exemplary embodiment, keyword quality inspection rules are used to detect the audio text and semantic information results of the audio file, and the detection results corresponding to the audio file under each audio quality inspection rule are obtained, including: there are preset keywords or keyword sequences in the keyword quality inspection rules, when keyword detection is performed on the audio text and semantic information results of the audio file using the keyword quality inspection rules, when the preset keywords or keyword sequences are detected in the audio text or semantic information results of the audio file, the detection result of the audio file under the keyword quality inspection rules is determined to be abnormal; when the preset keywords or keyword sequences are not detected in the audio text or semantic information results of the audio file, the detection result of the audio file under the keyword quality inspection rules is determined to be normal. Further, when the preset keywords or keyword sequences are detected in the audio text or semantic information results of the audio file, the initial abnormality level of the audio file under the keyword quality inspection rules can also be determined based on the number of keywords or keyword sequences detected.

[0082] In an exemplary embodiment, emotion quality inspection rules are used to detect the recorded text and semantic information results of the recorded file, and the detection results corresponding to the recorded file under each recorded quality inspection rule are obtained, including: using emotion quality inspection rules to detect emotions on the recorded text and semantic information results of the recorded file, when it is detected that the recorded text or semantic information results of the recorded file represent negative emotions, the detection result of the recorded file under the emotion quality inspection rule is determined to be abnormal; when it is detected that the recorded text or semantic information results of the recorded file represent positive emotions, the detection result of the recorded file under the emotion quality inspection rule is determined to be normal. Furthermore, when it is detected that the recorded text or semantic information results of the recorded file represent negative emotions, the initial abnormal level of the recorded file under the sentence meaning quality inspection rule can also be determined based on the degree of negativity.

[0083] In an exemplary embodiment, a speech speed quality inspection rule is used to detect the recording text and semantic information results of the recording file, and the detection result corresponding to the recording file under each recording quality inspection rule is obtained, including: using the speech speed quality inspection rule to detect the speech speed of the recording text and semantic information results of the recording file, when it is detected that the speech speed represented in the recording text or semantic information results of the recording file is not within the normal speech speed range, the detection result of the recording file under the speech speed quality inspection rule is determined to be abnormal; when it is detected that the speech speed represented in the recording text or semantic information results of the recording file is within the normal speech speed range, the detection result of the recording file under the speech speed quality inspection rule is determined to be normal. Furthermore, when it is detected that the speech speed represented in the recording text or semantic information results of the recording file is not within the normal speech speed range, it is also possible to determine the initial abnormal level corresponding to the speech speed range based on the speech speed range in which the speech speed is located, that is, to determine the initial abnormal level of the recording file under the speech speed quality inspection rule.

[0084] In an exemplary embodiment, the interruption quality inspection rule is adopted to detect the recording text and semantic information results of the recording file, and the detection result corresponding to the recording file under each recording quality inspection rule is obtained, including: the interruption quality inspection rule is adopted to detect the interruption of the recording text and semantic information results of the recording file, and when it is detected that the number of interruptions represented in the recording text or semantic information results of the recording file is not within the normal number range, the detection result of the recording file under the interruption quality inspection rule is determined to be abnormal; when it is detected that the number of interruptions represented in the recording text or semantic information results of the recording file is within the normal number range, the detection result of the recording file under the interruption quality inspection rule is determined to be normal. Further, when it is detected that the number of interruptions represented in the recording text or semantic information results of the recording file is not within the normal number range, it is also possible to determine the initial abnormal level corresponding to the number based on the number of interruptions, that is, to determine the initial abnormal level of the recording file under the interruption quality inspection rule.

[0085] In an exemplary embodiment, a volume quality inspection rule is used to detect the recording text and semantic information results of a recording file, and the detection result corresponding to the recording file under each recording quality inspection rule is obtained, including: using the volume quality inspection rule to detect the volume of the recording text and semantic information results of the recording file, when it is detected that the recording text or semantic information results of the recording file represent that the volume is not within the normal volume range, the detection result of the recording file under the volume quality inspection rule is determined to be abnormal; when it is detected that the recording text or semantic information results of the recording file represent that the volume is within the normal volume range, the detection result of the recording file under the volume quality inspection rule is determined to be normal. Furthermore, when it is detected that the recording text or semantic information results of the recording file represent that the volume is not within the normal volume range, the abnormal initial level corresponding to the volume can also be determined based on the volume, that is, the abnormal initial level of the recording file under the volume quality inspection rule is determined.

[0086] In the above embodiment, by configuring the recording quality inspection rules, the recording text and semantic information results of the recording file can be efficiently detected, thereby improving the efficiency of quality inspection of the recording file and obtaining the abnormal level of the recording file. Furthermore, by configuring multiple recording quality inspection rules, the recording text and semantic information results of the recording file can be accurately detected from multiple dimensions, thereby determining the accurate abnormal level of the recording file.

[0087] In one embodiment, the detection result includes an abnormality mark and an initial abnormality level; according to the detection result corresponding to the recording file under each recording quality inspection rule, the abnormality level of the recording file is determined, including:

[0088] If the abnormal identification corresponding to the recording file under the M recording quality inspection rules meets the preset conditions, the highest abnormal initial level among the abnormal initial levels is obtained; and the highest abnormal initial level is determined as the abnormal level of the recording file.

[0089] The detection result including an abnormality mark means that the detection result represents an abnormality, that is, when the detection result represents an abnormality, the detection result includes an abnormality mark; when the detection result represents a normal state, the detection result does not include an abnormality mark. The preset condition refers to the condition that the abnormality mark in the detection result corresponding to the recording rule must meet.

[0090] Specifically, M recording quality inspection rules are adopted to detect the recording text and semantic information results of the recording file respectively. In the process of obtaining the detection results corresponding to the recording file under the M recording quality inspection rules, the detection results corresponding to the recording file under the M recording quality inspection rules may include the following three situations, that is, the abnormal identification may include the following three situations:

[0091] A. The corresponding detection results under any recording quality inspection rule contain abnormal identification.

[0092] The corresponding detection results under B and M recording quality inspection rules do not contain abnormal identification.

[0093] C. The corresponding detection results under all recording quality inspection rules contain abnormal identification.

[0094] It is only necessary to determine the abnormality level of the recording file when the abnormal identification corresponding to the recording file under the M recording quality inspection rules meets one of the above preset conditions. Obtain the corresponding detection results under the M recording quality inspection rules, and analyze the detection results to obtain the abnormal detection results containing the abnormal identification. If the corresponding detection results under the M recording quality inspection rules do not contain the abnormal identification, there is no need to obtain its abnormal initial level. Based on the abnormal initial level in the abnormal detection result, obtain the highest abnormal initial level in the abnormal initial level, and determine the highest abnormal initial level as the abnormal level of the recording file.

[0095] For example, the initial abnormality levels include, from low to high, level 1, level 2, and level 3. Using recording quality inspection rule A, recording quality inspection rule B, and recording quality inspection rule C, the recording file is quality inspected to obtain the detection result of the recording file under recording quality inspection rule A, the detection result of the recording file under recording quality inspection rule B, and the detection result of the recording file under recording quality inspection rule C. If the detection result of the recording file under recording quality inspection rule A, the detection result of the recording file under recording quality inspection rule B, and the detection result of the recording file under recording quality inspection rule C all contain abnormal identification, and the initial abnormality level in the detection result of the recording file under recording quality inspection rule A is level 1, the detection result of the recording file under recording quality inspection rule B is level 2, and the detection result of the recording file under recording quality inspection rule C is level 3, then the highest initial abnormality level, that is, level 3, is determined as the abnormality level of the recording file.

[0096] In this embodiment, when the abnormal identification corresponding to the recording file under M types of recording quality inspection rules meets the preset conditions, the highest abnormal initial level among the abnormal initial levels is obtained, and the highest abnormal initial level is determined as the abnormal level of the recording file. The abnormal level of the recording file can be efficiently determined based on the detection results corresponding to the M types of recording quality inspection rules.

[0097] In one embodiment, when the abnormal identification corresponding to the recording file under each recording quality inspection rule meets the preset conditions, the most abnormal initial level in the abnormal initial levels can also be determined as the abnormal value of the recording file; the abnormal initial levels in the abnormal initial levels can also be converted into a score, and based on the average value of the score, the score is converted into a corresponding abnormal level, and the abnormal level is determined as the abnormal level of the recording file. In other words, the specific means of determining the abnormal level of the recording file according to the detection results corresponding to the recording file under each recording quality inspection rule can be set by the user.

[0098] In one embodiment, the preset condition includes any one of the following: the detection results of the recording file under M recording quality inspection rules all include abnormal identification; the detection results of the recording file under M recording quality inspection rules include at least one abnormal identification.

[0099] Specifically, when the preset condition is that the detection results of the recording file under M types of recording quality inspection rules all include abnormal identification, it means that the detection results corresponding to the M types of recording quality inspection rules are all abnormal. At this time, the recording file is abnormal and has an abnormal level; when the preset condition is that the detection results of the recording file under M types of recording quality inspection rules include at least one abnormal identification, it means that only one recording quality inspection rule needs to have a detection result corresponding to an abnormality. At this time, the recording file is abnormal and can have an abnormal level.

[0100] Therefore, when the abnormal identification corresponding to the recording file under the M recording quality inspection rules meets the preset conditions, the highest abnormal initial level among the abnormal initial levels can be determined as the abnormal level of the recording file.

[0101] When the abnormal identification corresponding to the recording file under M types of recording quality inspection rules does not meet the preset conditions, for example, the detection results corresponding to the recording file under M types of recording quality inspection rules do not include abnormal identification, that is, they are all normal, the recording file is also normal, and there is no need to determine the abnormal level.

[0102] In this embodiment, by setting preset conditions, the recording quality inspection server can determine the highest initial abnormal level as the abnormal level of the recording file only when the abnormal identification corresponding to the recording file under M types of recording quality inspection rules meets the preset conditions. This avoids the situation where the abnormal level of the recording file is determined when the recording file does not contain abnormal identification under M types of recording quality inspection rules, avoids time loss caused by invalid operations, and improves the efficiency of recording quality inspection.

[0103] In one embodiment, the detection result includes an abnormal initial level, and M recording quality inspection rules are used to detect the recording text and semantic information results of the recording file, and the detection result corresponding to the recording file under each recording quality inspection rule is obtained, including:

[0104] For any recording quality inspection rule, the target threshold interval of the recording text and semantic information results of the recording file is detected according to the threshold interval preset in the recording quality inspection rule; the initial abnormal level corresponding to the target threshold interval is determined as the initial abnormal level of the recording file.

[0105] Among them, each recording quality inspection rule also presets multiple threshold intervals, and each threshold interval corresponds to an initial abnormality level. For example, for the keyword quality inspection rule, three threshold intervals are preset, including a first threshold interval, a second threshold interval, and a third threshold interval; if there are 1-3 preset keywords in the recording text, it is in the first threshold interval, and the initial abnormality level corresponding to the first threshold interval is level 1; if there are 4-6 preset keywords in the recording text, it is in the second threshold interval, and the initial abnormality level corresponding to the second threshold interval is level 2; if there are 7-9 preset keywords in the recording text, it is in the third threshold interval, and the initial abnormality level corresponding to the third threshold interval is level 3.

[0106] Specifically, for any recording quality inspection rule, the target threshold interval of the recording text and semantic information results of the recording file is determined, and the initial abnormality level corresponding to the target threshold interval is determined. For example, the recording text and semantic information results of the recording file are detected to determine how many preset keywords exist in the recording text. When there are 8 preset keywords in the recording text, the target threshold interval is 7-9 preset keywords. At this time, the initial abnormality level corresponding to the target threshold interval is level 3, and level 3 is used as the initial abnormality level corresponding to the keyword quality inspection rule.

[0107] In an actual application, for the speech speed quality inspection rule, four threshold intervals are preset. If the speech speed is 10-40 words / min, it is in the first threshold interval, and the initial abnormality level corresponding to the first threshold interval is level 1; if the speech speed is 41-60 words / min, it is in the second threshold interval, and the initial abnormality level corresponding to the second threshold interval is level 2; if the speech speed is 200-299 words / min, it is in the third threshold interval, and the initial abnormality level corresponding to the third threshold interval is level 3; if the speech speed is 300-400 words / min, it is in the fourth threshold interval, and the initial abnormality level corresponding to the fourth threshold interval is level 4.

[0108] The recording text and semantic information results of the recording file are detected to determine the actual speaking speed in the semantic information results. When the actual speaking speed is 55 words / min, the target threshold interval is the second threshold interval. At this time, the initial abnormal level corresponding to the second threshold interval is level 2, and level 2 is used as the initial abnormal level corresponding to the speaking speed quality inspection rule.

[0109] In this embodiment, by determining the target threshold range of the recording text and semantic information results of the recording file, the initial abnormality level corresponding to the recording quality inspection rule can be accurately divided to accurately determine the abnormality degree of the recording file itself under the recording quality inspection rule.

[0110] In a specific application example, Figure 3The figure shows a configuration diagram of recording quality inspection rules and preset conditions in the process of quality inspection of recordings using M recording quality inspection rules. Among them, the M recording quality inspection rules can be configured as at least one of sentence meaning quality inspection rules, keyword quality inspection rules, emotion quality inspection rules, speech speed quality inspection rules, interruption quality inspection rules, and volume quality inspection rules. When using M recording quality inspection rules to perform quality inspection on recordings, the preset conditions can be configured to determine the abnormal conditions of the recording file. For example, when the corresponding detection result under at least one recording quality inspection rule contains an abnormal mark, the recording file is abnormal; or, when the corresponding detection results under all recording quality inspection rules contain abnormal marks, the recording file is abnormal. Further, the detection result corresponding to each recording quality inspection rule is determined by whether the threshold interval of the recording text and semantic information results of the recording file reaches the preset threshold interval when detecting each recording quality inspection rule. For example, if the abnormal threshold interval is reached, it can be determined that the detection result is abnormal. Then, the abnormal threshold interval is subdivided, and the initial abnormal level corresponding to the detection result can be determined. Furthermore, the quality inspection of the recording can be performed on every sentence in the recording, that is, M recording quality inspection rules can be adopted for each sentence to obtain the inspection result. For example, the sentence meaning inspection 1 in the figure is the quality inspection of the first sentence based on the sentence meaning quality inspection rules; the sentence meaning inspection 2 is the quality inspection of the first sentence based on the sentence meaning quality inspection rules; the keyword detection 1 is the quality inspection of the first sentence based on the keyword quality inspection rules; the keyword detection 2 is the quality inspection of the first sentence based on the keyword quality inspection rules.

[0111] In one embodiment, Figure 4 As shown, S400 includes:

[0112] S420, determining the agent quality inspection quantitative value according to the abnormal level of the recording file and the agent recording characteristics.

[0113] S440, determining a user quality inspection quantization value according to the abnormality level of the recording file and the user recording characteristics.

[0114] S460, determining a quantitative value of the interaction quality inspection between the agent and the user according to the agent recording characteristics and the user recording characteristics.

[0115] S480, determining an abnormality quantization value of the recording file according to the seat quality inspection quantization value, the user quality inspection quantization value, and the interaction quality inspection quantization value.

[0116] Among them, the agent quality inspection quantitative value refers to the score value of the abnormality level of the agent within the preset historical time period. The user quality inspection quantitative value refers to the score value of the abnormality level of the user within the preset historical time period. The interaction quality inspection quantitative value refers to the score value of the abnormality level of the interaction process between the user and the agent in the current recording file. The agent recording characteristics include at least one of the abnormality level of the agent's historical recordings within the preset historical time period, the total duration of the recordings corresponding to the agent, and the call duration between the agent and the user corresponding to the current recording file; the user recording characteristics include at least one of the abnormality level of the user's historical recordings within the preset historical time period, the total duration of the recordings corresponding to the user, and the call duration between the user and the agent corresponding to the current recording file.

[0117] Specifically, the seat quality inspection quantitative value is determined according to the abnormality level of the recording file and the characteristics of the seat recording, including: determining the degree of seat abnormality according to the abnormality level of the recording file and the abnormality level of the seat historical recording within a preset historical time period, and quantifying the degree of seat abnormality with a score to obtain the seat quality inspection quantitative value.

[0118] The user quality inspection quantitative value is determined according to the abnormality level of the recording file and the user recording characteristics, including: determining the user abnormality degree according to the abnormality level of the recording file and the abnormality level of the user's historical recordings within a preset historical time period, and quantifying the user abnormality degree with a score to obtain the user quality inspection quantitative value.

[0119] Determine a quantitative value of the interaction quality inspection between the agent and the user according to the agent recording features and the user recording features, including: determine a quality inspection result of the interaction time between the agent and the user according to features related to time in the agent recording features and features related to time in the user recording features, and quantify the quality inspection result of the interaction time between the agent and the user using a score to obtain a quantitative value of the interaction quality inspection.

[0120] Starting from the three dimensions of the seat quality inspection quantization value, the user quality inspection quantization value and the interactive quality inspection quantization value, determine the abnormal quantization value of the recording file at a deeper level. Among them, the abnormal quantization value of the recording file is determined according to the seat quality inspection quantization value, the user quality inspection quantization value and the interactive quality inspection quantization value, including but not limited to: taking the weighted sum of the seat quality inspection quantization value, the user quality inspection quantization value and the interactive quality inspection quantization value to determine the abnormal quantization value of the recording file. Furthermore, the sum of the preset weights of the seat quality inspection quantization value, the preset weights of the user quality inspection quantization value and the preset weights of the interactive quality inspection quantization value needs to be 1. In practical applications, the preset weights of the seat quality inspection quantization value generally do not exceed 0.6, and the preset weights of the user quality inspection quantization value generally do not exceed 0.3.

[0121] In an actual application, the preset weight of the agent quality inspection quantization value is set to 0.6, the preset weight of the user quality inspection quantization value is set to 0.3, and the preset weight of the interactive quality inspection quantization value is set to 0.1. Then the abnormal quantization value of the recording file = 0.6*agent quality inspection quantization value+0.3*user quality inspection quantization value and +0.1*interaction quality inspection quantization value.

[0122] In this embodiment, by quantifying the abnormality level of the agent, the abnormality level of the user, and the abnormality level of the interaction process between the user and the agent for quality inspection, a deeper and more comprehensive quality inspection of the recording file is performed from a unified multi-dimensional perspective, thereby improving the accuracy of the quality inspection of the recording file.

[0123] In one embodiment, the recording feature includes an abnormality level of the agent's historical recording within a preset historical time period; determining the agent quality inspection quantitative value according to the abnormality level of the recording file and the agent recording feature includes:

[0124] According to the abnormality level of the agent's historical recordings within the preset historical time period, obtain the agent's historical abnormal recordings with the same abnormality level as the recording file; according to the agent's historical abnormal recordings, determine the average daily number of abnormal recordings of the agent within the preset historical time period; according to the preset weight of the preset historical time period, perform weighted summation on the average daily number of abnormal recordings to obtain the agent quality inspection quantitative value.

[0125] The average number of abnormal recordings per day is the ratio of the number of abnormal recordings per seat in the preset historical time period to the total number of days. There may be more than one preset historical time period. The seat quality inspection quantification value can be obtained by quantifying the average number of abnormal recordings per day.

[0126] Specifically, first, the historical recordings of the seat within a preset historical time period are obtained, including: obtaining the historical recordings within the historical time period, and based on the seat identification transmitted through the message bus, obtaining the historical recordings of the seat corresponding to the seat identification and the abnormality level corresponding to the historical recordings of the seat from the historical recordings within the historical time period. It should be noted that the abnormality level corresponding to the historical recordings of the seat is also obtained after quality inspection of the historical recordings of the seat, and the process of quality inspection of the historical recordings of the seat is consistent with the above-mentioned quality inspection process of the current recording file, namely, including: converting the historical recordings of the seat into texts of the historical recordings of the seat; performing semantic analysis on the texts of the historical recordings of the seat to obtain the seat semantic information results of the historical recordings of the seat; using M recording quality inspection rules to perform quality inspection and analysis on the recording texts and seat semantic information results of the historical recordings of the seat to obtain the abnormality level of the historical recordings of the seat.

[0127] Secondly, according to the abnormality level of the seat historical recordings within the preset historical time period, the seat historical abnormal recordings with the same abnormality level as the recording file are obtained, and according to the seat historical abnormal recordings, the average daily number of abnormal recordings of the seat within the preset historical time period is determined. For example, the seat historical recordings within the preset historical time period include: seat historical recording Z, seat historical recording X, seat historical recording C, and seat historical recording V, wherein the abnormality level of seat historical recording Z is level 1, the abnormality level of seat historical recording X is level 3, the abnormality level of seat historical recording C is level 2, the abnormality level of seat historical recording V is level 1, and the abnormality level of the current recording file is level 1. It is determined that the seat historical abnormal recordings with the same abnormality level as the recording file are seat historical recording Z and seat historical recording V, and the average daily number of abnormal recordings of the seat within the preset historical time period is determined to be 2. In actual applications, the number of seat historical recordings is greater, but the actual calculation process is based on the same principle as the above calculation process.

[0128] It should be explained that the reason why the abnormal historical recordings of the agents with the same abnormal level as the recording files are obtained is that the average daily abnormal number of the abnormal historical recordings of the agents with the same abnormal level as the recording files is sufficient to reflect the abnormal degree of the abnormal historical recordings of the agents.

[0129] Further, the preset historical time period may be at least one historical time period, for example, the preset historical time period may be within the past 30 days and within the past 7 days. In the case where the preset historical time period includes within the past 30 days and within the past 7 days, the agent quality inspection quantitative value is determined according to the abnormality level of the recording file and the agent recording characteristics, including: the first step: according to the abnormality level of the agent's historical recordings in the past 30 days, obtain the agent's historical abnormal recordings with the same abnormality level as the recording file, and determine the average daily number of abnormal recordings of the agent in the past 30 days according to the agent's historical abnormal recordings; the second step: according to the abnormality level of the agent's historical recordings in the past 7 days, obtain the agent's historical abnormal recordings with the same abnormality level as the recording file; according to the agent's historical abnormal recordings, determine the average daily number of abnormal recordings of the agent in the past 7 days; the third step: according to the first preset weight in the past 30 days and the second preset weight in the past 7 days, perform weighted summation on the average daily number of abnormal recordings to obtain the agent quality inspection quantitative value.

[0130] Generally speaking, when the preset historical time period is multiple historical time periods, a smaller preset weight can be set for the earlier historical time periods, and a larger preset weight can be set for the more recent historical time periods, so that when the abnormal degree of the agent is quantitatively evaluated, more attention is paid to the recent abnormalities. For example, when the preset historical time period includes the past 30 days and the past 7 days, the preset weight for the past 30 days can be set to 0.3, and the preset weight for the past 7 days can be set to 0.7. At this time, the agent quality inspection quantitative value = 0.3*the average daily number of abnormal recordings of the agent in the past 30 days + 0.7*the average daily number of abnormal recordings of the agent in the past 7 days.

[0131] In this embodiment, by calculating the weighted sum of the number of historical abnormal recordings of the agent with the same abnormal level as the recording file in different historical time periods, the abnormality level of the agent can be accurately and uniformly quantitatively evaluated, so that the obtained agent quality inspection quantitative value is less affected by subjective factors and has higher objectivity and accuracy.

[0132] In one embodiment, the seat quality inspection quantification value is determined based on the abnormal level of the recording file and the characteristics of the seat recording, and also includes: obtaining the seat historical abnormal recording with the same abnormal level as the recording file according to the abnormal level of the seat historical recording within the preset historical time period; determining the number of abnormal recordings of the seat within the preset historical time period according to the seat historical abnormal recording; and performing weighted summation on the number of abnormal recordings according to the preset weight of the preset historical time period to obtain the seat quality inspection quantification value. In other words, the seat quality inspection quantification value can be obtained not only by the weighted sum of the average daily number of abnormal recordings within different historical time periods, but also by the weighted sum of the number of abnormal recordings within different historical time periods. The specific determination method can be determined by the quality inspector.

[0133] In one embodiment, the recording feature includes the abnormality level of the user's historical recordings in different historical time periods; determining the user quality inspection quantitative value according to the abnormality level of the recording file and the user's recording feature includes:

[0134] According to the abnormality level of the user's historical recordings within the preset historical time period, obtain the user's historical abnormal recordings with the same abnormality level as the recording file; according to the user's historical abnormal recordings, determine the average daily number of abnormal recordings of the user within the preset historical time period; according to the preset weight of the preset historical time period, perform weighted summation on the average daily number of abnormal recordings to obtain the user quality inspection quantitative value.

[0135] Among them, there is more than one preset historical time period.

[0136] Specifically, first, the user historical recordings within a preset historical time period are obtained, including: obtaining the historical recordings within the historical time period, and based on the user identifier transmitted through the message bus, obtaining the user historical recordings corresponding to the user identifier and the abnormality level corresponding to the user historical recordings from the historical recordings within the historical time period. It should be noted that the abnormality level corresponding to the user historical recording is also obtained after quality inspection of the user historical recording, and the process of quality inspection of the user historical recording is consistent with the above-mentioned quality inspection process of the current recording file, that is, including: converting the user historical recording into a user historical recording text; performing semantic analysis on the user historical recording text to obtain the user semantic information result of the user historical recording; using M recording quality inspection rules to perform quality inspection and analysis on the recording text and user semantic information result of the user historical recording to obtain the abnormality level of the user historical recording.

[0137] Secondly, according to the abnormality level of the user's historical recordings within the preset historical time period, obtain the user's historical abnormal recordings with the same abnormality level as the recording file, and determine the average daily number of abnormal recordings of the user within the preset historical time period based on the user's historical abnormal recordings. For example, the user's historical recordings within the preset historical time period include: user historical recordings Q, user historical recordings W, and user historical recordings E, wherein the abnormality level of user historical recordings Q is level 1, the abnormality level of user historical recordings W is level 1, the abnormality level of user historical recordings E is level 1, and the abnormality level of the current recording file is level 1. It is determined that the user's historical abnormal recordings with the same abnormality level as the recording file are user historical recordings Q, user historical recordings W, and user historical recordings E, and it is determined that the average daily number of abnormal recordings of the user within the preset historical time period is 3.

[0138] Further, similar to the process of obtaining the seat quality inspection quantitative value, the preset historical time period can be at least one historical time period, for example, the preset historical time period can be within the past 30 days and within the past 7 days. In the case where the preset historical time period includes within the past 30 days and within the past 7 days, the user quality inspection quantitative value is determined according to the abnormal level of the recording file and the user recording characteristics, including: the first step: according to the abnormal level of the user's historical recordings of the user in the past 30 days, obtain the user's historical abnormal recordings with the same abnormal level as the recording file, and determine the user's average daily abnormal recording number in the past 30 days according to the user's historical abnormal recordings; the second step: according to the abnormal level of the user's historical recordings of the user in the past 7 days, obtain the user's historical abnormal recordings with the same abnormal level as the recording file; according to the user's historical abnormal recordings, determine the user's average daily abnormal recording number in the past 7 days; the third step: according to the first preset weight in the past 30 days and the second preset weight in the past 7 days, perform weighted summation on the average daily abnormal recording number to obtain the user quality inspection quantitative value.

[0139] Similar to the process of obtaining the quantitative value of the agent quality inspection, when the preset historical time period is multiple historical time periods, a smaller preset weight can be set for the earlier historical time period, and a larger preset weight can be set for the more recent historical time period, so that when the user's abnormality is quantitatively evaluated, more attention is paid to recent abnormalities. For example, when the preset historical time period includes the past 30 days and the past 7 days, the preset weight for the past 30 days can be set to 0.3, and the preset weight for the past 7 days can be set to 0.7. At this time, the user quality inspection quantitative value = 0.3*the average number of abnormal recordings of the user in the past 30 days + 0.7*the average number of abnormal recordings of the user in the past 7 days.

[0140] In this embodiment, by calculating the weighted sum of the number of historical abnormal recordings of users with the same abnormal level as the recording files in different historical time periods, an accurate and unified quantitative evaluation of the user's abnormality level can be performed, so that the obtained user quality inspection quantitative value is less affected by subjective factors and has higher objectivity and accuracy.

[0141] In one embodiment, determining the user quality inspection quantization value based on the abnormal level of the recording file and the user recording characteristics also includes: obtaining the user's historical abnormal recordings with the same abnormal level as the recording file based on the abnormal level of the user's historical recordings in a preset historical time period; determining the number of abnormal recordings of the user in the preset historical time period based on the user's historical abnormal recordings; and performing weighted summation of the number of abnormal recordings according to the preset weight of the preset historical time period to obtain the user quality inspection quantization value. In other words, the user quality inspection quantization value can be obtained not only by the weighted sum of the average daily number of abnormal recordings in different historical time periods, but also by the weighted sum of the number of abnormal recordings in different historical time periods.

[0142] In one embodiment, determining the interaction quality inspection quantization value between the agent and the user based on the agent recording features and the user recording features includes:

[0143] Based on the agent recording characteristics, the duration of all recording files of the agent in the current time period is obtained, and the duration of the agent's current call is determined based on the duration of all recording files of the agent in the current time period; based on the agent recording characteristics and the user recording characteristics, the duration of all recording files between the agent and the user in the current time period is obtained; based on the duration of all recording files between the agent and the user in the current time period, the duration of the interactive call between the agent and the user is determined; based on the ratio of the interactive call duration to the agent's current call duration, the interactive quality inspection quantitative value is determined.

[0144] The recording feature includes the duration of the recording files in the current time period. The agent recording feature refers to the recording feature of the agent, that is, the duration of all the recording files of the agent in the current time period. The user recording feature refers to the recording feature of the user, that is, the duration of all the recording files of the user in the current time period.

[0145] Specifically, the abnormality evaluation related to the duration of the interaction between the agent and the user is determined by the duration of all recording files of the agent in the current time period and the duration of all recording files between the agent and the user in the current time period. Generally, the current time period refers to the current day.

[0146] More specifically, based on the agent recording characteristics, the duration of all recording files of the agent in the current time period is obtained, and the current call duration of the agent is determined based on the duration of all recording files of the agent in the current time period. In other words, all recording files of the agent corresponding to the agent ID on that day are obtained, and the duration corresponding to each recording file is determined, and then the duration corresponding to each recording file is added together to obtain the current call duration of the agent.

[0147] Based on the user recording feature, the duration of all recording files of the user in the current time period is obtained. Since the agent and the user will have a call, that is, the agent recording feature includes the duration of the recording file of the call with a specific user, and the user recording feature also includes the duration of the call with a specific agent. Therefore, based on the agent recording feature and the user recording feature, all the recording files of the call between the specific agent and the specific user can be determined, and then the duration of all the recording files between the agent and the user in the current time period can be obtained.

[0148] The duration of the interactive call between the agent and the user is determined based on the duration of all recording files between the agent and the user in the current time period, that is, all recording files between the agent corresponding to the agent ID and the user corresponding to the agent ID in the day are obtained, and the duration corresponding to each recording file is determined, and then the duration corresponding to each recording file is superimposed to obtain the duration of the interactive call between the agent and the user. It should be explained that the agent corresponding to the agent ID corresponding to the current recording file and the user corresponding to the agent ID may not have talked only once, but may have talked many times. Therefore, it is necessary to obtain the duration of all recording files between the agent corresponding to the agent ID and the user corresponding to the agent ID in the current time period to determine the duration of the interactive call between the agent and the user in the current time period.

[0149] Furthermore, the ratio of the duration of all recording files between the agent and the user in the current time period to the duration of all recording files of the agent in the current time period is determined as the interaction quality inspection quantization value between the agent and the user.

[0150] In an actual application, the quality inspection is performed on recording file 1, and the two parties in the conversation are seat A and user B. All recording files of seat A on that day are recording file 1, recording file 2, recording file 3, recording file 4, and recording file 5, among which recording file 1 and recording file 3 are recording files between seat A and user B.

[0151] Get the duration of all recording files of Agent A on that day, that is, get the total duration of recording files 1, 2, 3, 4, and 5, and get the current call duration of Agent A is 200 minutes; get the duration of all recording files in which the two parties of the conversation are Agent A and User B on that day, that is, get the total duration of recording files 1 and 3, and get the interactive call duration between Agent A and User B is 60 minutes; at this time, the interactive quality inspection quantification value is 60 minutes / 200 minutes=0.6.

[0152] In this embodiment, the interaction quality inspection quantification value can be accurately determined by the ratio of the interaction call duration between the agent and the user in the current time period to the current call duration of the agent in the current time period. For each recording file, the interaction quality inspection quantification value is determined by a unified calculation process, which is less affected by subjective factors and has higher objectivity and accuracy.

[0153] In one embodiment, when a user makes a call to an agent, the abnormality level of the recording file can also be used to assign a corresponding agent to the caller. Specifically, the server responds to the target user's call processing request, obtains the risk assessment value of the caller and the latest abnormality assessment values ​​of all currently idle agents; obtains the target idle agent with the smallest sum of the risk assessment value of the caller and the latest abnormality assessment values ​​of each idle agent; and assigns the caller's call to the target idle agent for processing.

[0154] Furthermore, obtaining a risk assessment value for the incoming caller includes: obtaining the number of abnormal power calls and the total number of power calls made by the incoming caller within a first historical time period, and the number of abnormal power calls made by the incoming caller within a second historical time period; determining the risk assessment value of the incoming caller based on the number of abnormal power calls and the total number of power calls made by the incoming caller within the first historical time period, and the number of abnormal power calls made by the incoming caller within the second historical time period.

[0155] When the first historical time is within the past 30 days and the second historical time is within the past 7 days, obtain the number of abnormal calls and the total number of calls made by the caller in the past 30 days, and obtain the number of abnormal calls made by the caller in the past 7 days. Abnormal calls refer to calls with abnormal identification obtained by evaluating the calls made by the caller using the recording quality inspection rules. At this time, the risk assessment value of the caller = the number of abnormal calls made by the caller in the past 30 days / the total number of calls made by the caller in the past 30 days*the number of abnormal calls made by the caller in the past 7 days.

[0156] Obtain the latest abnormality assessment value of all currently idle seats, including: for any idle seat, obtaining the abnormality assessment value of the idle seat within the third historical time length and the estimated abnormality assessment value of each day within the fourth historical time length; determine the latest abnormality assessment value of the idle seat according to the abnormality assessment value of the idle seat within the third historical time length and the estimated abnormality assessment value of each day within the fourth historical time length.

[0157] In the case where the third historical time period includes the last month and the fourth historical time period includes the current month, obtain the abnormal evaluation value of each idle seat in the last month and the estimated abnormal evaluation value of each day in the current month; determine the latest abnormal evaluation value of the idle seat based on the abnormal evaluation value of the idle seat in the last month and the estimated abnormal evaluation value of each day in the current month. The abnormal evaluation value of the idle seat in the last month means that the idle seat has an initial evaluation value every month. When the idle seat has a recording file that represents an abnormality in the month, determine the abnormality level of the recording file that represents the abnormality, and based on the abnormality level of the abnormal recording file, deduct points from the initial evaluation value of the idle seat, and each abnormality level corresponds to a deduction value.

[0158] Therefore, the latest abnormal evaluation value of the current idle seat = the abnormal evaluation value of the idle seat in the previous month * the first weighting coefficient + (the number of days from the 1st of the month to the present) * the number of violations corresponding to the recording of the seat on that day * the second weighting coefficient. The sum of the first weighting coefficient and the second weighting coefficient is 1. In actual applications, the first weighting coefficient can be 0.3 and the second weighting coefficient can be 0.7.

[0159] For example, let today be December 6th, the number of violations of agent N today is 1, the number of violations of agent N in the past 7 days is 2, the number of violations of agent N in the past 30 days is 3, the number of calls from user D in the past 30 days is 4, and the initial evaluation value of agent N last month is 0.

[0160] If there were 3 violations in the recording of agent N last month, and each violation corresponds to a deduction of 1 point, then the abnormal evaluation value of agent N last month is 0-3*1=-3 points. If the first weighting coefficient is 0.3 and the second weighting coefficient is 0.7, then the latest abnormal evaluation value of agent N = -3 *0.3+6*3*0.7. The abnormal evaluation value of caller D = 3 / 4*2.

[0161] In another example, Figure 5As shown, there are three seats N, M, and F, and the latest anomaly evaluation values ​​of the three seats N, M, and F are 11.7, 5, and 12 respectively. When caller D calls, the anomaly evaluation value of caller D is 1.5. The anomaly evaluation value of caller D is added to the latest anomaly evaluation values ​​of seats N, M, and F respectively, and the added results are sorted from small to large, and the sorting results are 6.5, 13.2, and 13.5. It can be determined that the sum of the anomaly evaluation value of caller D and the latest anomaly evaluation value of seat M is the smallest, so the call from caller D is assigned to seat M.

[0162] In one embodiment, Figure 6 As shown in FIG. 1 , the specific process of the recording quality inspection method includes:

[0163] S1. The agent answers the user's call through the agent terminal. After the call is hung up, the recording of each call will be stored locally.

[0164] S2. Convert the formats of multiple recording files of the day stored locally through the agent server. For example, the multiple recording files include recording file 1, recording file 2, recording file 3, recording file 4, ... recording file n, etc.

[0165] S3. Upload to the cloud storage system in a unified format to obtain the recording addresses where multiple recording files are stored.

[0166] S4. The agent server sends the recording addresses of the multiple recording files, and the agent identification and user identification corresponding to each recording file to the message bus, and the message bus sends the recording addresses of the multiple recording files, and the agent identification and user identification corresponding to each recording file to the recording quality inspection server. When sending the recording addresses of the multiple recording files through the message bus, the recording addresses of the multiple recording files can be sorted and sent according to the acquisition time of the recording files.

[0167] S5. The recording quality inspection server extracts the recording addresses of the multiple recording files, and information such as the seat identification and user identification corresponding to each recording file.

[0168] S6. Based on the recording addresses of the multiple recording files, download the recording files corresponding to the multiple recording addresses from the cloud storage system.

[0169] S7. Perform ASR translation on the multiple recording files to obtain recording texts corresponding to the multiple recording files.

[0170] Furthermore, the recording quality inspection server performs recording quality inspection on multiple recording texts one by one. Taking the recording file 1 obtained when the seat X answers the call of the caller Y and hangs up the call as an example, the specific steps of performing quality inspection on any one of the multiple recording texts are as follows:

[0171] S8. Perform NLP semantic analysis on the recorded text 1' corresponding to the recorded file 1 to obtain the semantic information results of the recorded text 1'. The semantic information results include: the sentence meaning, keywords, emotions, speaking speed, interruption, etc. of the recorded text 1'.

[0172] S9. Use the sentence meaning recording quality inspection rule to detect the recording text 1' and the semantic information results, and obtain the detection result corresponding to the recording text 1' under the sentence meaning recording quality inspection rule, including an abnormal mark, and the initial abnormal level in the detection result is level 1. Similarly, use the keyword recording quality inspection rule and the speech speed recording quality inspection rule to detect the recording text 1' and the semantic information results, and obtain the detection result corresponding to the recording text 1' under the speech speed recording quality inspection rule and the keyword recording quality inspection rule, including an abnormal mark, and the initial abnormal level in the detection result corresponding to the speech speed recording quality inspection rule is level 1, and the initial abnormal level in the detection result corresponding to the keyword recording quality inspection rule is level 3.

[0173] At this time, the highest initial abnormality level, that is, level 3, is determined as the abnormality level of the audio recording 1'.

[0174] S10. Manually review the abnormality level of the recording file.

[0175] S11. When the review result is consistent with the abnormality level of the recording file, the abnormality level of the recording text 1' is set to level 3.

[0176] S12: Obtain historical recordings of agent X in the past 30 days, historical recordings of agent X in the past 7 days, historical recordings of caller Y in the past 30 days, and historical recordings of caller Y in the past 7 days. And determine the abnormality level of each historical recording of agent X in the past 30 days, the abnormality level of each historical recording of agent X in the past 7 days, the abnormality level of each historical recording of caller Y in the past 30 days, and the abnormality level of each historical recording of caller Y in the past 7 days.

[0177] Determine the average daily number of abnormalities with level 3 in each historical recording of agent X in the past 30 days, the average daily number of abnormalities with level 3 in each historical recording of agent X in the past 7 days, the average daily number of abnormalities with level 3 in each historical recording of caller Y in the past 30 days, and the average daily number of abnormalities with level 3 in each historical recording of caller Y in the past 7 days.

[0178] In addition, it is also necessary to obtain the total duration of all recording files of the agent on that day, and use the total duration of all recording files of the agent on that day to determine the current call duration of the agent; and obtain the interactive call duration of all recording files of calls between the agent and the user on that day.

[0179] Expressions based on the abnormal quantization values ​​of the recording file:

[0180] The abnormality quantification value of the recording file = a1*agent quality inspection quantification value+a2*user quality inspection quantification value+a3*interaction quality inspection quantification value; where a1+a2+a3=1.

[0181] User quality inspection quantification value = b1*the average daily number of abnormalities with level 3 in each historical recording of user Y in the past 30 days + b2*the average daily number of abnormalities with level 3 in each historical recording of user Y in the past 30 days; where b1+ b2=1, b1<b2.

[0182] Agent quality inspection quantification value = c1*average number of abnormalities with level 3 in each historical recording of agent X in the past 30 days + c2*average number of abnormalities with level 3 in each historical recording of agent X in the past 7 days; where c1+c2=1, c1<c2.

[0183] Interaction quality inspection quantitative value = interactive call duration of all recorded files of calls between the agent and the user on that day / total duration of all recorded files of the agent on that day.

[0184] Then the anomaly quantification value of the recording file = a1*(c1*the average daily number of anomalies with level 3 in historical recordings of agent X in the past 30 days + c2*the average daily number of anomalies with level 3 in historical recordings of agent X in the past 7 days)+a2*b1*(the average daily number of anomalies with level 3 in historical recordings of user Y in the past 30 days + b2*the average daily number of anomalies with level 3 in historical recordings of user Y in the past 30 days)+a3*(the interactive call duration of all recording files of calls between the agent and the user on that day / the total duration of all recording files of the agent on that day).

[0185] Furthermore, the above method is used to obtain the abnormal quantification value of each recording file on that day, and the recording files are sorted from large to small according to the abnormal quantification value of each recording file, and the recording files ranked first are regarded as high-risk recordings that need to be processed urgently. For example, when the sorting results are recording file 4, recording file 5, ..., recording file n, recording file 2, ..., recording file 1, it means that recording file 4 is the recording file that needs to be processed first.

[0186] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0187] Based on the same inventive concept, the embodiment of the present application also provides a recording quality inspection device for implementing the recording quality inspection method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more recording quality inspection device embodiments provided below can refer to the limitations of the recording quality inspection method above, and will not be repeated here.

[0188] In an exemplary embodiment, Figure 7 As shown, a recording quality inspection device is provided, including: a determination module 200, a processing module 400 and a quality inspection module 600, wherein:

[0189] A determination module 200 is used to determine the abnormality level of the recording file, as well as the seat recording features and user recording features in the recording file;

[0190] The processing module 400 is used to determine the abnormality quantification value of the recording file according to the abnormality level of the recording file, the agent recording characteristics and the user recording characteristics;

[0191] The quality inspection module 600 is used to determine the quality inspection result of the recording file based on the abnormal quantitative value of the recording file.

[0192] In one embodiment, the determination module 200 is also used to convert the recording file into a recording text; perform semantic analysis on the recording text to obtain the semantic information result of the recording file; perform quality inspection and analysis on the recording text and semantic information result of the recording file to obtain the abnormality level of the recording file.

[0193] In one embodiment, the determination module 200 is also used to adopt M recording quality inspection rules to detect the recording text and semantic information results of the recording file, and obtain the detection results corresponding to the recording file under the M recording quality inspection rules; according to the detection results corresponding to the recording file under each recording quality inspection rule, determine the abnormality level of the recording file, where M is a positive integer greater than 1.

[0194] In one embodiment, the detection result includes an abnormality identification and an abnormality initial level; the determination module 200 is also used to obtain the highest abnormality initial level among the abnormality initial levels if the abnormality identification corresponding to the recording file under M types of recording quality inspection rules meets the preset conditions; and determine the highest abnormality initial level as the abnormality level of the recording file.

[0195] In one embodiment, the preset condition includes any one of the following: the detection results of the recording file under M recording quality inspection rules all include abnormal identification; the detection results of the recording file under M recording quality inspection rules include at least one abnormal identification.

[0196] In one embodiment, the detection result includes an initial abnormality level, and the determination module 200 is also used to detect the target threshold interval of the recording text and semantic information results of the recording file according to the threshold interval preset in the recording quality inspection rule for any recording quality inspection rule; determine the initial abnormality level corresponding to the target threshold interval as the initial abnormality level of the recording file.

[0197] In one embodiment, the processing module 400 is further used to determine the agent quality inspection quantization value based on the abnormality level of the recording file and the agent recording characteristics; determine the user quality inspection quantization value based on the abnormality level of the recording file and the user recording characteristics; determine the interaction quality inspection quantization value between the agent and the user based on the agent recording characteristics and the user recording characteristics; determine the abnormality quantization value of the recording file based on the agent quality inspection quantization value, the user quality inspection quantization value and the interaction quality inspection quantization value.

[0198] In one embodiment, the agent recording feature includes the abnormality level of the agent's historical recordings within a preset historical time period; the processing module 400 is also used to obtain the agent's historical abnormal recordings with the same abnormality level as the recording file based on the abnormality level of the agent's historical recordings within the preset historical time period; determine the average daily number of abnormal recordings of the agent within the preset historical time period based on the agent's historical abnormal recordings; and perform weighted summation on the average daily number of abnormal recordings according to the preset weight of the preset historical time period to obtain the agent quality inspection quantitative value.

[0199] In one embodiment, the user recording feature includes the abnormality level of the user's historical recordings in different historical time periods; the processing module 400 is also used to obtain the user's historical abnormal recordings with the same abnormality level as the recording file according to the abnormality level of the user's historical recordings in a preset historical time period; determine the average daily number of abnormal recordings of the user in the preset historical time period based on the user's historical abnormal recordings; and perform weighted summation on the average daily number of abnormal recordings according to the preset weights of the preset historical time period to obtain a user quality inspection quantification value.

[0200] In one embodiment, the processing module 400 is also used to obtain the duration of all recording files of the agent in the current time period based on the agent recording characteristics, and determine the duration of the agent's current call based on the duration of all recording files of the agent in the current time period; obtain the duration of all recording files between the agent and the user in the current time period based on the agent recording characteristics and the user recording characteristics; determine the duration of the interactive call between the agent and the user based on the duration of all recording files between the agent and the user in the current time period; and determine the interactive quality inspection quantification value based on the ratio of the interactive call duration to the agent's current call duration.

[0201] Determine the modules in the above recording quality inspection device, and all or part of them can be implemented by software, hardware, or a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0202] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as recording text, initial quality inspection results and final quality inspection results. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a recording quality inspection method is implemented.

[0203] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0204] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0205] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0206] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0207] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0208] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0209] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0210] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A recording quality inspection method, It is characterized in that The method comprises: Determine the abnormality level of the recording file, as well as the agent recording features and the user recording features in the recording file; Determine an abnormality quantification value of the recording file according to the abnormality level of the recording file, the agent recording feature and the user recording feature; Based on the abnormal quantification value of the recording file, a quality inspection result of the recording file is determined.

2. The method according to claim 1, wherein determining the abnormality level of the recording file, include: Converting the recording file into a recording text; Performing semantic analysis on the recorded text to obtain semantic information results of the recorded file; A quality inspection and analysis is performed on the recording text and semantic information results of the recording file to obtain an abnormality level of the recording file.

3. The method according to claim 2, It is characterized in that The quality inspection and analysis of the recording text and semantic information results of the recording file to obtain the abnormality level of the recording file includes: Using M recording quality inspection rules, the recording text and semantic information results of the recording file are detected to obtain the detection result corresponding to the recording file under each recording quality inspection rule; The abnormality level of the recording file is determined according to the detection results corresponding to the recording file under M recording quality inspection rules, where M is a positive integer greater than 1.

4. The method according to claim 3, It is characterized in that The detection result includes an abnormality mark and an initial abnormality level; the abnormality level of the recording file is determined according to the detection result corresponding to each recording quality inspection rule of the recording file, including: If the abnormal identification corresponding to the recording file under the M recording quality inspection rules meets the preset conditions, the highest abnormal initial level among the abnormal initial levels is obtained; The highest initial abnormality level is determined as the abnormality level of the recording file.

5. The method according to claim 4, It is characterized in that The preset condition includes any one of the following: The detection results of the recording files under the M recording quality inspection rules all include abnormal identification; The detection results of the recording file under M recording quality inspection rules include at least one abnormality mark.

6. The method according to claim 3, It is characterized in that The detection result includes an abnormal initial level, and the M recording quality inspection rules are used to detect the recording text and semantic information results of the recording file, and the detection result corresponding to the recording file under each recording quality inspection rule is obtained, including: For any recording quality inspection rule, according to the threshold interval preset in the recording quality inspection rule, the target threshold interval of the recording text and semantic information results of the recording file is detected; An initial abnormality level corresponding to the target threshold interval is determined as the initial abnormality level of the recording file.

7. The method according to any one of claims 1 to 6, It is characterized in that The determining, according to the abnormality level of the recording file, the agent recording feature and the user recording feature, the abnormality quantification value of the recording file includes: Determine the agent quality inspection quantitative value according to the abnormal level of the recording file and the agent recording characteristics; Determining a user quality inspection quantification value according to the abnormality level of the recording file and the user recording characteristics; Determining a quantitative value of the interaction quality inspection between the agent and the user according to the agent recording features and the user recording features; The abnormality quantization value of the recording file is determined according to the seat quality inspection quantization value, the user quality inspection quantization value and the interaction quality inspection quantization value.

8. The method according to claim 7, It is characterized in that The agent recording feature includes an abnormality level of the agent historical recording within a preset historical time period; determining the agent quality inspection quantitative value according to the abnormality level of the recording file and the agent recording feature includes: According to the abnormality level of the seat historical recording within the preset historical time period, obtain the seat historical abnormal recording with the same abnormality level as the recording file; According to the historical abnormal recordings of the seat, determine the average daily number of abnormal recordings of the seat in the preset historical time period; According to the preset weight of the preset historical time period, the average daily number of abnormal recordings is weighted and summed to obtain the seat quality inspection quantification value.

9. The method according to claim 7, It is characterized in that The user recording feature includes an abnormality level of the user's historical recording within a preset historical time period; and determining the user quality inspection quantitative value according to the abnormality level of the recording file and the user recording feature includes: According to the abnormality level of the user's historical recordings within the preset historical time period, obtaining the user's historical abnormal recordings with the same abnormality level as the recording file; Determine the average daily number of abnormal recordings of the user in the preset historical time period based on the historical abnormal recordings of the user; According to the preset weight of the preset historical time period, the average daily number of abnormal recordings is weighted and summed to obtain the user quality inspection quantification value.

10. The method according to claim 7, It is characterized in that The step of determining the interaction quality inspection quantization value between the agent and the user according to the agent recording feature and the user recording feature includes: Based on the agent recording feature, the duration of all recording files of the agent in the current time period is obtained, and the current call duration of the agent is determined according to the duration of all recording files of the agent in the current time period; Based on the seat recording feature and the user recording feature, obtaining the duration of all recording files between the seat and the user in the current time period; Determine the duration of the interactive call between the agent and the user according to the duration of all recording files between the agent and the user in the current time period; The interaction quality inspection quantization value is determined according to the ratio of the interaction call duration to the current call duration of the agent.

11. A recording quality inspection device, It is characterized in that The device comprises: A determination module, used to determine the abnormality level of the recording file, as well as the agent recording features and the user recording features in the recording file; A processing module, configured to determine an abnormality quantification value of the recording file according to the abnormality level of the recording file, the agent recording feature and the user recording feature; The quality inspection module is used to determine the quality inspection result of the recording file based on the abnormal quantitative value of the recording file.

12. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 10 are implemented.

13. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

14. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

Images