Knowledge base updating method, device, electronic device and storage medium for intelligent customer service

Abstract

The present disclosure provides a knowledge base update method, device, electronic device and storage medium for intelligent customer service, which belongs to the field of artificial intelligence technology. The method includes: screening out candidate recording files according to user satisfaction with multiple recording files; obtaining abnormal recording files based on preprocessed files corresponding to the candidate recording files; and updating the knowledge base based on updated question and answer pairs generated based on target question data in the abnormal recording files. The present disclosure does not need to play the recording files completely. Based on the user's satisfaction with the answer data provided by the intelligent customer service, the candidate recording files are screened, and on this basis, the candidate recording files are preprocessed to obtain abnormal recording files. The above-mentioned screening and preprocessing processes are used to greatly shorten the annotation time of abnormal recording files and improve the update speed of the knowledge base.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a knowledge base updating method, device, electronic device and storage medium for intelligent customer service. Background Art

[0002] With the development of AI (Artificial Intelligence) technology, AI voice assistants are becoming more and more popular in daily life, and the interaction between people and intelligent voice is becoming simpler and more efficient. Enterprises use intelligent customer service to communicate with users instead of manual customer service, saving labor costs and improving enterprise service efficiency. When working, intelligent customer service first identifies the user's question data, then queries the answer data corresponding to the question data from the question and answer pairs stored in the knowledge base, and then provides the queried answer data to the user. In order to ensure the comprehensiveness and accuracy of the question and answer pairs in the knowledge base, so as to better serve users, the knowledge base of intelligent customer service needs to be updated.

[0003] The relevant technology mainly uses the following method to update the knowledge base of intelligent customer service: play the audio files recorded during the interaction between the intelligent customer service and the user; during the playback of the audio files, special annotation personnel mark the abnormal audio files whose answer data need to be updated; when all the audio files are played, obtain the updated question and answer pairs of each abnormal audio file; based on the updated question and answer pairs, update the knowledge base.

[0004] However, it takes a long time to play a recording file completely, and one annotator cannot listen to multiple recording files at the same time, which results in a longer annotation time for abnormal recording files, and further slows down the update speed of the knowledge base. Summary of the invention

[0005] The disclosed embodiments provide a method, device, electronic device and storage medium for updating the knowledge base of intelligent customer service, which can shorten the annotation time of abnormal recording files and improve the updating speed of the knowledge base. The technical solution is as follows:

[0006] In a first aspect, a method for updating a knowledge base of an intelligent customer service is provided, the method comprising:

[0007] Determine user satisfaction status of the multiple recording files according to user feedback data of the multiple recording files, where the recording files are files recorded during the interaction between the user and the intelligent customer service, and the user satisfaction status is used to reflect the user's satisfaction status with the answer data obtained by the intelligent customer service from the knowledge base;

[0008] Filtering at least one candidate recording file from the multiple recording files according to user satisfaction status of the multiple recording files;

[0009] Preprocessing the at least one candidate recording file to obtain at least one preprocessed file, wherein the preprocessed file is a text file or a candidate recording file that does not include invalid voice data;

[0010] Based on the at least one preprocessed file, obtaining at least one abnormal recording file whose answer data needs to be updated;

[0011] Based on the target question data in the at least one abnormal recording file, generating at least one updated question-answer pair, the updated question-answer pair including the target question data and updated answer data;

[0012] The knowledge base is updated based on the at least one updated question-answer pair.

[0013] In another embodiment of the present disclosure, determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes:

[0014] Obtain evaluation data of each recording file from user feedback data of each recording file;

[0015] When the evaluation index indicated by the evaluation data of any recording file exceeds the preset index range, it is determined that the user satisfaction status of the recording file is unsatisfactory.

[0016] In another embodiment of the present disclosure, determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes:

[0017] From the user feedback data of each recording file, query the complaint data of each recording file;

[0018] When complaint data of any recording file is queried, it is determined that the user satisfaction status of the recording file is unsatisfactory.

[0019] In another embodiment of the present disclosure, determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes:

[0020] When it is detected that the interaction mode of the user in any recording file is switched from the intelligent customer service to the manual customer service, it is determined that the user satisfaction status of the recording file is dissatisfied.

[0021] In another embodiment of the present disclosure, determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes:

[0022] Obtain human voice data from user feedback data of each recording file;

[0023] For any recording file, determine the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file;

[0024] When the interaction emotion is a specified emotion, it is determined that the user satisfaction status of the recording file is dissatisfied.

[0025] In another embodiment of the present disclosure, determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes:

[0026] Extracting target voice features from the voice data in the recording file;

[0027] The target speech feature is input into the emotion recognition model, and the interaction emotion between the user and the intelligent customer service during the recording of the recording file is output. The emotion recognition model is used to recognize the user's emotion based on the speech feature.

[0028] In another embodiment of the present disclosure, determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes:

[0029] Extracting a specified keyword from the human voice data in the recording file;

[0030] When the specified keyword is extracted, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the audio file is the specified emotion.

[0031] In another embodiment of the present disclosure, determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes:

[0032] Obtaining at least one of the number of repetitions of the question data and the number of voice call requests received from the user within a preset time period from the human voice data in the recording file;

[0033] When at least one of the number of repetitions of the question data and the number of voice call requests reaches a preset number, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is a specified emotion.

[0034] In another embodiment of the present disclosure, preprocessing the at least one candidate recording file to obtain at least one preprocessed file includes:

[0035] Each candidate recording file is converted from a voice form to a text form to obtain at least one pre-processed file.

[0036] In another embodiment of the present disclosure, preprocessing the at least one candidate recording file to obtain at least one preprocessed file includes:

[0037] Each candidate recording file is segmented to obtain multiple recording segments;

[0038] At least one of the blank data and the target answer data in the multiple recording segments is deleted to obtain a preprocessed file, where the target answer data is all or part of the answer data included in the candidate recording file.

[0039] In another embodiment of the present disclosure, updating the knowledge base based on the at least one updated question-answer pair includes:

[0040] Clustering the question data in the knowledge base using the target question data in the at least one updated question-answer pair as a cluster center to obtain at least one question data set;

[0041] The answer data corresponding to each question data in each question data set is updated to the updated answer data corresponding to the corresponding cluster center.

[0042] In another embodiment of the present disclosure, the method further includes:

[0043] When it is detected that the interaction mode with the user is changed from the intelligent customer service to the manual customer service, a manual question and answer pair provided by the manual customer service is obtained, wherein the manual question and answer pair includes question data that needs to be answered by the manual customer service and manual answer data provided;

[0044] Based on the manual question and answer pairs, the knowledge base is updated.

[0045] In another embodiment of the present disclosure, the method further includes:

[0046] Based on the preset screening conditions, at least one target recording file is screened out from the normal recording files;

[0047] Obtain verification results for answer data in each target recording file;

[0048] Based on the verification result, the knowledge base is updated.

[0049] In a second aspect, a knowledge base updating device for intelligent customer service is provided, the device comprising:

[0050] A determination module, configured to determine user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files, wherein the recording files are files recorded during the interaction between the user and the intelligent customer service, and the user satisfaction status is used to reflect the user's satisfaction status with the answer data obtained by the intelligent customer service from the knowledge base;

[0051] A screening module, configured to screen out at least one candidate recording file from the plurality of recording files according to user satisfaction status of the plurality of recording files;

[0052] A preprocessing module, used for preprocessing the at least one candidate recording file to obtain at least one preprocessed file, wherein the preprocessed file is a text file or a candidate recording file that does not include invalid voice data;

[0053] An acquisition module, configured to acquire, based on the at least one preprocessed file, at least one abnormal recording file whose answer data needs to be updated;

[0054] A generating module, configured to generate at least one updated question-answer pair based on the target question data in the at least one abnormal recording file, wherein the updated question-answer pair includes the target question data and updated answer data;

[0055] An updating module is used to update the knowledge base based on the at least one updated question-answer pair.

[0056] In another embodiment of the present disclosure, the determination module is used to obtain evaluation data of each recording file from user feedback data of each recording file; when the evaluation index indicated by the evaluation data of any recording file exceeds a preset index range, the user satisfaction status of the recording file is determined to be unsatisfactory.

[0057] In another embodiment of the present disclosure, the determination module is used to determine that the user satisfaction status of any recording file is dissatisfied when it is detected that the user interaction mode in any recording file is switched from the intelligent customer service to the manual customer service.

[0058] In another embodiment of the present disclosure, the determination module is used to query the complaint data of each recording file from the user feedback data of each recording file; when the complaint data of any recording file is queried, it is determined that the user satisfaction status of the recording file is dissatisfied.

[0059] In another embodiment of the present disclosure, the determination module is used to obtain vocal data from user feedback data of each recording file; for any recording file, the interaction emotion between the user and the intelligent customer service during the recording of the recording file is determined based on the vocal data of the recording file; when the interaction emotion is a specified emotion, the user satisfaction status of the recording file is determined to be dissatisfied.

[0060] In another embodiment of the present disclosure, the determination module is used to extract target voice features from the vocal data of the recording file; input the target voice features into an emotion recognition model, and output the interaction emotions between the user and the intelligent customer service during the recording of the recording file. The emotion recognition model is used to identify the user's emotions based on the voice features.

[0061] In another embodiment of the present disclosure, the determination module is used to extract specified keywords from the vocal data of the recording file; when the specified keywords are extracted, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is the specified emotion.

[0062] In another embodiment of the present disclosure, the determination module is used to obtain at least one of the number of repetitions of question data and the number of voice call requests received from the user within a preset time period from the vocal data of the recording file; when at least one of the number of repetitions of the question data and the number of voice call requests reaches a preset number, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is a specified emotion.

[0063] In another embodiment of the present disclosure, the preprocessing module is used to convert each candidate recording file from a voice form to a text form to obtain at least one preprocessed file.

[0064] In another embodiment of the present disclosure, the preprocessing module is used to segment each candidate recording file to obtain multiple recording segments; delete blank data in the multiple recording segments and at least one of the target answer data to obtain a preprocessed file, and the target answer data is all or part of the answer data contained in the candidate recording file.

[0065] In another embodiment of the present disclosure, the updating module is used to cluster the question data in the knowledge base with the target question data in the at least one updated question-answer pair as the cluster center to obtain at least one question data set; and update the answer data corresponding to each question data in each question data set to the updated answer data corresponding to the corresponding cluster center.

[0066] In another embodiment of the present disclosure, the acquisition module is further used to, when detecting that the interaction mode with the user is changed from the intelligent customer service to the manual customer service, acquire a manual question and answer pair provided by the manual customer service, wherein the manual question and answer pair includes question data that needs to be answered by the manual customer service and manual answer data provided;

[0067] The updating module is further used to update the knowledge base based on the manual question and answer pairs.

[0068] In another embodiment of the present disclosure, the screening module is further used to screen out at least one target recording file from the normal recording files based on a preset screening condition;

[0069] The acquisition module is further used to obtain the verification result of the answer data in each target recording file;

[0070] The updating module is further used to update the knowledge base based on the verification result.

[0071] In a third aspect, an electronic device is provided, comprising a processor and a memory, wherein the memory stores at least one program code, and the at least one program code is loaded and executed by the processor to implement the knowledge base update method of the intelligent customer service as described in the first aspect.

[0072] In a fourth aspect, a computer-readable storage medium is provided, wherein at least one program code is stored in the storage medium, and the at least one program code is loaded and executed by a processor to implement the knowledge base updating method of the intelligent customer service as described in the first aspect.

[0073] The technical solution provided by the embodiments of the present disclosure has the following beneficial effects:

[0074] There is no need to play the recording file in its entirety. Based on the user's satisfaction with the answer data provided by the intelligent customer service, candidate recording files are screened. On this basis, the candidate recording files are pre-processed and converted into text form or invalid voice data in the candidate recording files are removed, so that labeling personnel can quickly mark abnormal recording files, greatly shortening the labeling time of abnormal recording files and improving the updating speed of the knowledge base. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0076] Figure 1 It is a flow chart of a knowledge base updating method of an intelligent customer service provided by an embodiment of the present disclosure;

[0077] Figure 2 is a flow chart of another method for updating the knowledge base of intelligent customer service provided by an embodiment of the present disclosure;

[0078] Figure 3 It is a structural diagram of another knowledge base updating device for intelligent customer service provided by an embodiment of the present disclosure;

[0079] Figure 4 A structural block diagram of an electronic device provided by an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0080] In order to make the objectives, technical solutions and advantages of the present disclosure more clear, the embodiments of the present disclosure will be further described in detail below with reference to the accompanying drawings.

[0081] It can be understood that the terms "each", "multiple", and "any" used in the embodiments of the present disclosure include two or more, each refers to each of the corresponding multiple, and any refers to any one of the corresponding multiple. For example, the multiple words include 10 words, and each word refers to each word of the 10 words, and any word refers to any one of the 10 words.

[0082] With the development of artificial intelligence technology, intelligent customer service has been widely developed, especially in the fields of after-sales service, customer screening, complaint consultation, product consultation, etc. in logistics, hotels, finance, education, insurance, banking, health, government and other industries. Intelligent customer service refers to an industry-oriented application developed on the basis of large-scale knowledge processing. It is suitable for technical industries such as large-scale knowledge processing, natural language understanding, knowledge management, automatic question and answer systems, reasoning, etc. Intelligent customer service not only provides enterprises with fine-grained knowledge management, but also establishes a fast and effective technical means based on natural language between enterprises and massive users. At the same time, it can also provide enterprises with the statistical analysis information required for refined management. In order to better serve users, the knowledge base of intelligent customer service needs to be updated in a timely manner. The disclosed embodiment provides a knowledge base update method for intelligent customer service, see Figure 1 The method flow provided in the embodiment of the present disclosure includes:

[0083] 101. Determine user satisfaction status of the multiple recording files based on user feedback data of the multiple recording files.

[0084] Among them, the recording file is a file recorded during the interaction between the user and the intelligent customer service, and the user satisfaction status is used to reflect the user's satisfaction with the answer data provided by the intelligent customer service.

[0085] 102. According to user satisfaction status of the multiple recording files, at least one candidate recording file is selected from the multiple recording files.

[0086] 103. Preprocess at least one candidate recording file to obtain at least one preprocessed file.

[0087] The preprocessed file is a text file or a candidate recording file that does not include invalid voice data.

[0088] 104. Based on at least one pre-processed file, obtain at least one abnormal recording file that needs to be updated for answer data.

[0089] 105. Generate at least one updated question-answer pair based on target question data in at least one abnormal recording file.

[0090] The updated question-answer pair includes the target question data and the updated answer data.

[0091] 106. Based on at least one updated question-answer pair, update the knowledge base.

[0092] The method provided by the embodiment of the present disclosure does not need to play the recording file completely. Based on the user's satisfaction with the answer data provided by the intelligent customer service, the candidate recording files are screened. On this basis, the candidate recording files are pre-processed to convert the candidate recording files into text form or invalid voice data in the candidate recording files are removed, so that the labeling personnel can quickly label abnormal recording files, which greatly shortens the labeling time of abnormal recording files and improves the updating speed of the knowledge base.

[0093] In another embodiment of the present disclosure, determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes:

[0094] Obtain evaluation data of each recording file from user feedback data of each recording file;

[0095] When the evaluation index indicated by the evaluation data of any recording file exceeds the preset index range, it is determined that the user satisfaction status of the recording file is unsatisfactory.

[0096] In another embodiment of the present disclosure, determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes:

[0097] From the user feedback data of each recording file, query the complaint data of each recording file;

[0098] When complaint data of any recording file is queried, it is determined that the user satisfaction status of the recording file is unsatisfactory.

[0099] In another embodiment of the present disclosure, determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes:

[0100] When it is detected that the user interaction mode in any recording file is switched from intelligent customer service to manual customer service, it is determined that the user satisfaction status of the recording file is dissatisfied.

[0101] In another embodiment of the present disclosure, determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes:

[0102] Obtain human voice data from user feedback data of each recording file;

[0103] For any recording file, determine the interaction emotions between the user and the intelligent customer service during the recording process based on the voice data of the recording file;

[0104] When the interaction emotion is the specified emotion, the user satisfaction status of the recording file is determined to be dissatisfied.

[0105] In another embodiment of the present disclosure, determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes:

[0106] Extract target speech features from the human voice data in the recording file;

[0107] The target speech features are input into the emotion recognition model, and the interactive emotions between the user and the intelligent customer service during the recording process are output. The emotion recognition model is used to identify the user's emotions based on the speech features.

[0108] In another embodiment of the present disclosure, determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes:

[0109] Extract specified keywords from the human voice data in the recording file;

[0110] When the specified keyword is extracted, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the audio file is the specified emotion.

[0111] In another embodiment of the present disclosure, determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes:

[0112] Obtaining at least one of the number of repetitions of the question data and the number of voice call requests received from the user within a preset time period from the human voice data in the recording file;

[0113] When at least one of the number of repetitions of the question data and the number of voice call requests reaches a preset number, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is a specified emotion.

[0114] In another embodiment of the present disclosure, preprocessing at least one candidate recording file to obtain at least one preprocessed file includes:

[0115] Each candidate recording file is converted from a voice form to a text form to obtain at least one pre-processed file.

[0116] In another embodiment of the present disclosure, preprocessing at least one candidate recording file to obtain at least one preprocessed file includes:

[0117] Each candidate recording file is segmented to obtain multiple recording segments;

[0118] Delete blank data in multiple recording segments and at least one of the target answer data to obtain a preprocessed file, where the target answer data is all or part of the answer data included in the candidate recording file.

[0119] In another embodiment of the present disclosure, updating the knowledge base based on at least one updated question-answer pair includes:

[0120] Taking the target question data in at least one updated question-answer pair as the cluster center, clustering the question data in the knowledge base to obtain at least one question data set;

[0121] The answer data corresponding to each question data in each question data set is updated to the updated answer data corresponding to the corresponding cluster center.

[0122] In another embodiment of the present disclosure, the method further comprises:

[0123] When it is detected that the interaction mode with the user changes from intelligent customer service to manual customer service, a manual question and answer pair provided by the manual customer service is obtained, where the manual question and answer pair includes question data that the manual customer service needs to answer and manual answer data provided;

[0124] Based on manual question and answer pairs, the knowledge base is updated.

[0125] In another embodiment of the present disclosure, the method further comprises:

[0126] Based on the preset screening conditions, at least one target recording file is screened out from the normal recording files;

[0127] Obtain verification results for answer data in each target recording file;

[0128] Based on the verification results, the knowledge base is updated.

[0129] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present disclosure, and will not be described in detail here.

[0130] The disclosed embodiment provides a method for updating the knowledge base of an intelligent customer service. Taking an electronic device executing the disclosed embodiment as an example, the electronic device may be a terminal with certain computing capabilities, such as a smart phone, a tablet computer, a laptop computer, a desktop computer, etc. The electronic device may also be a server, which may be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers. Figure 2 , the method flow provided by the embodiment of the present disclosure includes:

[0131] 201. The electronic device determines user satisfaction status of the plurality of recording files based on user feedback data of the plurality of recording files.

[0132] Among them, the recording file is a file recorded during the interaction between the user and the intelligent customer service, and the recording file includes the user's question data, the intelligent customer service's answer data and other interaction data between the user and the intelligent customer service. User feedback data includes evaluation data, complaint data, etc., which can be generated during the interaction between the user and the intelligent customer service, or after the user interacts with the intelligent customer service. For example, after the user interacts with the intelligent customer service, the electronic device sends a request message for evaluating the interaction in the form of email, short message, instant message, etc., and the user sends the user feedback data to the electronic device in a corresponding form according to the real feeling during the interaction. The evaluation data is the evaluation given by the user for the answer data provided by the intelligent customer service after the interaction with the intelligent customer service is completed. The evaluation data includes evaluation scores, evaluation levels, etc. If a 10-point system is adopted, the evaluation scores can be 8 points, 9 points, 10 points, etc. If a 100-point system is adopted, the evaluation scores can be 80 points, 90 points, 100 points, etc. The evaluation levels include multiple levels such as very satisfied, satisfied, general, poor, and poor. Complaint data is data generated when users complain to the backend when they are dissatisfied with the answer data provided by the intelligent customer service. The complaint data may include question data, answer data given by the intelligent customer service, reasons for dissatisfaction, etc. User satisfaction status is used to reflect the user's satisfaction with the answer data provided by the intelligent customer service. The user satisfaction status includes satisfaction and dissatisfaction.

[0133] In another embodiment of the present disclosure, when the electronic device determines the user satisfaction status of the plurality of recording files according to the user feedback data of the plurality of recording files, the following method may be adopted:

[0134] The electronic device obtains evaluation data of each recording file from user feedback data of each recording file.

[0135] Considering that if the evaluation score or evaluation level in the user feedback data is high, it is likely that the answer data provided by the intelligent customer service can solve the user's problem, and the user's satisfaction with the question given by the intelligent customer service is high. On the contrary, if the evaluation score or evaluation level in the user feedback data is low, it is likely that the answer data provided by the intelligent customer service cannot solve the user's problem, and the user's satisfaction with the question given by the intelligent customer service is low. In view of this, the electronic device can obtain the evaluation number from the user feedback data of each recording file, and then judge the user's satisfaction with the answer data provided by the intelligent customer service in the recording file based on the evaluation data.

[0136] When the evaluation index indicated by the evaluation data of any recording file exceeds the preset index range, the electronic device determines that the user satisfaction status of the recording file is unsatisfactory.

[0137] When the evaluation data of the recording file is an evaluation score, the preset indicator range can be a preset score range. Taking the 100 system as an example, the preset score range can be 80 to 85 points, 90 to 95 points, and so on. That is, when the evaluation score of the recording file exceeds the preset score range, the electronic device determines that the user satisfaction of the recording file is unsatisfactory. When the evaluation data of the recording file is an evaluation grade, the preset indicator range can be a preset grade range. That is, when the evaluation grade of the recording file exceeds the preset grade range, the electronic device determines that the user satisfaction of the recording file is unsatisfactory. The method provided in the embodiment of the present disclosure performs a preliminary screening of the recording files based on the evaluation data of the recording files, screens out the recording files with lower evaluation scores or evaluation grades, and then further verifies these recording files, thereby reducing the annotation cost and annotation time of abnormal recording files.

[0138] In another embodiment of the present disclosure, when the electronic device determines the user satisfaction status of the plurality of recording files according to the user feedback data of the plurality of recording files, the following method may also be used:

[0139] The electronic device queries the complaint data of each recording file from the user feedback data of each recording file.

[0140] Usually, after the intelligent customer service provides the answer data to the user, if the user is not satisfied with the answer data provided by the intelligent customer service, the user may complain to the intelligent customer service, but if the user is relatively satisfied with the answer data provided by the intelligent customer service, the user is unlikely to complain to the intelligent customer service. In view of this, the electronic device can judge the user satisfaction status of the recording file by querying whether the user feedback data of the recording file includes complaint data.

[0141] When the complaint data of any recording file is queried, the electronic device determines that the user satisfaction status of the recording file is unsatisfactory.

[0142] The electronic device queries the user feedback data of each recording file. When complaint data is queried from the user feedback data of any recording file, the electronic device determines that the user satisfaction status of the recording file is unsatisfactory.

[0143] Taking into account that users may not always complain after obtaining unsatisfactory answer data, when the electronic device does not find complaint data from the user feedback data of the recording file, the electronic device will not determine that the user satisfaction status of the recording file is satisfied, but will further judge in combination with the evaluation data or other data, so as to more accurately determine whether the user is satisfied with the answer data.

[0144] In another embodiment of the present disclosure, when the user is not satisfied with the answer data given by the intelligent customer service, the user is usually transferred to the manual customer service, and the manual customer service solves the problem. Therefore, when the electronic device determines the user satisfaction status of the recording file, it can detect the user's interaction mode in the recording text. When it is detected that the user's interaction mode in the recording file is transferred from the intelligent customer service to the manual customer service, it can be determined that the user satisfaction status of the recording file is dissatisfied.

[0145] In another embodiment of the present disclosure, when the electronic device determines the user satisfaction status of the plurality of recording files according to the user feedback data of the plurality of recording files, the following method may also be used:

[0146] The electronic device obtains human voice data from user feedback data of each recording file.

[0147] During the interaction between the intelligent customer service and the user, the user usually gives certain voice feedback for each answer data given by the intelligent customer service. The electronic device records the first time period from when the user starts speaking to when the intelligent customer service starts speaking, and records the second time period from when the intelligent customer service starts speaking to when the user starts speaking. The electronic device divides the recording file according to the time indicated by the first time period to obtain the user's vocal data, and the electronic device divides the recording file according to the time indicated by the second time period to obtain the voice data of the intelligent customer service.

[0148] For any recording file, the electronic device determines the interaction emotions between the user and the intelligent customer service during the recording process of the recording file based on the human voice data of the recording file.

[0149] In a possible implementation, when the electronic device determines the interaction emotions between the user and the intelligent customer service during the recording of the recording file based on the human voice data of the recording file, the following method can be used:

[0150] The electronic device extracts target voice features from the human voice data in the recording file.

[0151] Among them, the target voice features include at least one of volume features and speech speed features. Usually when users are angry or do not get satisfactory answers, they will unconsciously increase the volume, speed up the speech, etc. to express their impatience. In view of this, the electronic device can judge the interactive emotions between the user and the intelligent customer service by extracting the target voice features from the human voice data of the recording file. Specifically, when the electronic device extracts the target voice features from the human voice data of the recording file, it can first segment the human voice data to obtain multiple human voice segments, and then extract the target voice features from each human voice segment. When the electronic device segments the human voice data, it takes the starting position of the human voice data as the starting point and segments it as the time window with a preset duration to obtain multiple human voice segments with a preset duration. Among them, the preset duration can be 1 second, 2 seconds, etc.

[0152] The electronic device inputs the target speech features into the emotion recognition model and outputs the interactive emotions between the user and the intelligent customer service during the recording process of the audio file.

[0153] Among them, the emotion recognition model is used to identify the user's emotions based on voice features. The electronic device inputs the extracted target voice features into the emotion recognition model in sequence, and recognizes based on the emotion recognition model to output the interaction emotions between the user and the intelligent customer service. If the input multiple target voice features show that the volume gradually increases with the increase of input time, or the speaking speed becomes faster and faster, the emotion recognition model outputs the interaction emotions between the user and the intelligent customer service as the specified emotions. The specified emotions include negative emotions such as anger, dissatisfaction, and impatience.

[0154] Among them, the emotion recognition model can be trained based on human voice sample data with different emotions. For example, the electronic device obtains human voice sample data labeled with different emotions, and extracts the target voice features of each human voice sample data, and then inputs the target voice features of each human voice sample data into the initial emotion recognition model. The initial emotion recognition model can be a deep learning model, and the model parameters of the initial emotion recognition model are set with initial values. Based on the initial emotion recognition model, the user's emotions in each human voice sample data are recognized, and the recognized emotions corresponding to each human voice sample data are output. Next, the electronic device inputs the recognized emotions and labeled emotions corresponding to each human voice sample data into a pre-constructed target loss function, and outputs the target loss function value of each human voice sample. If the target loss function value does not meet the preset threshold, the parameter values ​​of the model parameters of the initial emotion recognition model are continuously adjusted until the obtained target loss function value meets the preset threshold, and the initial emotion recognition model corresponding to the parameter values ​​of the model parameters when the preset threshold is met is determined as the trained emotion recognition model.

[0155] In another possible implementation, when the electronic device determines the interaction emotions between the user and the intelligent customer service during the recording of the recording file based on the human voice data of the recording file, the following method can be used:

[0156] The electronic device extracts the specified keywords from the human voice data in the recording file.

[0157] The designated keywords include interjections used to express the user's impatience, such as "cough", "oops", etc., and words that directly express the user's emotions, such as "hate", "angry", "dislike", "bored", etc., and also include the user's uncivilized language, and may also include some negative sentences, such as "Can you answer my question", "Can you solve it", etc.

[0158] The electronic device inputs the human voice data of the recording file into the keyword recognition model for recognition, and outputs the specified keyword. The keyword recognition model is used to recognize the specified keyword from the human voice data. The recognition process of the keyword recognition model is: converting the human voice data into text data, and then traversing the text data, and when the text data includes the specified keyword, the specified keyword is output. When the electronic device recognizes the specified keyword from the human voice data based on the keyword recognition model, it can first divide the human voice data into multiple human voice segments, and then input the multiple human voice segments into the keyword recognition model in turn, and output the specified keyword. By dividing the human voice data, the recognition speed of the specified keyword is improved.

[0159] When the specified keyword is extracted, the electronic device determines that the interaction emotion between the user and the intelligent customer service during the recording of the audio file is the specified emotion.

[0160] When the user is angry or gets a satisfactory answer, he or she may say some words that express impatience. The electronic device extracts the specified keywords from the human voice data in the recording file. If the specified keywords can be extracted, the electronic device can determine that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is the specified emotion. If the specified keywords are not extracted, the electronic device needs to further judge in combination with the evaluation data or other data to more accurately determine whether the user is satisfied with the question data.

[0161] In a possible implementation, when the electronic device determines the interaction emotions between the user and the intelligent customer service during the recording of the recording file based on the human voice data of the recording file, the following method can be used:

[0162] The electronic device obtains at least one of the number of repetitions of the question data and the number of voice call requests received from the user within a preset time period from the human voice data in the recording file.

[0163] When the answer obtained by the user is not the answer he wants, the user usually repeats his question to prompt the intelligent customer service to give a corresponding answer to his question, or the user continuously sends voice call requests to the intelligent customer service within a preset time period. In view of this, the electronic device can determine the interaction emotions between the user and the intelligent customer service during the recording of the recording file by obtaining at least one of the number of repetitions of the question data or the number of voice call requests received from the user within a preset time period. Among them, the preset time period can be 3 minutes, 5 minutes, etc.

[0164] When at least one of the number of repetitions of the question data and the number of voice call requests reaches a preset number, the electronic device determines that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is a specified emotion.

[0165] The preset number of times may be 2 times, 3 times, etc., and the present disclosure embodiment does not specifically limit the preset number of times. When the number of repetitions of the question data reaches the preset number of times, or the number of voice call requests reaches the preset number of times, or the number of repetitions of the question data and the number of voice call requests both reach the preset number of times, the electronic device determines that the interaction request between the user and the intelligent customer service during the recording of the recording file is a specified emotion.

[0166] When the interaction emotion is the specified emotion, the electronic device determines that the user satisfaction status of the recording file is dissatisfied.

[0167] When the interaction emotion between the user and the intelligent customer service is a specified emotion, it means that the interaction process is not pleasant and the user's problem is most likely not solved. At this time, the electronic device can determine that the user satisfaction status of the recording file is dissatisfied.

[0168] It should be noted that the above is described by taking the example of determining the user's satisfaction with the recording file according to any one of the evaluation data, complaint data, the interaction mode with the user being switched from intelligent customer service to manual customer service, and the interaction emotions between the user and the intelligent customer service. In practical applications, the electronic device can also first query whether the user feedback data of the recording file includes complaint data. If the user feedback data of the recording file includes complaint data, the electronic device directly determines that the user is dissatisfied with the recording file. If the user feedback data of the recording file does not include complaint data, the electronic device obtains evaluation data. If the evaluation data can be obtained, but the evaluation index indicated by the evaluation data is not within the preset index range, it is determined that the user is dissatisfied with the recording file; if the evaluation data is not obtained, the electronic device queries whether the interaction mode with the user in the recording file is switched from intelligent customer service to manual customer service. If the interaction mode with the user is switched from intelligent customer service to manual customer service, it is determined that the user is dissatisfied with the recording file; if the interaction mode with the user is not queried to be switched from intelligent customer service to manual customer service, the electronic device extracts human voice data from the recording file, and based on the human voice data, determines the interaction emotions between the user and the intelligent customer service during the recording of the recording file. When the interaction emotion is a specified emotion, it is determined that the recording file is dissatisfied.

[0169] 202. The electronic device selects at least one candidate recording file from the multiple recording files according to user satisfaction of the multiple recording files.

[0170] When the user satisfaction status of the recording file is unsatisfactory, in order to better serve the user, the electronic device can select the recording file as a candidate recording file, thereby specifically checking the candidate recording file to improve the user's service satisfaction.

[0171] 203. The electronic device pre-processes at least one candidate recording file to obtain at least one pre-processed file.

[0172] In a possible implementation, the electronic device can convert each candidate recording file from voice form to text form to obtain at least one pre-processed file. Due to the characteristics of human beings, reading speed is faster than listening speed. By converting the candidate recording files into text form, the electronic device enables the annotator to quickly mark the abnormal recording files whose answer data needs to be updated, thereby improving the annotator's annotation speed for the candidate recording files.

[0173] In another possible implementation, the electronic device may process each candidate recording file in segments to obtain multiple recording segments, and then delete the blank data in the multiple recording segments and at least one of the target answer data to obtain a preprocessed file, where the target answer data is all or part of the answer data contained in the candidate recording file. Since the questions answered by the intelligent customer service have fixed templates, the annotators do not need to listen to the entire answer data in the actual analysis process to know the content. In order to increase the annotation speed of the annotators, the electronic device can delete part of the question voice data. Taking into account the network delay, it takes a certain amount of query time for the intelligent customer service to query the answer data, and it also takes a certain amount of time for the user to give feedback on the answer data given by the intelligent customer service, resulting in some blank voice data included in the recorded recording file. These data will extend the annotation time of the annotators. In order to increase the annotation speed of the annotators, the electronic device can delete these blank voice data.

[0174] Furthermore, after the electronic device deletes blank data in multiple recording segments and at least one item of target answer data, it can also add subtitles to the remaining audio segments, so that when the preprocessed file is played, the subtitles corresponding to the preprocessed file can be synchronously displayed to improve the annotation speed of the annotator.

[0175] 204. The electronic device obtains at least one abnormal recording file whose answer data needs to be updated based on at least one preprocessed file.

[0176] The electronic device provides at least one preprocessed file to a labeler for review. When the labeler finds that the answer data of any preprocessed file is not accurate enough, if the preprocessed file is a text file, the labeler operates the electronic device to treat the candidate recording file corresponding to the preprocessed file as an abnormal recording file; if the preprocessed file is a candidate recording file that does not include invalid voice data, the labeler operates the electronic device to treat the preprocessed file as an abnormal recording file.

[0177] 205. Based on the target question data in at least one abnormal recording file, the electronic device generates at least one updated question-answer pair.

[0178] For each abnormal recording file, the electronic device obtains target question data from the abnormal recording file, and then obtains updated answer data corresponding to the target question data by querying information or from other technical personnel, thereby forming an updated question-answer pair with the target question data and the updated answer data.

[0179] 206. The electronic device updates the knowledge base based on at least one updated question-answer pair.

[0180] When at least one updated question-answer pair is obtained, the electronic device may directly store the at least one updated question-answer pair, thereby updating the knowledge base.

[0181] In another implementation of the present disclosure, the electronic device may also cluster the question data in the knowledge base using the target question data in at least one updated question-answer pair as a cluster center to obtain at least one question data set, each question data set including at least two question data, and then update the answer data corresponding to each question data in each question data set to the updated answer data corresponding to the corresponding cluster center, thereby realizing batch update of data in the knowledge base.

[0182] Furthermore, to avoid storing a large amount of redundant data in the electronic device, in response to the performance of the server, the electronic device queries the redundant data in the knowledge base and deletes the queried redundant data from the knowledge base. When querying the redundant data, the electronic device can compare the question data of one question-answer pair with the question data of another question-answer pair. If the question data of the two question-answer pairs are the same, the question-answer pair with the later storage time is retained, and the question-answer pair with the earlier storage time is deleted.

[0183] 207. Based on preset filtering conditions, the electronic device filters out at least one target recording file from normal recording files.

[0184] The preset screening conditions include that the average speech speed exceeds the preset speed range, etc. The preset speed range is a range greater than the first speed and less than the second speed. The first speed may be 20 words / minute, 30 words / minute, etc., and the second speed may be 40 words / minute, 50 words / minute, etc. Considering the large number of normal recording files, in order to improve the verification speed of normal recording files, the electronic device may randomly obtain at least one normal recording file, or extract at least one normal recording file according to a preset ratio, and then based on the preset screening conditions, select at least one target voice file from the obtained normal recording files, and then provide it to the annotation personnel for manual verification to ensure that the data in the knowledge base are accurate data, thereby improving the service quality of the intelligent customer service.

[0185] 208. The electronic device obtains the verification result of the answer data in each target recording file.

[0186] The annotator verifies each target recording file. For any target recording file, if the answer data of the target recording file is inaccurate, the annotator obtains updated answer data corresponding to the question data of the target recording file by querying information or from other technical personnel, and inputs the updated answer data into the electronic device, thereby triggering the electronic device to generate a verification result that needs to be updated based on the question data and the updated answer data in the target recording file; if the answer data of the target recording file is accurate, the annotator will trigger the electronic device to generate a verification result that does not need to be updated.

[0187] 209. Based on the verification result, the electronic device updates the knowledge base.

[0188] When the verification result of the target recording file is a verification result that needs to be updated, the electronic device updates the knowledge base based on the updated answer data and question data; when the verification result of the target recording file is a verification result that does not need to be updated, the electronic device will not update the knowledge base.

[0189] In another embodiment of the present disclosure, when it is detected that the mode of interaction with the user is changed from intelligent customer service to manual customer service, the electronic device obtains a manual question and answer pair provided by the manual customer service, which includes question data that the manual customer service needs to answer and manual answer data provided, etc., and then based on the manual question and answer pair, the knowledge base is updated to improve the accuracy of the data in the knowledge base to better serve the user.

[0190] It should be noted that the above example uses voice interaction as an example of the interaction between the intelligent customer and the user. Of course, the intelligent customer service and the user can also interact in text form. When text form is used for interaction, when updating the knowledge base, the above method can also be used to judge the user satisfaction with the text file, and then based on the user satisfaction, select candidate text files from multiple text files, and then pre-process the candidate text files, delete the template answer data, and then mark the abnormal text files from the pre-processed files, so as to optimize the knowledge base by updating the answer data in the non-text file.

[0191] The method provided by the embodiment of the present disclosure does not need to play the recording file completely. Based on the user's satisfaction with the answer data provided by the intelligent customer service, the candidate recording files are screened. On this basis, the candidate recording files are pre-processed to convert the candidate recording files into text form or invalid voice data in the candidate recording files are removed, so that the labeling personnel can quickly label abnormal recording files, which greatly shortens the labeling time of abnormal recording files and improves the updating speed of the knowledge base.

[0192] See also Figure 3The present disclosure provides a knowledge base updating device for intelligent customer service, the device comprising:

[0193] The determination module 301 is used to determine the user satisfaction status of the multiple recording files according to the user feedback data of the multiple recording files, where the recording files are files recorded during the interaction between the user and the intelligent customer service, and the user satisfaction status is used to reflect the user's satisfaction with the answer data obtained by the intelligent customer service from the knowledge base;

[0194] A screening module 302, configured to screen out at least one candidate recording file from the plurality of recording files according to user satisfaction status of the plurality of recording files;

[0195] A preprocessing module 303 is used to preprocess at least one candidate recording file to obtain at least one preprocessed file, where the preprocessed file is a text file or a candidate recording file that does not include invalid voice data;

[0196] An acquisition module 304, configured to acquire, based on at least one preprocessed file, at least one abnormal recording file whose answer data needs to be updated;

[0197] A generating module 305, configured to generate at least one updated question-answer pair based on the target question data in the at least one abnormal recording file, the updated question-answer pair including the target question data and the updated answer data;

[0198] The updating module 306 is used to update the knowledge base based on at least one updated question-answer pair.

[0199] In another embodiment of the present disclosure, the determination module 301 is used to obtain evaluation data of each recording file from user feedback data of each recording file; when the evaluation index indicated by the evaluation data of any recording file exceeds a preset index range, the user satisfaction status of the recording file is determined to be unsatisfactory.

[0200] In another embodiment of the present disclosure, the determination module 301 is used to query the complaint data of each recording file from the user feedback data of each recording file; when the complaint data of any recording file is queried, it is determined that the user satisfaction status of the recording file is dissatisfied.

[0201] In another embodiment of the present disclosure, the determination module 301 is used to determine that the user satisfaction status of any recording file is dissatisfied when it is detected that the user interaction mode in any recording file is switched from intelligent customer service to manual customer service.

[0202] In another embodiment of the present disclosure, the determination module 301 is used to obtain vocal data from the user feedback data of each recording file; for any recording file, the interaction emotion between the user and the intelligent customer service during the recording of the recording file is determined based on the vocal data of the recording file; when the interaction emotion is a specified emotion, the user satisfaction status of the recording file is determined to be dissatisfied.

[0203] In another embodiment of the present disclosure, the determination module 301 is used to extract target voice features from the vocal data of the recording file; input the target voice features into the emotion recognition model, and output the interactive emotions between the user and the intelligent customer service during the recording of the recording file. The emotion recognition model is used to identify the user's emotions based on the voice features.

[0204] In another embodiment of the present disclosure, the determination module 301 is used to extract specified keywords from the human voice data of the recording file; when the specified keywords are extracted, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is the specified emotion.

[0205] In another embodiment of the present disclosure, the determination module 301 is used to obtain at least one of the number of repetitions of question data and the number of voice call requests received from the user within a preset time period from the vocal data of the recording file; when at least one of the number of repetitions of the question data and the number of voice call requests reaches the preset number, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is a specified emotion.

[0206] In another embodiment of the present disclosure, the preprocessing module 303 is used to convert each candidate recording file from a voice form to a text form to obtain at least one preprocessed file.

[0207] In another embodiment of the present disclosure, the preprocessing module 303 is used to segment each candidate recording file to obtain multiple recording segments; delete blank data in the multiple recording segments and at least one item of target answer data to obtain a preprocessed file, and the target answer data is all or part of the answer data contained in the candidate recording file.

[0208] In another embodiment of the present disclosure, the update module 306 is used to cluster the question data in the knowledge base with the target question data in at least one updated question-answer pair as the cluster center to obtain at least one question data set; and update the answer data corresponding to each question data in each question data set to the updated answer data corresponding to the corresponding cluster center.

[0209] In another embodiment of the present disclosure, the acquisition module 304 is further configured to, when detecting that the interaction mode with the user is changed from the intelligent customer service to the manual customer service, acquire a manual question and answer pair provided by the manual customer service, the manual question and answer pair including question data that the manual customer service needs to answer and manual answer data provided;

[0210] The updating module 306 is also used to update the knowledge base based on the manual question and answer pairs.

[0211] In another embodiment of the present disclosure,

[0212] The screening module 302 is further used to screen out at least one target recording file from the normal recording files based on a preset screening condition;

[0213] The acquisition module 304 is also used to obtain the verification result of the answer data in each target recording file;

[0214] The updating module 306 is further used to update the knowledge base based on the verification result.

[0215] To sum up, the device provided by the embodiment of the present disclosure does not need to play the recording file completely. It screens candidate recording files based on the user's satisfaction with the answer data provided by the intelligent customer service, and on this basis, pre-processes the candidate recording files to convert the candidate recording files into text form or removes invalid voice data in the candidate recording files, so that the labeling personnel can quickly label abnormal recording files, greatly shortening the labeling time of abnormal recording files and improving the updating speed of the knowledge base.

[0216] Figure 4 The structure block diagram of an electronic device 400 provided by an exemplary embodiment of the present disclosure is shown. Generally, the device 400 includes: a processor 401 and a memory 402 .

[0217] The processor 401 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 401 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 401 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 401 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 401 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0218] The memory 402 may include one or more computer-readable storage media, which may be non-transitory. The memory 402 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 402 is used to store at least one instruction, which is used to be executed by the processor 401 to implement the knowledge base update method of the intelligent customer service provided in the method embodiment of the present disclosure.

[0219] In some embodiments, the electronic device 400 may further optionally include: a peripheral device interface 403 and at least one peripheral device. The processor 401, the memory 402 and the peripheral device interface 403 may be connected via a bus or a signal line. Each peripheral device may be connected to the peripheral device interface 403 via a bus, a signal line or a circuit board. Specifically, the peripheral device includes: a power supply 404.

[0220] The peripheral device interface 403 may be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 401 and the memory 402. In some embodiments, the processor 401, the memory 402, and the peripheral device interface 403 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 401, the memory 402, and the peripheral device interface 403 may be implemented on a separate chip or circuit board, which is not limited in this embodiment.

[0221] The power supply 404 is used to power various components in the electronic device 400. The power supply 404 can be an alternating current, a direct current, a disposable battery, or a rechargeable battery. When the power supply 404 includes a rechargeable battery, the rechargeable battery can be a wired rechargeable battery or a wireless rechargeable battery. A wired rechargeable battery is a battery that is charged through a wired line, and a wireless rechargeable battery is a battery that is charged through a wireless coil. The rechargeable battery can also be used to support fast charging technology.

[0222] Those skilled in the art will understand that Figure 4 The structure shown in the figure does not constitute a limitation on the electronic device 400, and may include more or less components than those shown in the figure, or combine some components, or adopt a different component arrangement.

[0223] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory including instructions, and the above instructions can be executed by a processor of the electronic device 400 to complete the above-mentioned intelligent customer service knowledge base update method. Optionally, the storage medium can be a non-transitory computer-readable storage medium, for example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0224] The electronic device provided by the embodiment of the present disclosure does not need to play the recording file completely. It screens candidate recording files based on the user's satisfaction with the answer data provided by the intelligent customer service, and on this basis, pre-processes the candidate recording files to convert the candidate recording files into text form or removes invalid voice data in the candidate recording files, so that the labeling personnel can quickly label abnormal recording files, which greatly shortens the labeling time of abnormal recording files and improves the updating speed of the knowledge base.

[0225] An embodiment of the present disclosure provides a computer-readable storage medium, in which at least one program code is stored. The at least one program code is loaded and executed by a processor to implement a knowledge base updating method for intelligent customer service.

[0226] The computer-readable storage medium provided in the embodiment of the present disclosure does not need to play the recording file in its entirety. It screens candidate recording files based on the user's satisfaction with the answer data provided by the intelligent customer service. On this basis, it pre-processes the candidate recording files to convert the candidate recording files into text form or removes invalid voice data in the candidate recording files, so that labeling personnel can quickly label abnormal recording files, greatly shortening the labeling time of abnormal recording files and improving the updating speed of the knowledge base.

[0227] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0228] The above description is only an optional embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A knowledge base updating method for intelligent customer service, characterized in that: The method comprises: Determine user satisfaction status of the multiple recording files according to user feedback data of the multiple recording files, where the recording files are files recorded during the interaction between the user and the intelligent customer service, and the user satisfaction status is used to reflect the user's satisfaction status with the answer data obtained by the intelligent customer service from the knowledge base; Filtering at least one candidate recording file from the multiple recording files according to user satisfaction status of the multiple recording files; Preprocessing the at least one candidate recording file to obtain at least one preprocessed file, wherein the preprocessed file is a text file or a candidate recording file that does not include invalid voice data; Based on the at least one preprocessed file, obtaining at least one abnormal recording file whose answer data needs to be updated; Based on the target question data in the at least one abnormal recording file, generating at least one updated question-answer pair, the updated question-answer pair including the target question data and updated answer data; Taking the target question data in the at least one updated question-answer pair as a cluster center, clustering the question data in the knowledge base to obtain at least one question data set, each question data set including at least two question data; The answer data corresponding to each question data in each question data set is updated to the updated answer data corresponding to the corresponding cluster center.

2. The method according to claim 1, characterized in that Determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes: Obtain evaluation data of each recording file from user feedback data of each recording file; When the evaluation index indicated by the evaluation data of any recording file exceeds the preset index range, it is determined that the user satisfaction status of the recording file is unsatisfactory.

3. The method according to claim 1, characterized in that Determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes: From the user feedback data of each recording file, query the complaint data of each recording file; When complaint data of any recording file is queried, it is determined that the user satisfaction status of the recording file is unsatisfactory.

4. The method according to claim 1, characterized in that: Determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes: When it is detected that the interaction mode of the user in any recording file is switched from the intelligent customer service to the manual customer service, it is determined that the user satisfaction status of the recording file is dissatisfied.

5. The method according to claim 1, characterized in that Determining user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files includes: Obtain human voice data from user feedback data of each recording file; For any recording file, determine the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file; When the interaction emotion is a specified emotion, it is determined that the user satisfaction status of the recording file is dissatisfied.

6. The method according to claim 5, characterized in that The step of determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes: Extracting target voice features from the voice data in the recording file; The target speech feature is input into the emotion recognition model, and the interaction emotion between the user and the intelligent customer service during the recording process of the recording file is output. The emotion recognition model is used to recognize the user's emotion based on the speech feature.

7. The method according to claim 5, characterized in that The step of determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes: Extracting a specified keyword from the human voice data in the recording file; When the specified keyword is extracted, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the audio file is the specified emotion.

8. The method according to claim 5, characterized in that The step of determining the interaction emotions between the user and the intelligent customer service during the recording of the recording file according to the human voice data of the recording file includes: Obtaining at least one of the number of repetitions of the question data and the number of voice call requests received from the user within a preset time period from the human voice data in the recording file; When at least one of the number of repetitions of the question data and the number of voice call requests reaches a preset number, it is determined that the interaction emotion between the user and the intelligent customer service during the recording of the recording file is a specified emotion.

9. The method according to claim 1, characterized in that: The preprocessing of the at least one candidate recording file to obtain at least one preprocessed file includes: Each candidate recording file is converted from a voice form to a text form to obtain at least one pre-processed file.

10. The method according to claim 1, characterized in that The preprocessing of the at least one candidate recording file to obtain at least one preprocessed file includes: Each candidate recording file is segmented to obtain multiple recording segments; At least one of the blank data and the target answer data in the multiple recording segments is deleted to obtain a preprocessed file, where the target answer data is all or part of the answer data included in the candidate recording file.

11. The method according to any one of claims 1 to 10, characterized in that The method further comprises: When it is detected that the interaction mode with the user is changed from the intelligent customer service to the manual customer service, a manual question and answer pair provided by the manual customer service is obtained, wherein the manual question and answer pair includes question data that needs to be answered by the manual customer service and manual answer data provided; Based on the manual question and answer pairs, the knowledge base is updated.

12. The method according to any one of claims 1 to 10, characterized in that The method further comprises: Based on the preset screening conditions, at least one target recording file is screened out from the normal recording files; Obtain verification results for answer data in each target recording file; Based on the verification result, the knowledge base is updated.

13. A knowledge base updating device for intelligent customer service, characterized in that: The device comprises: A determination module, configured to determine user satisfaction status of the plurality of recording files according to user feedback data of the plurality of recording files, wherein the recording files are files recorded during the interaction between the user and the intelligent customer service, and the user satisfaction status is used to reflect the user's satisfaction status with the answer data obtained by the intelligent customer service from the knowledge base; A screening module, configured to screen out at least one candidate recording file from the plurality of recording files according to user satisfaction status of the plurality of recording files; A preprocessing module, used for preprocessing the at least one candidate recording file to obtain at least one preprocessed file, wherein the preprocessed file is a text file or a candidate recording file that does not include invalid voice data; An acquisition module, configured to acquire, based on the at least one preprocessed file, at least one abnormal recording file whose answer data needs to be updated; A generating module, configured to generate at least one updated question-answer pair based on the target question data in the at least one abnormal recording file, wherein the updated question-answer pair includes the target question data and updated answer data; An updating module is used to cluster the question data in the knowledge base with the target question data in the at least one updated question-answer pair as the cluster center to obtain at least one question data set, each question data set including at least two question data; and update the answer data corresponding to each question data in each question data set to the updated answer data corresponding to the corresponding cluster center.

14. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein at least one program code is stored in the memory, and the at least one program code is loaded and executed by the processor to implement the knowledge base updating method of the intelligent customer service as described in any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that: At least one program code is stored in the storage medium, and the at least one program code is loaded and executed by the processor to implement the knowledge base updating method of the intelligent customer service as described in any one of claims 1 to 12.

Images