Service quality monitoring method and device

Abstract

The invention provides a service quality monitoring method and device. The method comprises the following steps: acquiring user voice and seat voice of a current round; performing emotion recognition on the user voice and the seat voice of the current round based on an emotion recognition model to obtain a user emotion result and a seat emotion result; and performing service quality evaluation based on the user emotion result and the seat emotion result of the current round, and determining a service quality monitoring result. According to the method provided by the invention, emotion recognition is performed on the user voice and the seat voice of the current round through the emotion recognition model to obtain the user emotion result and the seat emotion result, and then service quality evaluation is performed based on the user emotion result and the seat emotion result of the current round and the historical emotion result of the dialogue voice of the historical round. And the service quality monitoring result of the current round of dialogue voice is determined, so that timely, efficient and accurate service quality monitoring is realized, an enterprise can find a service quality problem in time, and the service quality of an agent is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and in particular, to a method and device for monitoring service quality. Background Art

[0002] With the development of e-commerce and the improvement of customer experience awareness, it has become increasingly important for enterprises to provide high-quality customer service. Traditional service quality monitoring means mostly rely on manual review or customer surveys after providing services to customers to monitor service quality.

[0003] However, the above traditional service quality monitoring means have lag and low efficiency, and due to the lack of a unified objective standard, subjective judgment is prone to cause deviation in the quality monitoring results. Summary of the Invention

[0004] The present invention provides a method and device for monitoring service quality to solve the defects of lag, low efficiency and low accuracy in service quality monitoring based on manual work in the prior art.

[0005] The present invention provides a method for monitoring service quality, including: Obtaining the current round of dialogue voice, where the current round of dialogue voice includes the user voice of the current round and the agent voice of the current round; Performing emotion recognition on the user voice of the current round and the agent voice of the current round respectively based on an emotion recognition model to obtain the user emotion result of the current round and the agent emotion result of the current round; Performing service quality evaluation based on the user emotion result of the current round, the agent emotion result of the current round and the historical emotion results of the historical round of dialogue voice to determine the service quality monitoring result of the current round of dialogue voice; The emotion recognition model is constructed based on a deep learning algorithm.

[0006] According to the method for monitoring service quality provided by the present invention, after obtaining the current round of dialogue voice, it further includes: Performing speech recognition on the current round of dialogue voice to obtain the current round of dialogue text, where the current round of dialogue text includes the user text of the current round and the agent text of the current round; Performing intent recognition on the user text of the current round to obtain the user intent of the current round; Matching to obtain a reference reply text corresponding to the user intent of the current round; Comparing the agent text of the current round with the reference reply text to obtain a reply quality evaluation result.

[0007] A service quality monitoring method provided by the present invention, comparing the seat text of the current round with the reference reply text to obtain a reply quality evaluation result, including: Extract the current reply process of the seat text of the current round and the reference reply process of the reference reply text respectively; Compare the current reply process with the reference reply process to obtain the reply quality evaluation result.

[0008] A service quality monitoring method provided by the present invention, comparing the current reply process with the reference reply process to obtain the reply quality evaluation result, including: Compare the current reply process with the reference reply process to obtain a process evaluation result; Extract the reply keywords of the seat text of the current round, and determine the keyword retrieval result based on the retrieval result of the reply keywords in the preset sensitive word library; Based on the process evaluation result and the keyword retrieval result, obtain the reply quality evaluation result.

[0009] A service quality monitoring method provided by the present invention, The historical emotion result includes the user emotion result of the historical round and the seat emotion result of the historical round; The service quality evaluation based on the user emotion result of the current round, the seat emotion result of the current round, and the historical emotion result of the historical round dialogue voice to determine the service quality monitoring result of the current round dialogue voice includes: Based on the user emotion result of the current round and the user emotion result of the historical round, determine the user emotion change pattern, and based on the seat emotion result of the current round and the seat emotion result of the historical round, determine the seat emotion change pattern; Determine the user emotion monitoring score corresponding to the user emotion change pattern and the seat emotion monitoring score corresponding to the seat emotion change pattern respectively; Based on the user emotion monitoring score, the seat emotion score, and the reply quality evaluation result, calculate to obtain the service quality monitoring result.

[0010] A service quality monitoring method provided by the present invention, calculating to obtain the service quality monitoring result based on the user emotion monitoring score, the seat emotion score, and the reply quality evaluation result, including: Conduct a service quality evaluation on the current round dialogue voice based on the basic evaluation index to obtain a basic evaluation score, and the basic evaluation index includes at least one of reply response time, problem resolution rate, and customer satisfaction; Based on the user emotion monitoring score, the agent emotion score, and the reply quality evaluation result, a process evaluation score is calculated; Based on the basic evaluation score and the process evaluation score, a service quality monitoring result is calculated.

[0011] The present invention also provides a service quality monitoring device, including: An acquisition unit that acquires the current round of dialogue voice, where the current round of dialogue voice includes the user voice of the current round and the agent voice of the current round; An emotion recognition unit that respectively performs emotion recognition on the user voice of the current round and the agent voice of the current round based on an emotion recognition model to obtain the user emotion result of the current round and the agent emotion result of the current round; A service quality monitoring unit that performs service quality evaluation based on the user emotion result of the current round, the agent emotion result of the current round, and the historical emotion results of the historical round of dialogue voice to determine the service quality monitoring result of the current round of dialogue voice; The emotion recognition model is constructed based on a deep learning algorithm.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the service quality monitoring method as described in any one of the above.

[0013] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the service quality monitoring method as described in any one of the above.

[0014] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the service quality monitoring method as described in any one of the above.

[0015] The service quality monitoring method and device provided by the present invention respectively perform emotion recognition on the user voice of the current round and the agent voice of the current round through an emotion recognition model to obtain the user emotion result of the current round and the agent emotion result of the current round, and then perform service quality evaluation based on the user emotion result of the current round, the agent emotion result of the current round, and the historical emotion results of the historical round of dialogue voice to determine the service quality monitoring result of the current round of dialogue voice, realizing timely, efficient, and accurate service quality monitoring, which is beneficial for enterprises to timely discover service quality problems and improve the service quality of agents. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0017] Figure 1 is one of the flow schematic diagrams of the service quality monitoring method provided by the present invention; Figure 2 is another flow schematic diagram of the service quality monitoring method provided by the present invention; Figure 3 is the structural schematic diagram of the service quality monitoring device provided by the present invention; Figure 4 is the structural schematic diagram of the electronic device provided by the present invention. Detailed implementation manners

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0019] To address the above problems, the present invention provides a service quality monitoring method to achieve real-time, efficient, and accurate service quality monitoring. Figure 1 is one of the flow schematic diagrams of the service quality monitoring method provided by the present invention. As Figure 1 shown, the method includes: Step 110, obtaining the current-round dialogue voice, where the current-round dialogue voice includes the user voice of the current round and the agent voice of the current round.

[0020] Here, the current-round dialogue voice refers to the voice in the current-round dialogue between the user and the agent, which may include the user voice of the current round and the agent voice of the current round. Among them, the user voice of the current round refers to the voice emitted by the user in the current-round dialogue for communicating with the agent, and the agent voice of the current round refers to the voice emitted by the agent in the current dialogue for communicating with the user.

[0021] Specifically, the current-round user speech and the current-round agent speech in the current round of conversation can be obtained in real time as the current-round conversation speech. It should be noted that the current-round conversation speech here can refer to all the conversations between the user and the agent regarding a certain topic, which can be defined based on the logical structure and integrity of the conversation content. For example, a complete process of question raising and answering, or a clear business request and response cycle. Or, the current-round conversation speech can also refer to a section of conversation starting from a specific time point and ending at the next specific time point. For example, the entire interaction process from when the user starts speaking to when the agent finishes responding, or from when the agent starts responding to before the user's next speech.

[0022] Step 120, perform emotion recognition on the current-round user speech and the current-round agent speech respectively based on the emotion recognition model to obtain the current-round user emotion result and the current-round agent emotion result; the emotion recognition model is constructed based on a deep learning algorithm.

[0023] Here, the emotion recognition model refers to a model constructed through a deep learning algorithm. It can be trained by a large number of labeled sample speeches to enable the initial emotion recognition model to learn to recognize different emotions, such as happy, angry, sad, frightened, surprised, disgusted, etc., and finally obtain an emotion recognition model that can accurately recognize the emotions in the conversation speech.

[0024] Specifically, the speaker of the current-round conversation speech can be distinguished through speech separation technology to obtain the current-round user speech and the current-round agent speech. Then, the current-round user speech and the current-round agent speech are input into the emotion recognition model, and the emotion recognition model performs emotion recognition on the current-round user speech and the current-round agent speech respectively to obtain the current-round user emotion result corresponding to the current-round user speech and the current-round agent emotion result corresponding to the current-round agent speech. It can be understood that by inputting the current-round conversation speech into the trained emotion recognition model, the emotion recognition model recognizes the tone of the current-round conversation speech to obtain an emotion recognition result that can reflect the speaker's tone and attitude, making the result of service quality monitoring more accurate and objective.

[0025] It should be noted that, compared with the prior art, it is difficult to achieve timely and accurate service quality monitoring by manually monitoring the dialogue voices of all dialogue rounds after the user and the seat complete all dialogue rounds. This makes it difficult to promptly stop the non-compliant behaviors of the seat during the customer service process, which is not conducive to the enterprise's timely discovery of service quality problems and taking measures to improve the service. However, the method provided by the embodiments of the present invention performs emotion recognition on the user voice and the seat voice of the current round in the dialogue voice of the current round through an emotion recognition model, improving the timeliness and accuracy of the service quality monitoring results, enabling the timely prevention of non-compliant behaviors of the seat during the customer service process, being conducive to the enterprise's timely discovery of service quality problems, and improving the service quality of the seat.

[0026] Step 130, perform service quality evaluation based on the user emotion result of the current round, the seat emotion result of the current round, and the historical emotion results of the historical round dialogue voices, and determine the service quality monitoring result of the current round dialogue voice.

[0027] Here, the historical emotion result refers to the user emotion result and the seat emotion result corresponding to each historical round in the historical round dialogue voice. Among them, the historical round dialogue voice can include multiple historical rounds, and there are corresponding user emotion results and seat emotion results for the dialogue voice of each historical round.

[0028] Specifically, the emotion change situations of the user and the seat can be reflected by the user emotion result of the current round, the seat emotion result of the current round, and the historical round dialogue voice. Thus, service quality evaluation can be performed through the emotion change situations of the user and the seat in multiple rounds to determine the service quality monitoring result of the current round dialogue voice.

[0029] It should be noted that during the seat's service to the user, a complete call can include one or multiple rounds of dialogue voices. Then, the service quality monitoring result corresponding to this complete call can be obtained through the service quality detection results of each round of dialogue voice included in this complete call. For example, in the actual service process, if a complete call only includes one round of dialogue voice, the service quality monitoring result of the current round dialogue voice can be used as the service quality monitoring result of this call; if a complete call only includes multiple rounds of dialogue voices, the service quality monitoring results of the current round dialogue voice and the historical round dialogue voices can be jointly used to obtain the service quality monitoring result of this call. That is, when the current round dialogue voice is the last round of the call, while outputting the service quality monitoring result of the current round dialogue voice, the service quality monitoring result of this call will also be output together.

[0030] It can be understood that by combining the user emotion results of the current round, the seat emotion results of the current round, and the historical emotion results of the historical round of conversation voice for service quality evaluation, the emotional changes of the user and the seat from the first round of conversation voice to the current round of conversation voice can be obtained, and thus a more comprehensive, accurate, and reliable service quality monitoring result can be achieved.

[0031] The method provided by the embodiments of the present invention performs emotion recognition on the user voice of the current round and the seat voice of the current round respectively through an emotion recognition model to obtain the user emotion results of the current round and the seat emotion results of the current round, and then performs service quality evaluation based on the user emotion results of the current round, the seat emotion results of the current round, and the historical emotion results of the historical round of conversation voice to determine the service quality monitoring result of the current round of conversation voice, realizing timely, efficient, and accurate service quality monitoring, which is beneficial for enterprises to timely discover service quality problems and improve the service quality of the seat.

[0032] Based on any of the above embodiments, after step 110, it further includes: Performing speech recognition on the current round of conversation voice to obtain the current round of conversation text, where the current round of conversation text includes the user text of the current round and the seat text of the current round; Performing intent recognition on the user text of the current round to obtain the user intent of the current round; Matching to obtain a reference reply text corresponding to the user intent of the current round; Comparing the seat text of the current round and the reference reply text to obtain a reply quality evaluation result.

[0033] Specifically, first, speech recognition can be performed on the user voice of the current round and the seat voice of the current round in the current round of conversation voice to obtain the user text of the current round corresponding to the user voice of the current round and the seat text of the current round corresponding to the seat voice of the current round. Then, the user text of the current round and the seat text of the current round can be used as the current round of conversation text. It should be noted that in order to improve the accuracy of subsequent language processing, text preprocessing can be performed on the initial recognition text obtained after performing speech recognition on the current round of conversation voice, including operations such as removing stop words and stemming, and using the text after text preprocessing as the current round of conversation text.

[0034] Then, intent recognition can be performed on the user text of the current round through natural language understanding technology to obtain the user intent of the current round. For example, the user text of the current round can be input into a pre-trained word embedding or deep learning model to perform text analysis on the user text of the current round and predict the user intent of the current round of the user text of the current round.

[0035] Furthermore, a reference response text corresponding to the user intent of the current round can be obtained by matching from a pre-constructed response text resource library. It should be noted that the reference response text can answer the necessary response elements of the user intent of the current round, and can be used to evaluate whether the response of the agent text in the current round meets the necessary response elements of the user intent of the current round, and then can reflect the service professional level of the agent.

[0036] Next, the agent text of the current round and the reference response text can be compared to obtain a response quality evaluation result. For example, the text similarity between the agent text of the current round and the reference response text can be calculated, such as cosine similarity, Jaccard similarity, or a more complex semantic matching algorithm can be used to achieve this. The response quality evaluation result here can be a quantitative score, such as 0 - 100 points, indicating the degree of closeness between the agent's response and the reference response. Or, the response quality evaluation result can also be a qualitative label, such as "excellent", "good", "average", "poor", used to represent the overall quality of the response.

[0037] The method provided by the embodiments of the present invention performs speech recognition on the obtained current-round dialogue voice to obtain the user text of the current round and the agent text of the current round. Then, by performing intent recognition on the user text of the current round, the user intent of the current round is obtained, and a reference response text corresponding to the user intent of the current round is matched to compare the agent text of the current round and the reference response text, so as to obtain a response quality evaluation result that can be used to reflect the professional level of the agent, realizing the monitoring of the service quality in the dimension of the agent's business level, avoiding response content that does not conform to the customer's demands, being beneficial to improving the business response level of the agent, and being able to provide professional and effective customer service for users in a timely manner.

[0038] Based on any of the above embodiments, comparing the agent text of the current round and the reference response text to obtain a response quality evaluation result includes: Respectively extract the current response process of the agent text of the current round and the reference response process of the reference response text; Compare the current response process and the reference response process to obtain the response quality evaluation result.

[0039] Specifically, the response process extraction can be separately performed on the seat text and the reference response text of the current round to obtain the current response process and the reference response process. Specifically, it can be to perform word segmentation and part-of-speech tagging on the seat text of the current round to identify the keywords and phrases in the seat text of the current round. Then, through dependency syntactic analysis, the sentence structure in the text can be parsed, the main information and logical relationships can be identified, and the current response process can be extracted. Similarly, the reference response process of the reference response text can be obtained through the same extraction method.

[0040] After obtaining the current response process and the reference response process, the response quality evaluation result can be obtained by comparing the current response process and the reference response process. Specifically, it can be to align the current response process and the reference response process to ensure that they are at the same logical level. Then, compare the nodes in the two processes one by one to evaluate whether the seat text of the current round covers all the key information points in the reference response text and whether the information is provided in the correct order, and then the response quality evaluation result can be obtained according to the comparison result.

[0041] The method provided by the embodiments of the present invention realizes timely discovery of the problems and deficiencies existing in the seat during the response process by separately extracting the current response process of the seat text of the current round and the reference response process of the reference response text, comparing the current response process and the reference response process, and obtaining the response quality evaluation result, so as to take corresponding improvement measures, which helps to improve the response ability of the seat and customer satisfaction.

[0042] Based on any of the above embodiments, the comparing the current response process and the reference response process to obtain the response quality evaluation result includes: Comparing the current response process and the reference response process to obtain a process evaluation result; Extracting the response keywords of the seat text of the current round, and determining the keyword retrieval result based on the retrieval result of the response keywords in the preset sensitive word library; Based on the process evaluation result and the keyword retrieval result, obtaining the response quality evaluation result.

[0043] Specifically, the process evaluation result can be obtained by comparing the nodes in the current response process and the reference response process and through the comparison results of each node. In addition, the response keywords in the seat text of the current round can be extracted, and then the keyword retrieval result can be determined through the retrieval result of the response keywords in the preset sensitive word library.

[0044] It should be noted that by searching the reply keywords in the preset sensitive word library, the keyword search results obtained can be used to reflect whether the agent text of the current round contains negative, sensitive, illegal and other words that do not comply with regulations.

[0045] Finally, the process evaluation results and the scores corresponding to the keyword search results can be weighted to calculate the response quality evaluation results that can reflect the agent's response ability and the degree of response standardization.

[0046] The method provided by the embodiment of the present invention obtains a process evaluation result by comparing the current reply process of the agent text of the current round with the reference reply process, and obtains a keyword retrieval result by performing a sensitive word search on the reply keywords in the agent text of the current round, thereby realizing quality monitoring of the reply process of the agent text of the current round and quality monitoring of the reply standardization degree of the agent text of the current round, thereby making the final service quality monitoring result more comprehensive.

[0047] Based on any of the above embodiments, the historical emotion results include the user emotion results of the historical rounds and the agent emotion results of the historical rounds; step 130 includes: Determine a user emotion change pattern based on the user emotion results of the current round and the user emotion results of the historical rounds, and determine an agent emotion change pattern based on the agent emotion results of the current round and the agent emotion results of the historical rounds; Respectively determining a user emotion monitoring score corresponding to the user emotion change pattern and an agent emotion monitoring score corresponding to the agent emotion change pattern; The service quality monitoring result is calculated based on the user emotion monitoring score, the agent emotion score and the response quality evaluation result.

[0048] Specifically, first, the user emotion change pattern can be determined by the user emotion results of the current round and the user emotion results of the historical rounds. And, the agent emotion change pattern can be determined based on the agent emotion results of the current round and the agent emotion results of the historical rounds. It should be noted that the emotion change pattern may include the emotion changes of two adjacent rounds of conversation from the first round to the current round of conversation. The user emotion change pattern or the agent emotion change pattern can be "from happy to angry", or "from angry to happy".

[0049] Then, the user emotion monitoring score corresponding to the user emotion change pattern and the agent emotion monitoring score corresponding to the agent emotion change pattern can be determined respectively according to the scoring rules corresponding to the preset emotion change patterns. For example, if the user emotion change pattern is "changing from anger to happiness", the corresponding user emotion monitoring score is 9 points; if the user emotion change pattern is "changing from happiness to anger", the corresponding user emotion monitoring score is 1 point. Another example, if the agent emotion change pattern is "changing from anger to happiness", the corresponding agent emotion monitoring score is 6 points; if the agent emotion change pattern is "changing from happiness to anger", the corresponding agent emotion monitoring score is 4 points.

[0050] Finally, the service quality monitoring score can be obtained through weighted calculation of the user emotion monitoring score, the agent emotion score, and the score corresponding to the reply quality evaluation result, as the service quality monitoring result of the current round of dialogue voice. Among them, the weights corresponding to the user emotion monitoring score, the agent emotion score, and the reply quality evaluation result score can be determined in advance. For example, the weights corresponding to the user emotion monitoring score, the agent emotion score, and the reply quality evaluation result score can be 0.5, 0.3, and 0.2 in sequence.

[0051] Based on any of the above embodiments, calculating the service quality monitoring result based on the user emotion monitoring score, the agent emotion score, and the reply quality evaluation result includes: Evaluating the service quality of the current round of dialogue voice based on the basic evaluation indicators to obtain a basic evaluation score, where the basic evaluation indicators include at least one of the reply response time, the problem-solving rate, and the customer satisfaction; Calculating a process evaluation score based on the user emotion monitoring score, the agent emotion score, and the reply quality evaluation result; Calculating the service quality monitoring result based on the basic evaluation score and the process evaluation score.

[0052] Here, the reply response time can be used to measure the time interval from when the user asks a question to when the agent gives a reply. A shorter response time usually means better service efficiency. The problem-solving rate can be used to evaluate the proportion of problems that the agent successfully solves for the user during the conversation. A high problem-solving rate indicates the effectiveness of the service. Customer satisfaction refers to measuring the degree of satisfaction of the user with the service quality of the agent through user feedback, such as scoring, satisfaction surveys, etc. Thus, the basic evaluation indicators composed of at least one of the reply response time, the problem-solving rate, and the customer satisfaction can provide an objective basis for the preliminary evaluation of service quality.

[0053] Specifically, for evaluating the service quality of the current round of conversation based on the reply response time, it can be first to calculate the reply response time of the agent to the user speech of the current round through the time interval between the user speech of the current round and the agent speech of the current round in the current round of conversation voice. Finally, the service quality can be evaluated according to the length of the reply response time, and the score corresponding to the reply response time can be determined.

[0054] For evaluating the service quality of the current round of conversation voice through the problem-solving rate, it can be to perform natural language processing on the current round of conversation text corresponding to the current round of conversation voice to obtain the solution rate corresponding to the user intention of the current round, and use the solution rate corresponding to the user intention of the current round as the problem-solving rate. Or, it can be obtained by acquiring the input of the user's problem-solving rate for the agent's customer service. Finally, the service quality can be evaluated according to the level of the problem-solving rate, and the score corresponding to the problem-solving rate can be determined.

[0055] In addition, for evaluating the service quality of the current round of conversation voice through customer satisfaction, it can be to obtain the customer satisfaction by receiving the user evaluation of the current round of conversation voice. Finally, the service quality can be evaluated according to the level of the customer satisfaction, and the score corresponding to the customer satisfaction can be determined.

[0056] Thus, the basic evaluation score can be obtained by cumulative calculation of the scores corresponding to at least one of the basic evaluation indicators including the reply response time, the problem-solving rate, and the customer satisfaction. For example, the scores corresponding to the reply response time, the problem-solving rate, and the customer satisfaction can be weighted and cumulatively calculated to obtain the final basic evaluation score.

[0057] Then, the process evaluation score can be calculated by combining the user emotion monitoring score, the agent emotion score, and the reply quality evaluation result. Further, a more comprehensive, reliable, and accurate service quality monitoring result can be calculated through the basic evaluation score and the process evaluation score.

[0058] The method provided by the embodiment of the present invention, on the basis of calculating the emotion monitoring score and the reply quality evaluation result, introduces basic evaluation indicators including at least one of the reply response time, the problem-solving rate, and the customer satisfaction, and evaluates the service quality of the current round of conversation voice through the basic evaluation indicators, so as to obtain a more comprehensive and objective service quality evaluation result. Furthermore, it can realize objective evaluation and monitoring of the service quality according to the conversation content and emotion between the agent and the user, which helps enterprises such as e-commerce, banks, and telecommunications to timely discover service quality problems and take measures to improve the service.

[0059] Based on any of the above embodiments, Figure 2 is the second flowchart of the service quality monitoring method provided by the present invention, asFigure 2 As shown, the method includes: after the agent starts a conversation with the user, the customer expresses their demands through the current conversation voice and recognizes the intention and emotion through the current conversation voice. Then, the agent replies, generates the agent voice for the current turn, and recognizes the reply content and emotion corresponding to the agent voice for the current turn. In each conversation turn, the same operations are performed on the current conversation voice for this turn and the agent voice for the current turn to monitor the service quality of this conversation turn until the conversation ends, and a service quality report is generated and reported or shared in real time. It should be noted that by generating the service quality report, the advantages and disadvantages in the agent's service can be pointed out, which is convenient for the agent to adjust their working state and for department work management. For different situations that occur during the service, real-time reporting is supported, which is convenient for identifying excellent cases to assist agent training, handle red-line accidents, etc.

[0060] It should be noted that the method provided by the embodiments of the present invention realizes comprehensive, accurate, timely and effective service quality monitoring by respectively performing emotion recognition and natural language processing on the user voice for the current turn and the agent voice for the current turn in each current turn of the conversation voice during the conversation between the agent and the user.

[0061] Based on any of the above embodiments, Figure 3 is a schematic structural diagram of the service quality monitoring device provided by the present invention, as Figure 3 shown, the device includes: An acquisition unit 310 that acquires the current turn of conversation voice, where the current turn of conversation voice includes the user voice for the current turn and the agent voice for the current turn; An emotion recognition unit 320 that respectively performs emotion recognition on the user voice for the current turn and the agent voice for the current turn based on an emotion recognition model to obtain the user emotion result for the current turn and the agent emotion result for the current turn; A service quality monitoring unit 330 that performs service quality evaluation based on the user emotion result for the current turn, the agent emotion result for the current turn, and the historical emotion results of the historical turn of conversation voice to determine the service quality monitoring result of the current turn of conversation voice; The emotion recognition model is constructed based on a deep learning algorithm.

[0062] The device provided by the embodiment of the present invention performs emotion recognition on the user speech of the current round and the agent speech of the current round through an emotion recognition model, obtains the user emotion result of the current round and the agent emotion result of the current round, and then conducts service quality evaluation based on the user emotion result of the current round, the agent emotion result of the current round, and the historical emotion results of the historical round of dialogue speech, determines the service quality monitoring result of the current round of dialogue speech, realizes timely, efficient, and accurate service quality monitoring, is conducive to the enterprise to timely discover service quality problems, and improves the service quality of the agent.

[0063] Based on any of the above embodiments, after the acquisition unit, there is further a natural language processing unit, and the natural language processing unit is specifically used for: Perform speech recognition on the current round of dialogue speech to obtain the current round of dialogue text, where the current round of dialogue text includes the user text of the current round and the agent text of the current round; Perform intent recognition on the user text of the current round to obtain the user intent of the current round; Match and obtain a reference reply text corresponding to the user intent of the current round; Compare the agent text of the current round with the reference reply text to obtain a reply quality evaluation result.

[0064] Based on any of the above embodiments, the natural language processing unit is further specifically used for: Extract the current reply process of the agent text of the current round and the reference reply process of the reference reply text respectively; Compare the current reply process with the reference reply process to obtain the reply quality evaluation result.

[0065] Based on any of the above embodiments, the natural language processing unit is further specifically used for: Compare the current reply process with the reference reply process to obtain a process evaluation result; Extract the reply keywords of the agent text of the current round, and determine a keyword search result based on the search result of the reply keywords in a preset sensitive word library; Based on the process evaluation result and the keyword search result, obtain the reply quality evaluation result.

[0066] Based on any of the above embodiments, the historical emotion results include the user emotion result of the historical round and the agent emotion result of the historical round, and the service quality monitoring unit is specifically used for: Based on the user emotion result of the current round and the user emotion result of the historical round, determine the user emotion change pattern, and based on the agent emotion result of the current round and the agent emotion result of the historical round, determine the agent emotion change pattern; Determine the user emotion monitoring score corresponding to the user emotion change pattern and the seat emotion monitoring score corresponding to the seat emotion change pattern respectively; Calculate the service quality monitoring result based on the user emotion monitoring score, the seat emotion score, and the reply quality evaluation result.

[0067] Based on any of the above embodiments, the service quality monitoring unit is further specifically configured to: Evaluate the service quality of the current round of dialogue voice based on basic evaluation indicators to obtain a basic evaluation score, where the basic evaluation indicators include at least one of reply response time, problem-solving rate, and customer satisfaction; Calculate a process evaluation score based on the user emotion monitoring score, the seat emotion score, and the reply quality evaluation result; Calculate the service quality monitoring result based on the basic evaluation score and the process evaluation score.

[0068] Figure 4 An example of a schematic physical structure diagram of an electronic device is shown as Figure 4 shown. The electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440. Among them, the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 can call the logical instructions in the memory 430 to execute the service quality monitoring method, which includes: obtaining the current round of dialogue voice, where the current round of dialogue voice includes the user voice of the current round and the seat voice of the current round; performing emotion recognition on the user voice of the current round and the seat voice of the current round respectively based on an emotion recognition model to obtain the user emotion result of the current round and the seat emotion result of the current round; performing service quality evaluation based on the user emotion result of the current round, the seat emotion result of the current round, and the historical emotion results of the historical round of dialogue voice to determine the service quality monitoring result of the current round of dialogue voice; the emotion recognition model is constructed based on a deep learning algorithm.

[0069] In addition, when the logical instructions in the above-mentioned memory 430 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0070] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the service quality monitoring method provided by the above-mentioned various methods. The method includes: obtaining the current-round dialogue voice, where the current-round dialogue voice includes the user voice of the current round and the agent voice of the current round; respectively performing emotion recognition on the user voice of the current round and the agent voice of the current round based on an emotion recognition model to obtain the user emotion result of the current round and the agent emotion result of the current round; performing service quality evaluation based on the user emotion result of the current round, the agent emotion result of the current round, and the historical emotion results of the historical-round dialogue voice to determine the service quality monitoring result of the current-round dialogue voice; the emotion recognition model is constructed based on a deep learning algorithm.

[0071] On another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the service quality monitoring method provided by the above-mentioned various methods. The method includes: obtaining the current-round dialogue voice, where the current-round dialogue voice includes the user voice of the current round and the agent voice of the current round; respectively performing emotion recognition on the user voice of the current round and the agent voice of the current round based on an emotion recognition model to obtain the user emotion result of the current round and the agent emotion result of the current round; performing service quality evaluation based on the user emotion result of the current round, the agent emotion result of the current round, and the historical emotion results of the historical-round dialogue voice to determine the service quality monitoring result of the current-round dialogue voice; the emotion recognition model is constructed based on a deep learning algorithm.

[0072] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0073] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or equivalently replace some of the technical features. However, these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring service quality, characterized in that: include: Acquire the current round of dialogue voice, where the current round of dialogue voice includes the current round of user voice and the current round of agent voice; Based on the emotion recognition model, emotion recognition is performed on the user voice of the current round and the agent voice of the current round respectively, to obtain the user emotion result of the current round and the agent emotion result of the current round; Performing a service quality assessment based on the user emotion result of the current round, the agent emotion result of the current round, and the historical emotion results of the conversation speech of the previous round, and determining the service quality monitoring result of the conversation speech of the current round; The emotion recognition model is built based on a deep learning algorithm.

2. The service quality monitoring method according to claim 1, characterized in that: The method of obtaining the current round of dialogue voice further includes: Performing speech recognition based on the current round of conversation speech to obtain the current round of conversation text, wherein the current round of conversation text includes the user text of the current round and the agent text of the current round; Performing intent recognition on the user text of the current round to obtain the user intent of the current round; Matching and obtaining a reference reply text corresponding to the user intention of the current round; Compare the agent text of the current round with the reference answer text to obtain an answer quality evaluation result.

3. The service quality monitoring method according to claim 2, characterized in that: The comparing the agent text of the current round with the reference answer text to obtain the answer quality evaluation result includes: Respectively extracting the current reply process of the agent text of the current round and the reference reply process of the reference reply text; The current reply process is compared with the reference reply process to obtain the reply quality evaluation result.

4. The service quality monitoring method according to claim 3, characterized in that: The comparing the current reply process with the reference reply process to obtain the reply quality evaluation result includes: Comparing the current response process with the reference response process to obtain a process evaluation result; Extracting the reply keywords of the agent text of the current round, and determining the keyword search results based on the search results of the reply keywords in the preset sensitive word library; Based on the process evaluation result and the keyword search result, the answer quality evaluation result is obtained.

5. The service quality monitoring method according to any one of claims 2 to 4, characterized in that: The historical emotion results include the user emotion results of historical rounds and the agent emotion results of historical rounds; The performing of service quality evaluation based on the user emotion result of the current round, the agent emotion result of the current round, and the historical emotion result of the conversation speech of the previous rounds to determine the service quality monitoring result of the conversation speech of the current round includes: Determine a user emotion change pattern based on the user emotion results of the current round and the user emotion results of the historical rounds, and determine an agent emotion change pattern based on the agent emotion results of the current round and the agent emotion results of the historical rounds; Respectively determining a user emotion monitoring score corresponding to the user emotion change pattern and an agent emotion monitoring score corresponding to the agent emotion change pattern; The service quality monitoring result is calculated based on the user emotion monitoring score, the agent emotion score and the response quality evaluation result.

6. The service quality monitoring method according to claim 5, characterized in that: The service quality monitoring result is calculated based on the user emotion monitoring score, the agent emotion score and the response quality evaluation result, including: Performing a service quality evaluation on the current round of conversation voice based on basic evaluation indicators to obtain a basic evaluation score, wherein the basic evaluation indicators include at least one of reply response time, problem resolution rate, and customer satisfaction; Calculate a process evaluation score based on the user emotion monitoring score, the agent emotion score and the answer quality evaluation result; Based on the basic evaluation score and the process evaluation score, a service quality monitoring result is calculated.

7. A service quality monitoring device, characterized in that: include: An acquisition unit, which acquires the current round of dialogue voice, wherein the current round of dialogue voice includes the current round of user voice and the current round of agent voice; An emotion recognition unit, which performs emotion recognition on the user voice of the current round and the agent voice of the current round respectively based on the emotion recognition model to obtain the user emotion result of the current round and the agent emotion result of the current round; A service quality monitoring unit, which performs service quality evaluation based on the user emotion result of the current round, the agent emotion result of the current round, and the historical emotion results of the conversation speech of the previous round, and determines the service quality monitoring result of the conversation speech of the current round; The emotion recognition model is built based on a deep learning algorithm.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the service quality monitoring method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the service quality monitoring method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the service quality monitoring method according to any one of claims 1 to 6 is implemented.

Images