Question and answer quality evaluation method, device and equipment based on artificial intelligence and medium

Abstract

The invention relates to the technical field of artificial intelligence, and discloses an artificial intelligence-based question and answer quality evaluation method, device and equipment and a medium, and the method comprises the steps: collecting a historical question and answer sheet, and splitting the historical question and answer sheet into a training set and a verification set; taking the parameter combination of the BP neural network as a search object of a hunter prey optimization algorithm, and taking the fitness of the parameter combination of the BP neural network on the verification set as a reference to obtain an optimized BP neural network; training the optimized BP neural network by using the training set to obtain a trained BP neural network; and inputting a to-be-tested question and answer sheet into the trained BP neural network to obtain an evaluation level, and adjusting the intelligent customer service system by using the screened question and answer sheet. According to the embodiment of the invention, grade evaluation is carried out on each question and answer sheet, and the intelligent customer service system is adjusted by using the screened question and answer sheet, so that the accuracy and user satisfaction of the intelligent customer service system are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based question and answer quality assessment method, device, equipment and medium. Background Art

[0002] Insurance customer service is one of the services closest to users in the insurance industry. It is not only related to the promotion of insurance knowledge, the expansion of insurance business, and pre-sales and after-sales consultation, but also directly affects the business volume and performance of insurance companies. With the development of science and technology, it has become a very common technical means to obtain data and use data modeling to solve corresponding problems. Some relatively simple tasks in insurance customer service, such as user consultation and user background assessment, can be replaced by intelligent customer service systems. Intelligent customer service or intelligent assistant is one of the most extensive and important ways to implement natural language processing (NLP) technology in actual scenarios. Among them, the question-answering engine is the core module of the intelligent customer service system. For the question-answering engine, the most commonly used technology is the retrieval-based question-answering technology, that is, some question-answering knowledge bases are defined in advance by business experts, a series of question-answering pairs, each standard question corresponds to an answer, and then the user's question is found through the semantic matching model technology. Synonymous questions are then replied with corresponding answers.

[0003] However, since each user has different personalities and preferences, the pre-set standard answers may not meet the needs of some user groups, resulting in low accuracy and user satisfaction of the intelligent customer service system. Therefore, it is also necessary to evaluate the question and answer quality of the intelligent user system to facilitate subsequent improvements to the intelligent customer service system and further improve user satisfaction with the intelligent customer service system. Summary of the invention

[0004] The present invention provides a question and answer quality assessment method, device, equipment and medium based on artificial intelligence to solve the technical problems of accuracy of question and answer sheets and low user satisfaction in intelligent customer service systems.

[0005] In the first aspect, a question-answering quality assessment method based on artificial intelligence is provided, comprising:

[0006] Collect historical Q&A forms and split them into a training set and a validation set. The historical Q&A forms include the number of Q&A requests processed, response time, successful resolution rate, user satisfaction, and chat duration;

[0007] The parameter combination of the BP neural network is used as the search object of the hunter-prey optimization algorithm, and the fitness of the parameter combination of the BP neural network on the verification set is used as a reference to update the parameter combination of the BP neural network, and iterate a preset number of times to obtain an optimized BP neural network, wherein the parameter combination includes the number of neurons in each layer, weight parameters, bias parameters, learning rate and regularization parameters;

[0008] Using the training set to train the optimized BP neural network to obtain a trained BP neural network;

[0009] The question and answer sheet to be tested is input into the trained BP neural network to obtain an evaluation grade, so as to screen each question and answer sheet according to the evaluation grade corresponding to each question and answer sheet of the intelligent customer service system, and use the screened question and answer sheets to adjust the intelligent customer service system.

[0010] In a second aspect, a question-answering quality assessment device based on artificial intelligence is provided, comprising:

[0011] A collection module is used to collect historical question-and-answer forms and split the historical question-and-answer forms into a training set and a validation set. The historical question-and-answer forms include the number of question-and-answer requests processed, response time, successful resolution rate, user satisfaction, and chat duration;

[0012] A parameter module, used to use the parameter combination of the BP neural network as a search object of the hunter-prey optimization algorithm, update the parameter combination of the BP neural network with reference to the fitness of the parameter combination of the BP neural network on the verification set, and perform a preset number of iterations to obtain an optimized BP neural network, wherein the parameter combination includes the number of neurons in each layer, weight parameters, bias parameters, learning rate and regularization parameters;

[0013] A training module, used to train the optimized BP neural network using the training set to obtain a trained BP neural network;

[0014] The evaluation module is used to input the question and answer form to be tested into the trained BP neural network to obtain an evaluation grade, so as to screen each question and answer form according to the evaluation grade corresponding to each question and answer form of the intelligent customer service system, and use the screened question and answer forms to adjust the intelligent customer service system.

[0015] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned artificial intelligence-based question and answer quality assessment method when executing the computer program.

[0016] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned artificial intelligence-based question and answer quality assessment method are implemented.

[0017] In the scheme implemented by the above-mentioned artificial intelligence-based question and answer quality assessment method, device, equipment and medium, historical question and answer forms are first collected and split into a training set and a verification set; then the parameter combination of the BP neural network is optimized using the training set and the hunter-prey optimization algorithm, and the fitness of the parameter combination of the BP neural network on the verification set is used as a reference. After multiple iterations, the optimized BP neural network is finally obtained; then the optimized BP neural network is trained using the training set to obtain a trained BP neural network; finally, the question and answer form to be tested is input into the trained BP neural network to obtain the evaluation level of the question and answer form to be tested, so as to adjust the intelligent customer service system according to the evaluation level.

[0018] The embodiment of the present invention utilizes the hunter-prey optimization algorithm in combination with the BP neural network to construct a question-answering quality evaluation model. Since the complexity of BP neural network training is relatively high, the global search capability of the hunter-prey optimization algorithm can effectively optimize the parameter combination of the BP neural network, so that the parameter combination of the optimized BP neural network can reach a better state. When the optimized BP neural network is subsequently trained using the training set, the complexity of training can be reduced, thereby improving the prediction accuracy and stability of the trained BP neural network.

[0019] In addition, each question and answer form is evaluated through the trained BP neural network, and the question and answer forms are screened according to the evaluated levels. The screened question and answer forms are used to adjust the intelligent customer service system, thereby improving the accuracy of the intelligent customer service system and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0021] Figure 1 is a schematic diagram of an application environment of a question-answering quality assessment method based on artificial intelligence in one embodiment of the present invention;

[0022] Figure 2 A schematic diagram of a process of a question-answer quality assessment method based on artificial intelligence provided by an embodiment of the present invention;

[0023] Figure 3 for Figure 2 A schematic diagram of a process of step S10;

[0024] Figure 4 for Figure 2 A schematic diagram of a process of step S20;

[0025] Figure 5 for Figure 2 A schematic diagram of a process of step S40;

[0026] Figure 6 is a structural schematic diagram of a question-answer quality assessment device based on artificial intelligence provided by an embodiment of the present invention;

[0027] Figure 7 is a schematic diagram of a structure of a computer device in one embodiment of the present invention;

[0028] Figure 8 It is another structural schematic diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] The embodiment of the present invention provides an artificial intelligence-based question and answer quality assessment method, which can be applied in the following aspects: Figure 1 In an application environment, the client communicates with the server through a network. The server can receive the question-and-answer form processed by the client or the consulting question input by the user through the client, and split the historical question-and-answer form into a training set and a verification set; use the parameter combination of the BP neural network as the search object of the hunter-prey optimization algorithm, and update the parameter combination of the BP neural network with reference to the fitness of the parameter combination of the BP neural network on the verification set, and perform a preset number of iterations to obtain the optimized BP neural network; use the training set to train the optimized BP neural network to obtain the trained BP neural network; input the question-and-answer form to be tested into the trained BP neural network to obtain an evaluation level, so as to screen each question-and-answer form according to the evaluation level corresponding to each question-and-answer form of the intelligent customer service system, and use the screened question-and-answer forms to adjust the intelligent customer service system.

[0031] In the present invention, for complex insurance entities such as insurance business, an artificial intelligence-based question-answer quality evaluation scheme can be used, and a hunter-prey optimization algorithm combined with a BP neural network can be used to construct a question-answer quality evaluation model. Since the complexity of BP neural network training is relatively high, the global search capability of the hunter-prey optimization algorithm can effectively optimize the parameter combination of the BP neural network, so that the parameter combination of the optimized BP neural network can reach a better state. When the optimized BP neural network is subsequently trained using a training set, the complexity of training can be reduced, thereby improving the prediction accuracy and stability of the trained BP neural network. In addition, each question-answer form is evaluated for a grade through the trained BP neural network, and the question-answer form is screened according to the evaluated grade, and the screened question-answer form is used to adjust the intelligent customer service system, thereby improving the accuracy of the intelligent customer service system and user satisfaction. The present invention is described in detail below through specific embodiments.

[0032] See also Figure 2 As shown, the method comprises the following steps:

[0033] S10: Collect historical Q&A sheets and split them into a training set and a validation set. The historical Q&A sheets include the number of processed Q&A requests, response time, successful resolution rate, user satisfaction, and chat duration.

[0034] In order to evaluate the quality of the questions and answers in the questionnaire, the embodiment of the present invention constructs an artificial intelligence model for evaluating the processing quality of the questionnaire. It is necessary to first consider the evaluation index system, and then quantify the features that can be used to build the artificial intelligence model from these indicators. The artificial intelligence model obtained based on these features can more accurately evaluate the quality of the questions and answers. The details are as follows:

[0035] First, we analyze the quality assessment dimensions. The quality of the Q&A form needs to be quantified from multiple dimensions, including user experience and service quality. Specific assessment indicators can be abstracted from these two dimensions. Among them, the user experience dimension can indicate that the user experience can intuitively show the quality of the Q&A service, usually with user satisfaction as a quantitative indicator; the service quality dimension can indicate that the service quality is mainly reflected in the professional ability of the Q&A during the service process, and usually professionals will conduct quality inspections and then give scores.

[0036] Further defining quantifiable evaluation indicators, for the above evaluation indicators, it is necessary to abstract specific quantifiable features. Combined with the work attributes of insurance customer service, the embodiment of the present invention defines the following indicator features:

[0037] Number of Q&A requests processed: The number of user inquiries or complaints processed during a specific time period.

[0038] Response time: The average time from when a user initiates a request to when customer service responds for the first time.

[0039] Success resolution rate: the proportion of user issues that are successfully resolved.

[0040] User satisfaction: Satisfaction ratings obtained through user feedback or surveys.

[0041] Chat duration: The average duration of each interaction with a user.

[0042] After determining the evaluation indicators, all historical Q&A forms processed by the intelligent customer service system are collected. The historical Q&A forms include the above-mentioned evaluation indicators, such as the number of Q&A requests processed, response time, successful resolution rate, user satisfaction, and chat duration.

[0043] The historical question-and-answer sheet is used as a training sample or a test sample, and all historical question-and-answer sheets are divided into two parts: a training set and a verification set. The verification set is used to verify the fitness of the parameter combination of the subsequent BP neural network, and the training set is used to perform quality assessment training on the subsequent BP neural network. The evaluation level of the historical question-and-answer sheet is also needed during training. The evaluation level is the evaluation score given by professionals to each question-and-answer sheet, and different scores correspond to different levels. For example, the evaluation score is from 1 to 100, and 1-10, 11-20, ..., 91-100 are divided into ten evaluation levels. The evaluation level is used as a label for the historical question-and-answer sheet and is used when training the BP neural network in the training set.

[0044] As an implementation, see Figure 3 As shown, step S10 includes the following steps:

[0045] S11, obtaining interaction data between the customer service and the user through the log of the intelligent customer service system, wherein the interaction data includes the number of question and answer requests processed in a preset time period, the response time of each question and answer request, the chat duration of each question and answer request, and whether the question and answer request is successfully resolved;

[0046] S12, collect user satisfaction for each Q&A request through email, SMS, and online surveys;

[0047] S13, cleaning, feature extraction, and normalization processing are performed on the interaction data and the user satisfaction to obtain the historical question and answer list.

[0048] In the embodiment of the present invention, the log of the intelligent customer service system is used to obtain the detailed data of the communication between the intelligent customer service and the user, and the detailed data is processed, such as converting the voice in the communication process into text, converting the pictures and emoticons sent by the user into text, and extracting the interaction data from the text finally converted. The interaction data includes the number of Q&A requests processed in the preset time period, the response time of each Q&A request, the chat time of each Q&A request, and whether the Q&A request is successfully resolved, wherein the preset time period can be a predefined period of time, such as 8:00 am to 5:00 pm, etc.; the response time of each Q&A request refers to the time difference between the time when the user asks a question and the intelligent customer service displays all the text answers in the text box; the chat time of the Q&A request refers to the time difference between the time when the user asks the first question and the time when the user gives feedback; whether the Q&A request is successfully resolved, in the intelligent customer service system, the Q&A request finally sends the user whether the Q&A request is successfully resolved, and the user has two options of "yes" and "no". According to the option of the user feedback, it is known whether the user successfully resolves the Q&A request.

[0049] In addition to the logs of the intelligent customer service system, the user satisfaction with the answer to each question and answer request of the intelligent customer service system is collected through any of the following methods: email, text message and online survey. Different survey methods can be used for users of different professions. For example, each user's satisfaction is first surveyed through email, and then the satisfaction of each user is collected through text message or online survey for users who do not reply to emails.

[0050] The interaction data and user satisfaction data collected initially are cleaned and feature extracted. The data cleaning step can remove duplicate, erroneous or irrelevant data. The feature extraction step can extract features that are effective for quality assessment from the original data. Finally, each historical question and answer form is collected.

[0051] S20: Using the parameter combination of the BP neural network as the search object of the hunter-prey optimization algorithm, and taking the fitness of the parameter combination of the BP neural network on the verification set as a reference, updating the parameter combination of the BP neural network, and performing a preset number of iterations to obtain an optimized BP neural network, wherein the parameter combination includes the number of neurons in each layer, weight parameters, bias parameters, learning rate and regularization parameters.

[0052] The output of the assessment result of the question and answer quality assessment model can be a score value, and its corresponding input should be the eigenvalue after the above-mentioned preprocessing. Therefore, a BP (Back Propagation) neural network is selected to construct the question and answer quality assessment model. Considering the high complexity of BP neural network training, a hunter prey optimization algorithm (Hunter Prey Optimization, HPO) is introduced in the embodiment of the present invention to optimize the BP neural network, and the global search capability of HPO is used to tune the parameters of the BP neural network.

[0053] As an implementation method, Figure 4 As shown, in step S20, the parameter combination of the BP neural network is used as the search object of the hunter-prey optimization algorithm, and the fitness of the parameter combination of the BP neural network on the verification set is used as a reference to update the parameter combination of the BP neural network, and iterate a preset number of times to obtain the optimized BP neural network. The steps include:

[0054] S21, for each iteration, using the current parameter combination of the BP neural network as a search object of the hunter-prey optimization algorithm to generate a new parameter combination;

[0055] S22, respectively calculating the fitness of the BP neural network on the training set under a new parameter combination, a current parameter combination, and a historical parameter combination;

[0056] S23, taking the parameter combination with the highest fitness as the optimization parameter combination corresponding to the current iteration;

[0057] S24, selecting the best parameter combination from the optimization parameter combination corresponding to each iteration, and obtaining the optimized BP neural network according to the best parameter combination.

[0058] In the embodiment of the present invention, the number of iterations of the hunter-prey optimization algorithm is preset as a preset number, and the specific value of the preset number can be determined according to actual conditions, and the embodiment of the present invention does not specifically limit this. Taking the current iteration process as an example for explanation, the process of each iteration is the same, and the current iteration process can be referred to.

[0059] It should be noted that before the hunter-prey optimization algorithm starts to iterate, it is necessary to initialize the parameter combination of the BP neural network and the parameters of the hunter-prey optimization algorithm, and set the initial values ​​of the parameter combination and the HPO parameters. The parameter combination includes the number of input layers, the number of hidden layers, the number of output layers of the BP neural network, and the weight of each layer of the neural network; the HPO parameters include the number of hunters, the number of prey, the search space, and the movement strategy.

[0060] After the initialization is completed, in each iteration of the hunter-prey optimization algorithm, the current parameter combination of the BP neural network is used as the search object of the hunter-prey optimization algorithm to generate a new parameter combination.

[0061] As an example, in step S21, the current parameter combination of the BP neural network is used as the search object of the hunter-prey optimization algorithm to generate a new parameter combination, and the steps include:

[0062] According to the current parameter combination of the BP neural network and the fitness of the BP neural network on the verification set, a preset movement strategy is adopted to generate the new parameter combination;

[0063] The preset movement strategy includes random walk or local search.

[0064] Specifically, in the hunter-prey optimization algorithm, each hunter generates a new position (i.e., a new parameter combination) based on the current parameter combination and the prey position using a pre-set movement strategy (such as random walk or local search, etc.); then the prey adjusts its position based on the current parameter combination and the historical optimal parameter combination to try to escape the hunter's tracking and obtain a new parameter combination; then the fitness is calculated to evaluate the fitness of the new parameter combination; the position is updated, and the new parameter combination obtained is used as the hunter's new position. For the prey, the better of its historical optimal parameter combination or the current parameter combination is usually selected.

[0065] Among them, calculating the fitness of each parameter combination refers to using the performance index of the current parameter combination on the validation set. The performance index can be any index that reflects the performance of the BP neural network, which can be determined according to actual conditions, and the embodiment of the present invention does not specifically limit this. For example, the performance index can be the accuracy rate, that is, calculating the accuracy rate of each parameter combination on the training set, and selecting the parameter combination with the highest accuracy rate as the optimization parameter combination corresponding to the current iteration.

[0066] According to this method, the optimization parameter combination of each iteration is calculated, and the best parameter combination is selected from the optimization parameter combination corresponding to each iteration, and the optimized BP neural network is obtained according to the best parameter combination.

[0067] The embodiment of the present invention utilizes the hunter-prey optimization algorithm in combination with the BP neural network to construct a question-answering quality evaluation model. Since the complexity of BP neural network training is relatively high, the global search capability of the hunter-prey optimization algorithm can effectively optimize the parameter combination of the BP neural network, so that the parameter combination of the optimized BP neural network can reach a better state. When the optimized BP neural network is subsequently trained using the training set, the complexity of training can be reduced, thereby improving the prediction accuracy and stability of the trained BP neural network.

[0068] S30: Using the training set to train the optimized BP neural network to obtain a trained BP neural network.

[0069] After the optimized BP neural network is determined, it can be understood that the optimal BP neural network parameter combination is determined, and then the optimized BP neural network is trained using the historical question and answer sheets in the training set to obtain the trained BP neural network, wherein the label of the historical question and answer sheet during the training process is the evaluation level.

[0070] In the embodiment of the present invention, the optimized parameter combination of the BP neural network is determined in advance by the HPO algorithm, so that it can quickly reach a convergence state in the subsequent training process, thereby reducing the time required for training and reducing the complexity of training.

[0071] S40: Input the question and answer form to be tested into the trained BP neural network to obtain an evaluation grade, so as to screen each question and answer form according to the evaluation grade corresponding to each question and answer form of the intelligent customer service system, and use the screened question and answer forms to adjust the intelligent customer service system.

[0072] Finally, the question and answer form to be tested is input into the trained BP neural network, and the evaluation level of the question and answer form to be tested is output, so that through the trained BP neural network, other question and answer forms with unknown evaluation levels can be output, and corresponding to all historical question and answer forms or newly predicted question and answer forms, the question and answer forms are screened according to their evaluation levels, and the screened question and answer forms are used to adjust the intelligent customer service system.

[0073] As an implementation method, Figure 5 As shown, in step S40, each question and answer form is screened according to the evaluation level corresponding to each question and answer form of the intelligent customer service system, and the screened question and answer form is used to adjust the intelligent customer service system, and the steps include:

[0074] S41, according to the evaluation level corresponding to each questionnaire, filter out the questionnaires whose evaluation levels do not meet the requirements, extract features of users in the filtered questionnaires and perform clustering to obtain user groups;

[0075] S42, correcting the answers in the screened question-and-answer sheet, and obtaining a correction library using the corrected question-and-answer sheet;

[0076] S43, receiving the user's consulting question, calculating the similarity between the user and the user group, and if the similarity is greater than a preset similarity threshold, obtaining a corrected answer based on the consulting question and the correction library, so that the intelligent customer service system feeds back the corrected answer to the user.

[0077] In an embodiment of the present invention, according to the evaluation level corresponding to each questionnaire, questionnaires whose evaluation levels do not meet the requirements are screened out. Here, screening out questionnaires whose evaluation levels do not meet the requirements refers to screening out questionnaires with low evaluation levels or low evaluation scores. The lower the score, the worse the quality of the answer to the questionnaire. The specific evaluation level or evaluation score that is considered low can be determined based on actual conditions, and the embodiment of the present invention does not make any specific limitation on this.

[0078] The steps of extracting features and clustering the users in the screened question and answer sheets to obtain a user group include:

[0079] The user group is obtained by using the K-Means algorithm based on the basic information of the users, where the basic information includes the gender, age and occupation of the users.

[0080] In the embodiment of the present invention, the method for extracting features and clustering users in the screened questionnaire can be to first extract basic information of the users from the questionnaire, such as user gender, user age, user occupation, etc. It is easy to understand that customers with the same gender, age and occupation are likely to have the same question-and-answer style. After collecting the basic information of the users, the K-Means algorithm is used to cluster the users to obtain various user groups.

[0081] Then, the answer content in the screened question and answer sheet is corrected to obtain the corrected question and answer sheet and a correction library, which includes the corrected question and answer sheet. As an example, the steps of correcting the answer content in the screened question and answer sheet and obtaining the correction library using the corrected question and answer sheet include:

[0082] Conducting manual telephone interviews with users in the selected questionnaires, and revising the answers based on the interview information;

[0083] The correction library is formed based on the corrected answer content.

[0084] When the screened questionnaire is revised in the embodiment of the present invention, a manual telephone return visit is conducted on the screened questionnaire, and the answer content is revised using the return visit information; and then a revision library is formed based on the revised answer content.

[0085] After obtaining the correction library. When the client receives the user's consultation question, it calculates the similarity between the user and the user group. If the similarity is greater than the preset similarity threshold, it means that the user's answer originally had a low score. Then, according to the consultation question and the correction library, a correction answer is obtained, and the intelligent customer service system feeds back the correction answer to the user.

[0086] In the above-mentioned artificial intelligence-based question and answer quality assessment method, historical question and answer sheets are first collected and split into a training set and a validation set; then the parameter combination of the BP neural network is optimized using the training set and the hunter-prey optimization algorithm, and the fitness of the parameter combination of the BP neural network on the validation set is used as a reference. After multiple iterations, the optimized BP neural network is finally obtained; then the optimized BP neural network is trained using the training set to obtain a trained BP neural network; finally, the question and answer sheet to be tested is input into the trained BP neural network to obtain the evaluation level of the question and answer sheet to be tested, so as to adjust the intelligent customer service system according to the evaluation level.

[0087] The embodiment of the present invention utilizes the hunter-prey optimization algorithm in combination with the BP neural network to construct a question-answering quality evaluation model. Since the complexity of BP neural network training is relatively high, the global search capability of the hunter-prey optimization algorithm can effectively optimize the parameter combination of the BP neural network, so that the parameter combination of the optimized BP neural network can reach a better state. When the optimized BP neural network is subsequently trained using the training set, the complexity of training can be reduced, thereby improving the prediction accuracy and stability of the trained BP neural network.

[0088] In addition, each question and answer form is evaluated through the trained BP neural network, and the question and answer forms are screened according to the evaluated levels. The screened question and answer forms are used to adjust the intelligent customer service system, thereby improving the accuracy of the intelligent customer service system and user satisfaction.

[0089] In one embodiment, a question-answer quality assessment device based on artificial intelligence is provided, and the question-answer quality assessment device based on artificial intelligence corresponds one-to-one to the question-answer quality assessment method based on artificial intelligence in the above embodiment. Figure 6 As shown, the AI-based question-answer quality assessment device includes a collection module 101, a parameter module 102, a training module 103, and an assessment module 104. Each functional module is described in detail as follows:

[0090] A collection module 101 is used to collect historical question-and-answer forms and split the historical question-and-answer forms into a training set and a validation set. The historical question-and-answer forms include the number of processed question-and-answer requests, response time, successful resolution rate, user satisfaction, and chat duration;

[0091] The parameter module 102 is used to use the parameter combination of the BP neural network as the search object of the hunter-prey optimization algorithm, update the parameter combination of the BP neural network with reference to the fitness of the parameter combination of the BP neural network on the verification set, and perform a preset number of iterations to obtain an optimized BP neural network, wherein the parameter combination includes the number of neurons in each layer, weight parameters, bias parameters, learning rate and regularization parameters;

[0092] A training module 103 is used to train the optimized BP neural network using the training set to obtain a trained BP neural network;

[0093] The evaluation module 104 is used to input the question and answer form to be tested into the trained BP neural network to obtain an evaluation grade, so as to screen each question and answer form according to the evaluation grade corresponding to each question and answer form of the intelligent customer service system, and use the screened question and answer forms to adjust the intelligent customer service system.

[0094] In one embodiment, the parameter module 102 is specifically configured to:

[0095] For each iteration, the current parameter combination of the BP neural network is used as a search object of the hunter-prey optimization algorithm to generate a new parameter combination;

[0096] Calculate the fitness of the BP neural network on the training set under the new parameter combination, the current parameter combination and the historical parameter combination respectively;

[0097] The parameter combination with the highest fitness is used as the optimization parameter combination corresponding to the current iteration;

[0098] The best parameter combination is selected from the optimization parameter combination corresponding to each iteration, and the optimized BP neural network is obtained according to the best parameter combination.

[0099] As an implementation mode, the current parameter combination of the BP neural network is used as a search object of the hunter-prey optimization algorithm to generate a new parameter combination, and the steps include:

[0100] According to the current parameter combination of the BP neural network and the fitness of the BP neural network on the verification set, a preset movement strategy is adopted to generate the new parameter combination;

[0101] The preset movement strategy includes random walk or local search.

[0102] In one embodiment, the evaluation module 104 is specifically configured to:

[0103] According to the evaluation level corresponding to each question and answer sheet, the question and answer sheets whose evaluation levels do not meet the requirements are screened out, and the users in the screened question and answer sheets are subjected to feature extraction and clustering to obtain user groups;

[0104] Correcting the answers in the screened question-and-answer sheets, and obtaining a correction library using the corrected question-and-answer sheets;

[0105] Receive the user's consulting question, calculate the similarity between the user and the user group, and if the similarity is greater than a preset similarity threshold, obtain a corrected answer based on the consulting question and the correction library, so that the intelligent customer service system feeds back the corrected answer to the user.

[0106] As an implementation mode, the steps of extracting features and clustering the users in the screened question and answer form to obtain a user group include:

[0107] The user group is obtained by using the K-Means algorithm based on the basic information of the users, where the basic information includes the gender, age and occupation of the users.

[0108] As an implementation mode, the steps of correcting the answers in the screened questionnaire and obtaining a correction library using the corrected questionnaire include:

[0109] Conducting manual telephone interviews with users in the selected questionnaires, and revising the answers based on the interview information;

[0110] The correction library is formed based on the corrected answer content.

[0111] In one embodiment, the acquisition module 101 is specifically used for:

[0112] Obtaining interaction data between the customer service and the user through the log of the intelligent customer service system, wherein the interaction data includes the number of question and answer requests processed in a preset time period, the response time of each question and answer request, the chat duration of each question and answer request, and whether the question and answer request is successfully resolved;

[0113] Collect user satisfaction for each Q&A request through email, SMS, or online surveys;

[0114] The interaction data and the user satisfaction are cleaned, feature extracted, and normalized to obtain the historical question and answer list.

[0115] The present invention provides a question-and-answer quality evaluation device based on artificial intelligence. First, historical question-and-answer sheets are collected, and the historical question-and-answer sheets are split into a training set and a verification set. Then, the parameter combination of a BP neural network is optimized by using the training set and a hunter-prey optimization algorithm, and the fitness of the parameter combination of the BP neural network on the verification set is used as a reference. After multiple iterations, the optimized BP neural network is finally obtained. Then, the optimized BP neural network is trained by using the training set to obtain a trained BP neural network. Finally, the question-and-answer sheet to be tested is input into the trained BP neural network to obtain an evaluation grade of the question-and-answer sheet to be tested, so as to adjust the intelligent customer service system according to the evaluation grade.

[0116] The embodiment of the present invention utilizes the hunter-prey optimization algorithm in combination with the BP neural network to construct a question-answering quality evaluation model. Since the complexity of BP neural network training is relatively high, the global search capability of the hunter-prey optimization algorithm can effectively optimize the parameter combination of the BP neural network, so that the parameter combination of the optimized BP neural network can reach a better state. When the optimized BP neural network is subsequently trained using the training set, the complexity of training can be reduced, thereby improving the prediction accuracy and stability of the trained BP neural network.

[0117] In addition, each question and answer form is evaluated through the trained BP neural network, and the question and answer forms are screened according to the evaluated levels. The screened question and answer forms are used to adjust the intelligent customer service system, thereby improving the accuracy of the intelligent customer service system and user satisfaction.

[0118] For the specific limitations of the question-and-answer quality assessment device based on artificial intelligence, please refer to the limitations of the question-and-answer quality assessment method based on artificial intelligence above, which will not be repeated here. Each module in the above-mentioned question-and-answer quality assessment device based on artificial intelligence can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0119] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the server side of a question and answer quality assessment method based on artificial intelligence.

[0120] In one embodiment, a computer device is provided. The computer device may be a client, and its internal structure diagram may be as follows: Figure 8As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the client side of a question and answer quality assessment method based on artificial intelligence

[0121] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0122] Collect historical Q&A forms and split them into a training set and a validation set. The historical Q&A forms include the number of Q&A requests processed, response time, successful resolution rate, user satisfaction, and chat duration;

[0123] The parameter combination of the BP neural network is used as the search object of the hunter-prey optimization algorithm, and the fitness of the parameter combination of the BP neural network on the verification set is used as a reference to update the parameter combination of the BP neural network, and iterate a preset number of times to obtain an optimized BP neural network, wherein the parameter combination includes the number of neurons in each layer, weight parameters, bias parameters, learning rate and regularization parameters;

[0124] Using the training set to train the optimized BP neural network to obtain a trained BP neural network;

[0125] The question and answer sheet to be tested is input into the trained BP neural network to obtain an evaluation grade, so as to screen each question and answer sheet according to the evaluation grade corresponding to each question and answer sheet of the intelligent customer service system, and use the screened question and answer sheets to adjust the intelligent customer service system.

[0126] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0127] Collect historical Q&A forms and split them into a training set and a validation set. The historical Q&A forms include the number of Q&A requests processed, response time, successful resolution rate, user satisfaction, and chat duration;

[0128] The parameter combination of the BP neural network is used as the search object of the hunter-prey optimization algorithm, and the fitness of the parameter combination of the BP neural network on the verification set is used as a reference to update the parameter combination of the BP neural network, and iterate a preset number of times to obtain an optimized BP neural network, wherein the parameter combination includes the number of neurons in each layer, weight parameters, bias parameters, learning rate and regularization parameters;

[0129] Using the training set to train the optimized BP neural network to obtain a trained BP neural network;

[0130] The question and answer sheet to be tested is input into the trained BP neural network to obtain an evaluation grade, so as to screen each question and answer sheet according to the evaluation grade corresponding to each question and answer sheet of the intelligent customer service system, and use the screened question and answer sheets to adjust the intelligent customer service system.

[0131] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the client side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0132] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0133] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0134] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A question-answering quality assessment method based on artificial intelligence, characterized in that: include: Collect historical Q&A forms and split them into a training set and a validation set. The historical Q&A forms include the number of Q&A requests processed, response time, successful resolution rate, user satisfaction, and chat duration; The parameter combination of the BP neural network is used as the search object of the hunter-prey optimization algorithm, and the fitness of the parameter combination of the BP neural network on the verification set is used as a reference to update the parameter combination of the BP neural network, and iterate a preset number of times to obtain an optimized BP neural network, wherein the parameter combination includes the number of neurons in each layer, weight parameters, bias parameters, learning rate and regularization parameters; Using the training set to train the optimized BP neural network to obtain a trained BP neural network; The question and answer sheet to be tested is input into the trained BP neural network to obtain an evaluation grade, so as to screen each question and answer sheet according to the evaluation grade corresponding to each question and answer sheet of the intelligent customer service system, and use the screened question and answer sheets to adjust the intelligent customer service system.

2. The question-answer quality assessment method based on artificial intelligence as claimed in claim 1, characterized in that: The step of taking the parameter combination of the BP neural network as the search object of the hunter-prey optimization algorithm, taking the fitness of the parameter combination of the BP neural network on the verification set as a reference, updating the parameter combination of the BP neural network, and performing a preset number of iterations to obtain the optimized BP neural network comprises: For each iteration, the current parameter combination of the BP neural network is used as a search object of the hunter-prey optimization algorithm to generate a new parameter combination; Calculate the fitness of the BP neural network on the training set under the new parameter combination, the current parameter combination and the historical parameter combination respectively; The parameter combination with the highest fitness is used as the optimization parameter combination corresponding to the current iteration; The best parameter combination is selected from the optimization parameter combination corresponding to each iteration, and the optimized BP neural network is obtained according to the best parameter combination.

3. The question-answer quality assessment method based on artificial intelligence as claimed in claim 2, characterized in that: The current parameter combination of the BP neural network is used as the search object of the hunter-prey optimization algorithm to generate a new parameter combination, the steps include: According to the current parameter combination of the BP neural network and the fitness of the BP neural network on the verification set, a preset movement strategy is adopted to generate the new parameter combination; The preset movement strategy includes random walk or local search.

4. The question-answer quality assessment method based on artificial intelligence as claimed in claim 1, characterized in that: The steps of screening each question and answer sheet according to the evaluation level corresponding to each question and answer sheet of the intelligent customer service system, and adjusting the intelligent customer service system by using the screened question and answer sheets include: According to the evaluation level corresponding to each question and answer sheet, the question and answer sheets whose evaluation levels do not meet the requirements are screened out, and the users in the screened question and answer sheets are subjected to feature extraction and clustering to obtain a user group; Correcting the answers in the screened question-and-answer sheets, and obtaining a correction library using the corrected question-and-answer sheets; Receive the user's consulting question, calculate the similarity between the user and the user group, and if the similarity is greater than a preset similarity threshold, obtain a corrected answer based on the consulting question and the correction library, so that the intelligent customer service system feeds back the corrected answer to the user.

5. The question-answer quality assessment method based on artificial intelligence as claimed in claim 4, characterized in that: The steps of extracting features and clustering the users in the screened question and answer sheets to obtain a user group include: The user group is obtained by using the K-Means algorithm based on the basic information of the users, where the basic information includes the gender, age and occupation of the users.

6. The question-answer quality assessment method based on artificial intelligence as claimed in claim 4, characterized in that: The steps of correcting the answers in the screened question-and-answer sheet and obtaining a correction library using the corrected question-and-answer sheet include: Conducting manual telephone interviews with users in the selected questionnaires, and revising the answers based on the interview information; The correction library is formed based on the corrected answer content.

7. The question-answer quality assessment method based on artificial intelligence according to any one of claims 1 to 6, characterized in that: The steps of collecting the historical question and answer sheet include: Obtaining interaction data between the customer service and the user through the log of the intelligent customer service system, wherein the interaction data includes the number of question and answer requests processed in a preset time period, the response time of each question and answer request, the chat duration of each question and answer request, and whether the question and answer request is successfully resolved; Collect user satisfaction for each Q&A request through email, SMS, or online surveys; The interaction data and the user satisfaction are cleaned, feature extracted, and normalized to obtain the historical question and answer list.

8. A question-answer quality assessment device based on artificial intelligence, characterized in that: include: A collection module is used to collect historical question-and-answer forms and split the historical question-and-answer forms into a training set and a validation set. The historical question-and-answer forms include the number of question-and-answer requests processed, response time, successful resolution rate, user satisfaction, and chat duration; A parameter module, used to use the parameter combination of the BP neural network as a search object of the hunter-prey optimization algorithm, update the parameter combination of the BP neural network with reference to the fitness of the parameter combination of the BP neural network on the verification set, and perform a preset number of iterations to obtain an optimized BP neural network, wherein the parameter combination includes the number of neurons in each layer, weight parameters, bias parameters, learning rate and regularization parameters; A training module, used to train the optimized BP neural network using the training set to obtain a trained BP neural network; The evaluation module is used to input the question and answer form to be tested into the trained BP neural network to obtain an evaluation grade, so as to screen each question and answer form according to the evaluation grade corresponding to each question and answer form of the intelligent customer service system, and use the screened question and answer forms to adjust the intelligent customer service system.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the question and answer quality assessment method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the question and answer quality assessment method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

Images