Seat verbal skill recommendation method and device, computer equipment and storage medium

Abstract

The embodiment of the invention belongs to the field of financial science and technology, and relates to a seat verbal skill recommendation method and device, computer equipment and a storage medium, and the method comprises the following steps: obtaining customer communication voice information and customer communication text information; performing first emotion analysis on the client communication voice information according to a preset first user emotion analysis model to obtain first emotion information; performing second emotion analysis on the client communication text information according to a preset second user emotion analysis model to obtain second emotion information; performing emotion calculation according to the first emotion information and the second emotion information to obtain emotion analysis information; obtaining product knowledge information, and generating a recommended verbal skill according to the product knowledge information and the sentiment analysis information; and sending the recommended verbal skill to a seat display interface for display. According to the method, accurate and specialized recommendation verbal skills can be effectively generated according to the emotional state of the customer, so that the service satisfaction of the customer is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, specifically to the field of financial technology, and in particular to a method, device, computer equipment and storage medium for recommending agent speech. Background Art

[0002] With the booming insurance industry, financial insurance companies have won wide market recognition and the trust of a large number of users with their diversified product lines, such as auto insurance and non-auto insurance. However, as the user base continues to grow, the challenges faced by junior agents in handling user inquiries and sales promotions are becoming increasingly prominent. In particular, since junior agents often lack in-depth understanding of complex operating processes and have limited training resources, the learning cost is high and it is difficult for them to quickly grasp the essence of the business.

[0003] More importantly, in the actual insurance product sales and Q&A, junior agents often have weak operational capabilities and poor speech organization skills, which make it impossible to accurately capture user needs and effectively convey product value, thus affecting the conversion rate of the list. In order to solve the problem of insufficient sales skills of agents, the traditional method is generally to have the system or professionals write corresponding sales skills for agents to train or practice. However, the sales skills provided by this method are often too rigid. When dealing with customers with various emotions or users with various roles, it is difficult to meet the actual sales and Q&A needs, resulting in poor sales skills. This not only reduces customer satisfaction, but also restricts the further improvement of the company's sales performance, bringing certain obstacles to the development of the insurance industry. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a method, device, computer equipment and storage medium for recommending agent speech to solve the problem of being unable to effectively provide accurate speech to agents.

[0005] In order to solve the above technical problems, the present application embodiment provides a method for recommending agent speech, which adopts the following technical solution:

[0006] Obtain customer communication voice information and customer communication text information;

[0007] Performing a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information;

[0008] Performing a second sentiment analysis on the customer communication text information according to a preset second user sentiment analysis model to obtain second sentiment information;

[0009] Performing sentiment calculation according to the first sentiment information and the second sentiment information to obtain sentiment analysis information;

[0010] Acquire product knowledge information, and generate recommendation words based on the product knowledge information and the sentiment analysis information;

[0011] The recommended words are sent to the agent display interface for display.

[0012] Furthermore, the step of obtaining the customer communication voice information and the customer communication text information specifically includes:

[0013] Obtaining an information extraction request, and extracting the original communication information of the customer from the database according to the information extraction request;

[0014] Performing text conversion on the original communication information of the customer to obtain the original text information of the customer;

[0015] The customer original communication information and the customer original text information are preprocessed to obtain the customer communication voice information and the customer communication text information.

[0016] Furthermore, the step of performing a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain the first emotion information specifically includes:

[0017] Extracting features of the customer's communication voice information to obtain voice intonation features, voice timbre features, and voice expression features;

[0018] Mapping the customer communication voice information to a preset user portrait according to the voice intonation feature, the voice timbre feature, and the voice expression feature to obtain a first user portrait result;

[0019] Acquire the preset first user sentiment analysis model according to the first user portrait result;

[0020] The speech intonation feature, the speech timbre feature, and the speech expression feature are input into the preset first user emotion analysis model to obtain the first emotion information.

[0021] Furthermore, the step of performing a second sentiment analysis on the customer communication text information according to a preset second user sentiment analysis model to obtain second sentiment information specifically includes:

[0022] Performing word embedding processing on the customer communication text information to obtain customer text embedding words;

[0023] Mapping the customer text embedding words to a preset user portrait to obtain a second user portrait result;

[0024] Acquire the preset second user sentiment analysis model according to the second user portrait result;

[0025] Performing vector transformation on the customer text embedding words to obtain a customer text vector;

[0026] The customer text vector is input into the preset second user sentiment analysis model to obtain the second sentiment information.

[0027] Furthermore, the step of performing sentiment calculation according to the first sentiment information and the second sentiment information to obtain sentiment analysis information specifically includes:

[0028] Obtaining a first weight and a second weight corresponding to the first emotion information and the second emotion information;

[0029] Calculating according to the first emotion information and the first weight to obtain a first emotion score;

[0030] Calculate according to the second emotion information and the second weight to obtain a second emotion score;

[0031] Calculating the mean of the first sentiment score and the second sentiment score to obtain the sentiment analysis score;

[0032] The sentiment analysis score is mapped to a preset sentiment analysis table to obtain the sentiment analysis information.

[0033] Furthermore, the step of acquiring product knowledge information and generating recommendation words according to the product knowledge information and the sentiment analysis information specifically includes:

[0034] Extract user demand information from preset user portraits;

[0035] Extracting the product knowledge information from a preset product knowledge graph according to the user demand information;

[0036] Acquire emotional speech text according to the emotional analysis information;

[0037] The product knowledge information and the emotional speech text are combined to obtain the recommended speech.

[0038] Furthermore, after the step of sending the recommended words to the agent display interface for display, the following steps are also included:

[0039] Obtain service feedback data;

[0040] Evaluate the effectiveness of the recommended speech skills according to the service feedback data to obtain a speech skills effectiveness evaluation score;

[0041] Determine whether the evaluation score of the speech effect reaches the preset expected value;

[0042] If the speech effect evaluation score does not reach the preset expected value, the preset first user sentiment analysis model and the preset second user sentiment analysis model are adjusted, and the adjusted sentiment analysis information is re-acquired, and then a new recommended speech is generated based on the adjusted sentiment analysis information, and the service feedback data of the new recommended speech is obtained for effect evaluation to obtain a new speech effect evaluation score, until the new speech effect evaluation score reaches the preset expected value.

[0043] In order to solve the above technical problems, the embodiment of the present application further provides a seat speech recommendation device, which adopts the following technical solution:

[0044] An information acquisition module is used to acquire voice information and text information of customer communication;

[0045] A first emotion analysis module, configured to perform a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information;

[0046] A second sentiment analysis module, used to perform a second sentiment analysis on the customer communication text information according to a preset second user sentiment analysis model to obtain second sentiment information;

[0047] A comprehensive sentiment analysis module, used for performing sentiment calculation according to the first sentiment information and the second sentiment information to obtain sentiment analysis information;

[0048] A speech generation module is used to obtain product knowledge information and generate recommended speech based on the product knowledge information and the sentiment analysis information;

[0049] The speech display module is used to send the recommended speech to the seat display interface for display.

[0050] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0051] A computer device includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the agent speech recommendation method as described in any one of the above items are implemented.

[0052] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0053] A computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the agent speech recommendation method as described in any one of the above items are implemented.

[0054] Compared with the prior art, the embodiment of the present application has the following beneficial effects: the embodiment obtains customer communication voice information and customer communication text information; performs a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information; performs a second emotion analysis on the customer communication text information according to a preset second user emotion analysis model to obtain second emotion information; performs emotion calculation according to the first emotion information and the second emotion information to obtain emotion analysis information; obtains product knowledge information, generates recommended speech according to the product knowledge information and the emotion analysis information; and sends the recommended speech to the agent display interface for display. Thus, accurate and professional recommended speech is effectively generated according to the customer's emotional state to improve the customer's service satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0057] Figure 2 A flowchart of an embodiment of a method for recommending agent speech according to the present application;

[0058] Figure 3 yes Figure 2 A flowchart of a specific implementation of step S10;

[0059] Figure 4 yes Figure 2 A flowchart of a specific implementation of step S20;

[0060] Figure 5 yes Figure 4 A flowchart of a specific implementation of step S30;

[0061] Figure 6 yes Figure 4 A flowchart of a specific implementation of step S40;

[0062] Figure 7 yes Figure 2 A flowchart of a specific implementation of step S50;

[0063] Figure 8 It is a structural schematic diagram of an embodiment of a seat speech recommendation device according to the present application;

[0064] Fig. 9 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0066] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor does it refer to non-related or alternative embodiments that are mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0067] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0068] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0069] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0070] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0071] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0072] It should be noted that the agent speech recommendation method provided in the embodiment of the present application is generally executed by a server, and accordingly, the agent speech recommendation device is generally arranged in the server.

[0073] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0074] Continue to refer Figure 2 , shows a flow chart of an embodiment of a method for recommending agent speech according to the present application. The method for recommending agent speech includes the following steps:

[0075] Step S10, obtaining customer communication voice information and customer communication text information;

[0076] In this embodiment, the customer communication voice information refers to the voice call recording file when the customer communicates with the service agent, and the voice call recording file includes all sound elements such as voice, intonation, speech speed, pauses, laughter, sighs, etc. when the service agent and the customer communicate. The customer communication text information refers to the text field converted from the voice call recording file, and the text field displays the text content of the conversation between the service agent and the customer, including the customer's questions, requests, feedback, etc., as well as the service agent's answers, suggestions, explanations, etc.

[0077] Step S20, performing a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information;

[0078] In this embodiment, the first emotion analysis is based on speech emotion recognition technology, and the preset first user emotion analysis model adopts a data-trained support vector machine (SVM) model. The support vector machine is a supervised learning algorithm that performs classification by finding a hyperplane to maximize the interval between different categories. The training of the preset first user emotion analysis model includes the following steps: data preparation: collecting and labeling speech data sets with emotion labels; feature extraction: extracting features such as intonation, timbre, and expression from speech signals; feature selection: selecting features useful for emotion recognition according to model requirements; model training: using the labeled data set to train the SVM model, and optimizing the classification performance by adjusting parameters.

[0079] Step S30, performing a second sentiment analysis on the customer communication text information according to a preset second user sentiment analysis model to obtain second sentiment information;

[0080] In this embodiment, the preset second user sentiment analysis model adopts a data-trained logistic regression model. Logistic regression is a statistical method widely used for classification problems. The output of linear regression is mapped to the (0,1) interval through the sigmoid function, thereby realizing the prediction of sentiment categories. The training of the preset second user sentiment analysis model includes the following steps: Data preprocessing: Obtain the customer's historical communication text, and perform text cleaning, word segmentation, and stop word removal on the customer's historical communication text. Feature extraction: Convert the preprocessed customer historical communication text into numerical features to obtain customer communication features. Model training: Use the training data set labeled with customer communication features to train the logistic regression model, and optimize the model performance by adjusting parameters (such as regularization terms, learning rate, etc.). Model evaluation: Use the test data set to evaluate the model's accuracy, recall rate, F1 score and other indicators.

[0081] Step S40, performing sentiment calculation according to the first sentiment information and the second sentiment information to obtain sentiment analysis information;

[0082] In this embodiment, sentiment analysis information refers to analysis information that represents the specific emotions of the customer, and corresponding sentiment levels can be set to represent it, such as mild dissatisfaction, general dissatisfaction, extreme dissatisfaction, etc. The first sentiment information corresponds to the first weight, and the second sentiment information corresponds to the second weight. The first weight and the second weight can use a prior learning method to dynamically adjust the weight according to new data input, or adjust the weight according to the result feedback data obtained after the output result. In specific implementation, the rationality of the weight can be verified by a cross-validation method, such as dividing the data set into multiple subsets, using different weight configurations for training and testing, comparing the performance differences under different configurations, and then judging whether the weight is reasonable by the gap between the performance difference and the expected result. At the same time, the sentiment analysis effects under different weight configurations can be compared through A / B testing, and the configuration with better performance can be selected as the final solution. The first weight and the second weight in the configuration are the first weight and the second weight used for the above-mentioned sentiment calculation.

[0083] Step S50, obtaining product knowledge information, and generating recommendation words according to the product knowledge information and the sentiment analysis information;

[0084] In this embodiment, product knowledge information is knowledge information about insurance products, specifically including an overview of insurance products: the name and type of insurance products (such as life insurance, health insurance, property insurance, accidental injury insurance, etc.); the basic coverage of insurance products and target customer groups. Insurance terms and coverage content: the main terms of the insurance contract, including insurance liability, exemption clauses, waiting period, hesitation period, etc.; specific coverage content, such as death compensation, major disease protection, medical expense reimbursement, property damage compensation, etc.; key parameters such as insurance amount, compensation ratio, and compensation limit. Insurance costs and payment methods: the premium calculation method of insurance products, which may be based on factors such as age, gender, health status, and insured amount; payment period and method, such as one-time payment, installment payment (annual payment, semi-annual payment, quarterly payment, monthly payment, etc.); premium adjustment mechanism, such as adjustment based on market interest rate or inflation rate. Insurance term and renewal regulations: the validity period of the insurance contract, including the start date and end date; renewal conditions and procedures, including premium adjustments and health disclosure requirements at the time of renewal; contract termination situations, such as expiration of the insurance term, completion of compensation, termination of the contract, etc. Claims process and requirements: how customers should report the case and submit claims materials after the insurance accident occurs; the process and time requirements for claims review; the payment method and time of claims. Value-added services and additional protection: value-added services that insurance products may provide, such as health management, emergency rescue, legal consultation, etc.; optional additional protection items, such as accidental injury medical treatment, premium exemption, etc. Laws, regulations and regulatory requirements: laws, regulations and regulatory requirements that insurance products must comply with, such as the Insurance Law, the Consumer Rights Protection Law, etc.; regulatory regulations and guidelines that need to be followed during the design and sales of insurance products. Insurance notices and risk warnings: important information that customers need to know before taking out insurance, such as health disclosure requirements, occupational category restrictions, etc.; potential risks and uncertainties of insurance products, such as the scope of application of exemption clauses and possible restrictions on compensation, etc.

[0085] Step S60: Send the recommended words to the agent display interface for display.

[0086] In this embodiment, the agent display interface refers to the display interface on the agent terminal corresponding to the service agent. After the recommended speech is generated, the system sends it to the corresponding agent display interface for display according to the agent ID, thereby helping the service agent to provide effective question-and-answer services to customers. In this embodiment, the above process of generating recommended speech and displaying the recommended speech on the agent display interface is carried out in real time. When the recommended speech is generated for the first time, it can be generated based on the historical communication voice information and historical communication text information between the service agent and the current customer. If the current customer does not have the corresponding historical communication voice information and historical communication text information, the preset speech template for the corresponding classified customer can be obtained based on the customer portrait to which the current customer belongs. The preset speech template is the recommended speech that has been generated for the classified customer.

[0087] This embodiment obtains customer communication voice information and customer communication text information; performs a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information; performs a second emotion analysis on the customer communication text information according to a preset second user emotion analysis model to obtain second emotion information; performs emotion calculation according to the first emotion information and the second emotion information to obtain emotion analysis information; obtains product knowledge information, generates recommended words according to the product knowledge information and the emotion analysis information; and sends the recommended words to the agent display interface for display. In this way, accurate and professional recommended words are effectively generated according to the customer's emotional state to improve the customer's service satisfaction.

[0088] Continue to refer Figure 3 In some optional implementations of this embodiment, step S10 includes the following steps:

[0089] Step S101, obtaining an information extraction request, and extracting the original communication information of the customer from the database according to the information extraction request;

[0090] In this embodiment, the customer's original communication information refers to the real-time recorded voice communication data between the service agent and the customer. The information extraction request is an information extraction identifier corresponding to the customer's original communication information in the database. The information extraction identifier can be an identifier of the information number corresponding to the customer's original communication information. The customer's original communication information is effectively obtained from the database through the information extraction request.

[0091] Step S102, converting the original communication information of the customer into text to obtain the original text information of the customer;

[0092] In this embodiment, the original communication information of the customer is subjected to noise reduction processing to remove interference factors such as background noise and echo, thereby improving the voice quality. If the voice quality after noise reduction processing is still not high, voice enhancement processing can be performed to further improve the voice quality of the original communication information of the customer. In this embodiment, the text conversion of the original communication information of the customer is realized based on the acoustic model of deep learning. By extracting features from the voice signal of the original communication information of the customer, the sound signal parameters reflecting the essential characteristics of the sound are extracted, and then according to the relevant theories of linguistics, the probability of the sound signal parameters corresponding to the possible phrase sequence is calculated, and the phrase sequence with high probability is used as the final output result to complete the conversion of the original communication information of the customer into the original text information of the customer.

[0093] Step S103, pre-processing the customer original communication information and the customer original text information to obtain the customer communication voice information and the customer communication text information.

[0094] In this embodiment, the preprocessing of the customer's original communication information includes noise reduction processing, voice enhancement, etc. After noise reduction processing and voice enhancement processing, clear customer communication voice information is obtained; the preprocessing of the customer's original text information includes text cleaning and text correction. Text cleaning can effectively remove noise (such as garbled characters, special symbols, etc.) in the customer's original text information, and unify the text format (such as uppercase and lowercase letters, punctuation marks, etc.). Text correction can correct errors generated in text fields (wrong words, wrong pronunciations, etc.).

[0095] This embodiment obtains an information extraction request, extracts the customer's original communication information from a database according to the information extraction request; performs text conversion on the customer's original communication information to obtain the customer's original text information; and pre-processes the customer's original communication information and the customer's original text information to obtain customer communication voice information and customer communication text information that can facilitate subsequent first sentiment analysis and second sentiment analysis.

[0096] Continue to refer Figure 4 In some optional implementations of this embodiment, step S20 includes the following steps:

[0097] Step S201, extracting features of the customer communication voice information to obtain voice intonation features, voice timbre features, and voice expression features;

[0098] In this embodiment, the feature extraction of the voice information of customer communication can be completed by STFT (Short Time Fourier Transform). By analyzing the frequency domain characteristics of the voice signal, the voice signal is decomposed into multiple frequency bands, so as to extract the voice intonation features, voice timbre features, and voice expression features in the voice information of customer communication. Among them, the voice intonation features include the change of fundamental frequency (F0), pitch contour, speaking speed, etc., which can reflect the emotional state of the speaker, such as the pitch rises when happy, the pitch falls when sad, and the speaking speed increases when angry. The voice timbre features include the sound quality and resonance peak of the sound, etc., which can reflect the emotional intensity and personality characteristics of the speaker. The voice expression features include pauses, stresses, filler words (such as "um", "ah"), etc. in the voice, which can reveal the speaker's thinking process, emotional fluctuations and attitude changes.

[0099] Step S202, mapping the customer communication voice information to a preset user portrait according to the voice intonation feature, the voice timbre feature, and the voice expression feature, to obtain a first user portrait result;

[0100] In this embodiment, the speech intonation features, speech timbre features, and speech expression features are quantified to obtain corresponding feature values, wherein the quantification of the speech intonation features includes: extracting the fundamental frequency contour, intonation change rate, etc., and quantizing them into specific values ​​or sequences; the quantification of the speech timbre features includes: extracting the spectrum shape, resonance peak frequency, spectrum centroid, etc., and also quantizing them into values ​​or feature vectors; the quantification of the speech expression features includes: analyzing energy changes, stress positions, pause patterns, etc., and quantizing them into expression feature vectors. In this embodiment, the preset user portrait is a user portrait constructed based on the extracted user features. By defining the dimensions of the user portrait, it is clear what aspects of information the user portrait contains, such as basic attributes, psychological state, behavior patterns, etc. Then set a specific label and corresponding threshold or range for each defined dimension so that the feature can be mapped to the corresponding label. After obtaining the feature values ​​and setting specific labels for the user portrait, each feature is mapped to the corresponding user portrait dimension according to the correspondence between the feature values ​​and the user portrait labels. The feature mapping results of multiple dimensions are then combined, and through logical reasoning or machine learning models, the user portrait label that the customer corresponding to the customer communication voice information is most likely to be determined, and the user portrait corresponding to the user portrait performance is used as the first user portrait result.

[0101] Step S203, obtaining the preset first user sentiment analysis model according to the first user portrait result;

[0102] In this embodiment, model matching is performed based on the classification information identifier corresponding to the first user portrait result to extract the corresponding first user sentiment analysis model. In this embodiment, multiple different first user sentiment analysis models are trained for different categories of user portraits, and the different first user sentiment analysis models are stored in a pre-built first sentiment analysis model library.

[0103] Step S204: input the voice intonation feature, the voice timbre feature, and the voice expression feature into the preset first user emotion analysis model to obtain the first emotion information.

[0104] In this embodiment, the preset first user emotion analysis model adopts a support vector machine (SVM) model trained with data. The speech intonation features, speech timbre features, and speech expression features are standardized and normalized, and then input into the support vector machine (SVM) model trained with data, thereby outputting the first emotion information representing different emotion expression results.

[0105] This embodiment extracts features from the customer communication voice information to obtain speech intonation features, speech timbre features, and speech expression features; maps the customer communication voice information to a preset user portrait based on the speech intonation features, speech timbre features, and speech expression features to obtain a first user portrait result; obtains the preset first user emotion analysis model based on the first user portrait result; inputs the speech intonation features, speech timbre features, and speech expression features into the preset first user emotion analysis model, thereby obtaining first emotion information after emotion analysis based on the speech intonation features, speech timbre features, and speech expression features, so as to facilitate subsequent emotion calculation based on the first emotion information.

[0106] Continue to refer Figure 5 In some optional implementations of this embodiment, step S30 includes the following steps:

[0107] Step S301, performing word embedding processing on the customer communication text information to obtain customer text embedding words;

[0108] In this embodiment, word embedding processing can be completed by a pre-trained word embedding model. The pre-trained word embedding model can adopt a pre-trained BERT model. The pre-trained BERT model is used to perform word embedding processing on customer communication text information, and the customer communication text information is divided into multiple sub-words. The sub-words are then encoded to obtain the embedding representations corresponding to the sub-words, and the embedding representations corresponding to the sub-words are then extracted to obtain customer text embedded words.

[0109] Step S302, mapping the customer text embedded words to a preset user portrait to obtain a second user portrait result;

[0110] In this embodiment, by defining the attributes of the preset user portrait, it is clear which key attributes the user portrait contains, such as age, gender, interests, purchasing power, consumption habits, brand preferences, etc., and then designing the structure and hierarchy of the user portrait according to business needs to ensure that it can fully and accurately reflect the user characteristics, thereby obtaining a user portrait with attributes and structure. The customer text embedded words are classified into the user portrait with attributes and structure through pre-defined classification rules. The pre-defined classification rules are the correspondence between words and user portrait categories set according to different attributes and structures of the user portrait. Through the above classification operation, the mapping process of the customer text embedded words in the preset user portrait is completed, and the user portrait category corresponding to the customer text embedded words is effectively obtained, and the user portrait category is used as the second user portrait result.

[0111] Step S303, obtaining the preset second user sentiment analysis model according to the second user portrait result;

[0112] In this embodiment, model matching is performed based on the classification information identifier corresponding to the second user portrait result to extract the corresponding second user sentiment analysis model. In this embodiment, multiple different second user sentiment analysis models are trained for different categories of user portraits, and the different second user sentiment analysis models are stored in a pre-built second sentiment analysis model library.

[0113] Step S304, performing vector transformation on the customer text embedded words to obtain a customer text vector;

[0114] In this embodiment, by aggregating the customer text embedded words, a set vector that can represent the entire text is obtained, and the set vector is the customer text vector. The above aggregation method may include average pooling, weighted average, etc. This embodiment uses weighted average to complete the vector transformation, and the specific steps include: obtaining the weights corresponding to the customer text embedded words; calculating the sum of the customer text embedded words; dividing the sum by the sum of the weights to obtain a weighted average vector; and using the weighted average vector as the customer text vector.

[0115] Step S305: input the customer text vector into the preset second user sentiment analysis model to obtain the second sentiment information.

[0116] In this embodiment, the preset second user emotion analysis model adopts a data-trained logistic regression model, which performs a linear transformation on the input customer text vector and applies the sigmoid function to map the result to the interval (0,1), thereby obtaining a probability value representing the emotional tendency. Then, according to the output probability value, a threshold value (initial setting is 0.2, 0.4, 0.6, 0.8) is set to classify the text into positive emotions (calm, joy, etc.) or negative emotions (anger, excitement, disappointment, etc.). For example, if the output probability value is 0.1 (less than 0.2), the second emotional information is joy, if the output probability value is 0.3 (0.2-0.4), the second emotional information is calm, if the output probability value is 0.5 (0.4-0.6), the second emotional information is excitement, if the output probability value is 0.7 (0.6-0.8), the second emotional information is disappointment, and if the output probability value is 0.9 (greater than 0.8), the second emotional information is anger. The higher the above probability value is, the more negative the corresponding emotion is. The above emotions and the probability value ranges corresponding to the emotions can be adjusted accordingly according to actual conditions.

[0117] This embodiment obtains customer text embedded words by performing word embedding processing on the customer communication text information; maps the customer text embedded words to a preset user portrait to obtain a second user portrait result; obtains the preset second user sentiment analysis model based on the second user portrait result; performs vector transformation on the customer text embedded words to obtain a customer text vector; inputs the customer text vector into the preset second user sentiment analysis model, thereby obtaining second sentiment information for performing specific sentiment analysis based on the customer text embedded words, so as to facilitate subsequent sentiment calculation based on the second sentiment information.

[0118] Continue to refer Figure 6 In some optional implementations of this embodiment, step S40 includes the following steps:

[0119] Step S401, obtaining a first weight and a second weight corresponding to the first emotion information and the second emotion information;

[0120] In this embodiment, the first weight is extracted from the database by the first information identifier corresponding to the first emotional information, and the second weight is extracted from the database by the second information identifier corresponding to the second emotional information.

[0121] Step S402, performing calculation according to the first emotion information and the first weight to obtain a first emotion score;

[0122] In this embodiment, the first emotional information is quantified to obtain a first emotional value. The quantification of the first emotional information can be performed according to the predefined emotional tendency (positive, negative or neutral) and intensity of the words in the preset emotional dictionary, and then the first emotional score is calculated based on the first emotional value and the first weight. The calculation formula of the first emotional score is: first emotional score = first emotional value * first weight. In specific implementation, if there are multiple first emotional information, the obtained first emotional values ​​are also multiple. In this case, the first emotional score should take the average of the multiple first emotional values ​​calculated.

[0123] Step S403, performing calculation according to the second emotion information and the second weight to obtain a second emotion score;

[0124] In this embodiment, the second emotional information is quantified to obtain the second emotional value. The quantification of the second emotional information can be performed according to the predefined emotional tendency (positive, negative or neutral) and intensity of the words in the preset emotional dictionary, and then the second emotional score is calculated based on the second emotional value and the second weight. The calculation formula of the second emotional score is: second emotional score = second emotional value * second weight. In the specific implementation, if there are multiple second emotional information, the obtained second emotional values ​​are also multiple. In this case, the second emotional score should take the average of the multiple second emotional values ​​calculated.

[0125] Step S404, calculating the mean of the first sentiment score and the second sentiment score to obtain the sentiment analysis score;

[0126] In this embodiment, the calculation formula of the sentiment analysis score is: sentiment analysis score=first sentiment score*second sentiment score / 2. The above formula is used to calculate the average of the first sentiment score and the second sentiment score to obtain the sentiment analysis score.

[0127] Step S405: Map the sentiment analysis score to a preset sentiment analysis table to obtain the sentiment analysis information.

[0128] In this embodiment, the preset sentiment analysis table defines sentiment classifications, descriptions, and corresponding score ranges, wherein the sentiment classifications include calmness, joy, anger, excitement, disappointment, etc. By defining the corresponding descriptions in the sentiment classifications, it helps in the subsequent generation of recommendation scripts. By defining the corresponding score ranges in the sentiment classifications, it is convenient to determine the sentiment classifications corresponding to the sentiment analysis scores according to the mapping conditions. After determining the sentiment classifications corresponding to the sentiment analysis scores in the preset sentiment analysis table, the corresponding sentiment classifications are output as sentiment analysis information.

[0129] This embodiment obtains the first weight and the second weight corresponding to the first emotional information and the second emotional information; calculates according to the first emotional information and the first weight to obtain a first emotional score; calculates according to the second emotional information and the second weight to obtain a second emotional score; calculates the mean of the first emotional score and the second emotional score to obtain the emotional analysis score; maps the emotional analysis score to a preset emotional analysis table, thereby obtaining emotional analysis information that performs comprehensive emotional analysis by integrating the first emotional information and the second emotional information, so as to facilitate the subsequent generation of recommendation words based on the emotional analysis information.

[0130] Continue to refer Figure 7 In some optional implementations of this embodiment, step S50 includes the following steps:

[0131] Step S501, extracting user demand information from a preset user portrait;

[0132] In this embodiment, the preset user portrait includes multiple user portrait categories. The user portrait category refers to the category corresponding to the user in the portrait. The user portrait category contains user demand information. The user demand information includes the user's income situation, eligible insurance types, insurance awareness and needs, risk tolerance, service preferences, purchase motivations, etc.

[0133] Step S502, extracting the product knowledge information from a preset product knowledge graph according to the user demand information;

[0134] In this embodiment, the user demand information is mapped to preset insurance product characteristics or functions to determine the insurance product label corresponding to the user demand information, and the insurance product label is used to match and extract the corresponding product knowledge information in the preset product knowledge graph.

[0135] Step S503, obtaining emotional speech text according to the emotional analysis information;

[0136] In this embodiment, the corresponding emotional speech text is obtained from the speech text database through the sentiment analysis information identifier of the sentiment analysis information. The emotional speech text is a preset text template, which can be manually edited and generated according to actual conditions and saved to the speech text database.

[0137] Step S504, combining the product knowledge information and the emotional speech text to obtain the recommended speech.

[0138] In this embodiment, core information such as key selling points, advantages, and special features are extracted from the product knowledge information as the product knowledge basis, and then the product knowledge basis and the emotional speech text are combined in a logical order, and combined and generated from the aspects of logical order, language style, and personalized adjustment to obtain recommended speech.

[0139] This embodiment extracts user demand information from a preset user portrait; extracts the product knowledge information from a preset product knowledge graph based on the user demand information; obtains an emotional speech text based on the sentiment analysis information; and combines the product knowledge information and the emotional speech text to obtain a recommended speech that integrates product knowledge and user demand, so as to provide effective speech to service agents.

[0140] In some optional implementations of this embodiment, step S60 includes the following steps:

[0141] Obtain service feedback data;

[0142] In this embodiment, the service feedback data is the insurance product promotion result information and customer feedback information collected after the recommendation words are sent to the agent display interface for display and the service agent and the customer communicate according to the recommendation words.

[0143] Evaluate the effectiveness of the recommended speech skills according to the service feedback data to obtain a speech skills effectiveness evaluation score;

[0144] In this embodiment, the service feedback data is quantitatively analyzed and qualitatively analyzed by the defined evaluation criteria to calculate the index score corresponding to the evaluation criteria, and the index score is used as the evaluation score of the speech effect, so as to achieve the evaluation of the effect of the recommended speech. Among them, quantitative analysis refers to the quantitative processing of product promotion result information and customer feedback data, and the calculation of the score or proportion of each index; qualitative analysis refers to the qualitative analysis of the subjective description in the customer feedback data, and the identification of the positive and negative evaluation of the customer's speech, as well as improvement suggestions. The indicators corresponding to the evaluation criteria include: success rate: the proportion of the number of successful sales to the total number of attempts, reflecting the effect of the speech in facilitating transactions. Customer satisfaction: the customer's evaluation of the content of the speech, service attitude, professionalism, etc., which can be measured by scoring or rating. Speech understanding: the degree of customer understanding of the content of the speech, which can be indirectly evaluated by the frequency of customer questions and the relevance of the questions. Trust improvement: whether the customer has a higher sense of trust in the product or service after communication can be observed through the customer's subsequent behavior (such as actively asking for more information, expressing willingness to buy, etc.).

[0145] Determine whether the evaluation score of the speech effect reaches the preset expected value;

[0146] In this embodiment, by comparing the speech effect evaluation score with the preset expected value, when the speech effect evaluation score is greater than or equal to the preset expected value, it means that the speech effect evaluation score reaches the preset expected value.

[0147] If the speech effect evaluation score does not reach the preset expected value, the preset first user sentiment analysis model and the preset second user sentiment analysis model are adjusted, and the adjusted sentiment analysis information is re-acquired, and then a new recommended speech is generated based on the adjusted sentiment analysis information, and the service feedback data of the new recommended speech is obtained for effect evaluation to obtain a new speech effect evaluation score, until the new speech effect evaluation score reaches the preset expected value.

[0148] This embodiment obtains service feedback data; evaluates the effect of the recommended speech according to the service feedback data to obtain a speech effect evaluation score; determines whether the speech effect evaluation score reaches the preset expected value; if the speech effect evaluation score does not reach the preset expected value, the preset first user sentiment analysis model and the preset second user sentiment analysis model are adjusted, the adjusted sentiment analysis information is re-acquired, and a new recommended speech is generated according to the adjusted sentiment analysis information, and the service feedback data of the new recommended speech is obtained for effect evaluation to obtain a new speech effect evaluation score, until the new speech effect evaluation score reaches the preset expected value. In this way, the first user sentiment analysis model and the second user sentiment analysis model are effectively adjusted and optimized according to the service feedback data collected after the implementation of the recommended speech, so as to obtain a better speech recommendation effect.

[0149] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0150] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0151] Further references Figure 8 , as a response to the above Figure 1 The present application provides an embodiment of a seat speech recommendation device, and the device embodiment is similar to Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0152] like Figure 8As shown, the seat speech recommendation device 700 described in this embodiment includes: an information acquisition module 701, a first sentiment analysis module 702, a second sentiment analysis module 703, a comprehensive sentiment analysis module 704, a speech generation module 705, and a speech display module 706. Among them:

[0153] The information acquisition module 701 is used to acquire the voice information and text information of customer communication;

[0154] A first emotion analysis module 702, configured to perform a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information;

[0155] A second sentiment analysis module 703, configured to perform a second sentiment analysis on the customer communication text information according to a preset second user sentiment analysis model to obtain second sentiment information;

[0156] A comprehensive sentiment analysis module 704, configured to perform sentiment calculation based on the first sentiment information and the second sentiment information to obtain sentiment analysis information;

[0157] The speech generation module 705 is used to obtain product knowledge information and generate recommended speech according to the product knowledge information and the sentiment analysis information;

[0158] The speech display module 706 is used to send the recommended speech to the seat display interface for display.

[0159] By adopting the above-mentioned agent speech recommendation device, this embodiment can obtain customer communication voice information and customer communication text information; perform a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information; perform a second emotion analysis on the customer communication text information according to a preset second user emotion analysis model to obtain second emotion information; perform emotion calculation according to the first emotion information and the second emotion information to obtain emotion analysis information; obtain product knowledge information, generate recommended speech according to the product knowledge information and the emotion analysis information; and send the recommended speech to the agent display interface for display. In this way, accurate and professional recommended speech is effectively generated according to the customer's emotional state to improve the customer's service satisfaction.

[0160] To solve the above technical problems, the present application also provides a computer device. Fig. 9 , Fig. 9 This is a basic structural block diagram of the computer device in this embodiment.

[0161] The computer device 8 includes a memory 81, a processor 82, and a network interface 83 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 8 with components 81-83, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0162] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0163] The memory 81 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 81 can be an internal storage unit of the computer device 8, such as a hard disk or memory of the computer device 8. In other embodiments, the memory 81 can also be an external storage device of the computer device 8, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 8. Of course, the memory 81 can also include both the internal storage unit of the computer device 8 and its external storage device. In this embodiment, the memory 81 is generally used to store the operating system and various application software installed on the computer device 8, such as computer-readable instructions of the agent speech recommendation method, etc. In addition, the memory 81 can also be used to temporarily store various types of data that have been output or are to be output.

[0164] The processor 82 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 82 is generally used to control the overall operation of the computer device 8. In this embodiment, the processor 82 is used to run computer-readable instructions stored in the memory 81 or process data, such as computer-readable instructions for running the agent speech recommendation method.

[0165] The network interface 83 may include a wireless network interface or a wired network interface. The network interface 83 is generally used to establish a communication connection between the computer device 8 and other electronic devices.

[0166] By adopting the above-mentioned computer device, this embodiment can obtain customer communication voice information and customer communication text information; perform a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information; perform a second emotion analysis on the customer communication text information according to a preset second user emotion analysis model to obtain second emotion information; perform emotion calculation according to the first emotion information and the second emotion information to obtain emotion analysis information; obtain product knowledge information, generate recommendation words according to the product knowledge information and the emotion analysis information; and send the recommendation words to the agent display interface for display. In this way, accurate and professional recommendation words are effectively generated according to the customer's emotional state to improve the customer's service satisfaction.

[0167] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the agent speech recommendation method as described above.

[0168] By adopting the above-mentioned computer-readable storage medium, this embodiment can obtain customer communication voice information and customer communication text information; perform a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information; perform a second emotion analysis on the customer communication text information according to a preset second user emotion analysis model to obtain second emotion information; perform emotion calculation according to the first emotion information and the second emotion information to obtain emotion analysis information; obtain product knowledge information, generate recommended words according to the product knowledge information and the emotion analysis information; and send the recommended words to the agent display interface for display. In this way, accurate and professional recommended words are effectively generated according to the customer's emotional state to improve the customer's service satisfaction.

[0169] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0170] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for recommending agent speech, characterized in that: The steps include: Obtain customer communication voice information and customer communication text information; Performing a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information; Performing a second sentiment analysis on the customer communication text information according to a preset second user sentiment analysis model to obtain second sentiment information; Performing sentiment calculation according to the first sentiment information and the second sentiment information to obtain sentiment analysis information; Acquire product knowledge information, and generate recommendation words based on the product knowledge information and the sentiment analysis information; The recommended words are sent to the agent display interface for display.

2. The method for recommending agent speech skills according to claim 1, characterized in that: The step of obtaining the customer communication voice information and the customer communication text information specifically includes: Obtaining an information extraction request, and extracting the original communication information of the customer from the database according to the information extraction request; Performing text conversion on the original communication information of the customer to obtain the original text information of the customer; The customer original communication information and the customer original text information are preprocessed to obtain the customer communication voice information and the customer communication text information.

3. The method for recommending agent speech skills according to claim 1, characterized in that: The step of performing a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain the first emotion information specifically includes: Extracting features of the customer's communication voice information to obtain voice intonation features, voice timbre features, and voice expression features; Mapping the customer communication voice information to a preset user portrait according to the voice intonation feature, the voice timbre feature, and the voice expression feature to obtain a first user portrait result; Acquire the preset first user sentiment analysis model according to the first user portrait result; The speech intonation feature, the speech timbre feature, and the speech expression feature are input into the preset first user emotion analysis model to obtain the first emotion information.

4. The method for recommending agent speech skills according to claim 1, characterized in that: The step of performing a second sentiment analysis on the customer communication text information according to a preset second user sentiment analysis model to obtain the second sentiment information specifically includes: Performing word embedding processing on the customer communication text information to obtain customer text embedding words; Mapping the customer text embedding words to a preset user portrait to obtain a second user portrait result; Acquire the preset second user sentiment analysis model according to the second user portrait result; Performing vector transformation on the customer text embedding words to obtain a customer text vector; The customer text vector is input into the preset second user sentiment analysis model to obtain the second sentiment information.

5. The method for recommending agent speech according to claim 1, characterized in that: The step of performing sentiment calculation according to the first sentiment information and the second sentiment information to obtain sentiment analysis information specifically includes: Obtaining a first weight and a second weight corresponding to the first emotion information and the second emotion information; Calculating according to the first emotion information and the first weight to obtain a first emotion score; Calculate according to the second emotion information and the second weight to obtain a second emotion score; Calculating the mean of the first sentiment score and the second sentiment score to obtain the sentiment analysis score; The sentiment analysis score is mapped to a preset sentiment analysis table to obtain the sentiment analysis information.

6. The method for recommending agent speech according to claim 1, characterized in that: The step of acquiring product knowledge information and generating recommendation words according to the product knowledge information and the sentiment analysis information specifically includes: Extract user demand information from preset user portraits; Extracting the product knowledge information from a preset product knowledge graph according to the user demand information; Acquire emotional speech text according to the emotional analysis information; The product knowledge information and the emotional speech text are combined to obtain the recommended speech.

7. The method for recommending agent speech skills according to claim 1, characterized in that: After the step of sending the recommended words to the agent display interface for display, the method further includes the following steps: Obtain service feedback data; Evaluate the effectiveness of the recommended speech skills according to the service feedback data to obtain a speech skills effectiveness evaluation score; Determine whether the evaluation score of the speech effect reaches the preset expected value; If the speech effect evaluation score does not reach the preset expected value, the preset first user sentiment analysis model and the preset second user sentiment analysis model are adjusted, and the adjusted sentiment analysis information is re-acquired, and then a new recommended speech is generated based on the adjusted sentiment analysis information, and the service feedback data of the new recommended speech is obtained for effect evaluation to obtain a new speech effect evaluation score, until the new speech effect evaluation score reaches the preset expected value.

8. A device for recommending agent speech, characterized in that: include: An information acquisition module is used to acquire voice information and text information of customer communication; A first emotion analysis module, configured to perform a first emotion analysis on the customer communication voice information according to a preset first user emotion analysis model to obtain first emotion information; A second sentiment analysis module, used to perform a second sentiment analysis on the customer communication text information according to a preset second user sentiment analysis model to obtain second sentiment information; A comprehensive sentiment analysis module, used for performing sentiment calculation according to the first sentiment information and the second sentiment information to obtain sentiment analysis information; A speech generation module, used to obtain product knowledge information and generate recommended speech based on the product knowledge information and the sentiment analysis information; The speech display module is used to send the recommended speech to the seat display interface for display.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the agent speech recommendation method as described in any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the agent speech recommendation method according to any one of claims 1 to 7 are implemented.

Images