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

Abstract

The invention relates to the technical field of voice semantic processing, is suitable for the financial field and the medical health field, and provides a seat verbal skill recommendation method and device, equipment and a storage medium. Inputting the target text data into a preset intention analysis model for analysis so as to determine target feature data used for representing the intention of the user; determining a target coping strategy for coping with the intention of the user based on a preset mapping rule table and the target feature data; jointly inputting the target text data and the target coping strategy into a preset verbal skill generation model for processing to generate a target verbal skill text suitable for answering the user; and sending the target verbal skill text to the seat, so that the seat performs voice interaction with the user according to the target verbal skill text. The invention aims to improve the matching degree between the seat verbal skill and the user intention, so that the accuracy of seat verbal skill recommendation is improved.

Description

Technical Field

[0001] This application relates to the field of speech and semantic processing technology, and in particular to a method, apparatus, device and storage medium for recommending agent dialogue scripts. Background Technology

[0002] In the business sector, many companies use call centers to handle sales, inquiries, and complaints. For example, in the financial sector, banks and insurance companies use call centers to provide customers with services such as account inquiries, business processing, and risk warnings. Similarly, in the healthcare sector, hospitals and health management institutions use call centers to provide patients with convenient services such as appointment booking, health consultations, and report interpretation. In the traditional model, call center agents make and answer calls to provide services. While this can meet some basic user needs, the accuracy of responses is often affected by the agent's professional skills, communication abilities, and personal experience.

[0003] With the development of artificial intelligence technology, more and more companies are introducing AI-assisted systems to improve the efficiency and quality of their agents' services. However, most current AI-assisted systems still rely on keyword extraction as their core method for user intent recognition. Meanwhile, users' speech patterns are often diverse, and factors such as ambiguity in language, the complexity of context, and users' individual language habits mean that the scripts generated by AI-assisted systems often do not match the user's actual intent. Summary of the Invention

[0004] The main purpose of this application is to provide a method, apparatus, device and storage medium for recommending agent scripts, which aims to improve the matching degree between agent scripts and user intentions, thereby improving the accuracy of agent script recommendations.

[0005] Firstly, this application provides a method for recommending agent scripts, including: Acquire target text data, which is obtained by text conversion based on voice data from agents and users; The target text data is input into a preset intention analysis model for analysis in order to determine the target feature data used to characterize the user's intention; Based on a preset mapping rule table and the target feature data, a target response strategy for responding to the user's intention is determined; wherein, the mapping rule table records the mapping relationship between multiple target feature data and multiple response strategies; The target text data and the target response strategy are input together into a preset script generation model for processing, so as to generate target script text suitable for responding to the user; The target script text is sent to the agent so that the agent can interact with the user via voice based on the target script text.

[0006] Secondly, this application also provides a call center script recommendation device, the call center script recommendation device comprising: The acquisition module is used to acquire target text data, which is obtained by text conversion based on the voice data of the agent and the user; The analysis module is used to input the target text data into a preset intention analysis model for analysis in order to determine the target feature data used to characterize the user's intention. The mapping module is used to determine a target response strategy for responding to the user's intention based on a preset mapping rule table and the target feature data; wherein, the mapping rule table records the mapping relationship between multiple target feature data and multiple response strategies; The generation module is used to input the target text data and the target response strategy into a preset script generation model for processing, so as to generate target script text suitable for responding to the user; The sending module is used to send the target script text to the agent, so that the agent can interact with the user via voice based on the target script text.

[0007] Thirdly, this application also provides a computer device, the computer device including a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the steps of the agent script recommendation method as described above.

[0008] Fourthly, this application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of the agent script recommendation method described above.

[0009] This application provides a method, apparatus, device, and storage medium for recommending agent dialogue scripts. The method involves inputting target text data obtained by text conversion of voice data between agents and users into a preset intention analysis model for analysis, thereby determining target feature data that more comprehensively and accurately represents user intentions. Based on a preset mapping rule table and the target feature data, and through the one-to-one mapping relationship between the target feature data and target response strategies in the mapping rule table, the method quickly and accurately determines the target response strategies for responding to user intentions. The target text data and target response strategies are then input into a preset dialogue script generation model for processing to generate target dialogue script text suitable for responding to users. Finally, the target dialogue script text is sent to the agent, allowing the agent to interact with the user via voice based on the target dialogue script text, thereby improving the matching degree between agent dialogue scripts and user intentions, and ultimately improving the accuracy of agent dialogue script recommendations. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This application provides a flowchart illustrating the steps of a method for recommending customer service scripts. Figure 2 for Figure 1 A flowchart illustrating one step of the method for recommending agent conversation scripts in the context of [the process]. Figure 3 for Figure 1 A flowchart illustrating another sub-step of the agent script recommendation method; Figure 4 A schematic block diagram of a call center script recommendation device provided in this application embodiment; Figure 5 for Figure 4 A schematic block diagram of a sub-module of the agent script recommendation device in the system; Figure 6 for Figure 4 A schematic block diagram of another sub-module of the agent script recommendation device in the system; Figure 7 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0012] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the described order. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0015] This application provides a method, apparatus, device, and storage medium for recommending agent dialogue scripts. The agent dialogue script recommendation method can be applied to a terminal device or a server. The terminal device can be an electronic device such as a mobile phone, tablet computer, laptop computer, desktop computer, personal digital assistant, or wearable device. The server can be a single server or a server cluster composed of multiple servers. The following explanation uses the application of the agent dialogue script recommendation method to a server as an example.

[0016] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0017] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating the steps of a call center script recommendation method provided in an embodiment of this application.

[0018] like Figure 1 As shown, the method for recommending agent scripts includes steps S101 to S105.

[0019] Step S101: Obtain the target text data.

[0020] The target text data can be obtained by converting the voice data between the agent and the user into text. The agent can be an insurance salesperson, customer service representative, etc., while the user can be the target of the agent's outbound calls or a potential customer who proactively contacts the agent.

[0021] It should be noted that in some special cases, such as when an agent initiates an outbound call and states their purpose (e.g., product introduction) after the call is connected, some users may not respond and hang up immediately. In such cases, there is no effective voice interaction content directly related to the user that can be analyzed and converted during the entire call, so there is no need to convert the voice data to obtain the target text data. Therefore, in this embodiment, the target text data should be the conversion result based on the voice data that includes at least one reply or response from the user during the telephone interaction.

[0022] It should be noted that the target text data converted from voice data includes not only the direct question-and-answer content between the user and the agent, but also information such as pauses, volume increases or decreases during the question-and-answer process. For example, a sudden increase in volume at a certain point may indicate a fluctuation in the user's emotions; while a longer pause may indicate that the user is hesitant or needs more time to think.

[0023] Understandably, spoken language typically contains richer information than written language, such as tone of voice and emotional fluctuations. Therefore, the target text data obtained by converting spoken data into text can provide more comprehensive user status information, which helps to more accurately identify user intent, emotions, or potential needs. This, in turn, allows for more appropriate strategies and scripts to be provided to agents in subsequent responses, thereby improving service quality and user experience.

[0024] In one embodiment, acquiring target text data includes: acquiring voice data of a conversation between an agent and a user; performing text conversion processing on the voice data to obtain voice text data to be analyzed; determining contextual voice text data containing the user's voice text data from the voice text data to be analyzed, and using the contextual voice text data as target text data.

[0025] Understandably, by determining the contextual speech-text data containing the user's speech-text data as the target text data, the problem of missing context caused by extracting only single sentences or isolated statements is effectively avoided, ensuring that the obtained target text data can fully restore the complete semantics that the user wants to express.

[0026] It should be noted that the timing of acquiring the target text data can be flexibly determined. Therefore, in addition to "converting speech to text before text recognition" as mentioned above, the target text data can also be obtained by "recognizing speech before text conversion." Specifically, after determining that the voice data packet of the conversation between the agent and the user contains user voice data, the voice data can be directly converted to text and used as the target text data. This effectively improves the efficiency of voice data processing and reduces unnecessary processing steps. Especially in scenarios with frequent dialogue interactions or requiring real-time responses, it can effectively shorten the conversion time from voice data to text data.

[0027] It should also be noted that, in order to further ensure the privacy and security of the aforementioned target text data and related information, the aforementioned target text data and related information can also be stored in a blockchain node. The technical solution of this application can also be applied to adding other data files stored on the blockchain. The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm.

[0028] Step S102: Input the target text data into the preset intention analysis model for analysis to determine the target feature data used to represent the user's intention.

[0029] In this context, intent analysis models can incorporate specific business knowledge through a large-scale model prompt project. This effectively enhances the model's understanding and execution capabilities for specific business-related tasks. When handling actual business dialogues, the model can automatically retrieve relevant domain knowledge to more accurately identify user intent. For example, in the financial sector, the intent analysis model can accurately identify whether a user is inquiring about investment products, account management, or fund transactions based on prompts. Similarly, in the healthcare sector, it can identify whether the user's needs involve health consultations, appointment booking, or medical services. By guiding the intent analysis model to access industry- or enterprise-specific knowledge bases, it can better understand industry terminology, rules, and customer needs, thereby more accurately identifying and responding to complex user requests.

[0030] Intention analysis models can also be obtained by combining natural language processing (NLP) techniques with deep learning models (such as LSTM, BERT, etc.), which can accurately analyze users' intentions through a deep understanding of target text data.

[0031] In one embodiment, the intent analysis model includes a first feature recognition network and a second feature recognition network for recognizing data representing different user characteristics. For example, the first feature recognition network is a convolutional neural network, capable of efficiently extracting first target feature data representing the user's call state, while the second feature recognition network is a recurrent neural network or a long short-term memory network, capable of accurately extracting second target feature data representing the user's emotions. The intent analysis model also includes a feature combination network for combining feature data. For example, the feature combination network is a fully connected layer in a convolutional neural network, capable of integrating feature data from different first and second feature recognition networks to generate a comprehensive user feature representation, thereby providing more comprehensive information for subsequent intent analysis and response strategy generation.

[0032] In one embodiment, such as Figure 2 As shown, step S102 includes sub-steps S1021 to S1023.

[0033] Sub-step S1021: Use the first feature recognition network to perform call status recognition on the target text data to obtain the first target feature data.

[0034] Among them, the first target feature data is used to characterize the user's call status at least.

[0035] In one embodiment, the first feature recognition network includes a first feature recognition subnetwork, a call status analysis subnetwork, a second feature recognition subnetwork, and a feature fusion subnetwork.

[0036] The system comprises three subnetworks: a first feature recognition subnetwork, which identifies preliminary features related to the call status from the raw target text data, such as customer feedback, questions, or replies; and a second feature recognition subnetwork, which analyzes and identifies the user's current call status category. Specifically, the second feature recognition subnetwork further analyzes the preliminary feature data extracted by the first feature recognition subnetwork to determine the user's status category, such as whether the user is in a "connected call," "discussing car insurance renewal," or "ended call" state. Finally, the second feature recognition subnetwork identifies and extracts the user's emotional features. Specifically, through deep understanding of the target text data, the second feature recognition subnetwork analyzes the user's tone, emotion, and emotional fluctuations in the conversation to help the system identify the user's emotional state, such as "happy" or "resistant," or "positive," "neutral," or "negative."

[0037] It should be noted that, depending on the actual input target text data, the first feature recognition network can selectively utilize one or more of its subnetworks to achieve feature extraction.

[0038] In one embodiment, analyzing target text data using a first feature recognition network to obtain first target feature data includes: using a first feature extraction sub-network to perform call state recognition on the target text data to obtain first sub-target feature data representing the user's call state; using a call state analysis sub-network to analyze the first sub-target feature data to determine the call state category corresponding to the first sub-target feature data, where the call state categories include a first category of call state and a second category of call state; if the call state category corresponding to the first sub-target feature data is the first category of call state, then the first sub-target feature data representing the user's call state is used as the first target feature data; if the call state category corresponding to the first candidate feature data is the first category of call state, then using a second feature extraction sub-network to perform user intention recognition on the target text data to obtain second sub-target feature data representing the user's intention; and using a feature fusion sub-network to perform feature fusion processing on the first target feature data and the second sub-target feature data to obtain first target feature data representing the user's call state and user intention.

[0039] Taking the insurance business scenario in the financial sector as an example, the call status is categorized into six states: "Call connected (initial state)," "Discussing the car insurance renewal plan," "Agreeing to renew the car insurance," "Discussing the comprehensive medical insurance," "Agreeing to purchase the comprehensive medical insurance," and "Call ended (terminated state)." The correlation between various user intentions corresponding to each call status is shown in Table 1 below. Among them, "Agreeing to renew the car insurance" and "Agreeing to purchase the comprehensive medical insurance" are defined as the first category of call status, and "Call connected," "Discussing the car insurance renewal plan," "Discussing the comprehensive medical insurance," and "Call ended" are defined as the second category of call status.

[0040] Table 1

[0041] For example, in Scenario 1, the call process between the user and the agent is as follows: Operator: Hello Ms. A, this is Business Manager B from XX Insurance Company. I'm calling to inform you that your car insurance is about to expire and I'd like to ask if you have any plans to renew it. User: I've recently started looking at car insurance from other companies.

[0042] In the above scenario, the system identifies and obtains the first sub-target feature data representing the user's "connected call" state, and determines the call state category corresponding to the first sub-target feature data as the second category of call state. Therefore, the system further performs user intention identification and determines the second sub-target feature data representing the user's intention as "contacted other insurance companies" according to Table 1. Finally, through feature fusion, the system obtains the first target feature data representing the user's call state as "connected call" and the user's intention as "contacted other insurance companies".

[0043] For example, in scenario two, the system identifies and obtains the first sub-target feature data that represents the user being in the "agree to renew car insurance" call state. In this case, the first sub-target feature data that represents the user being in the "agree to renew car insurance" call state is directly used as the first target feature data.

[0044] Sub-step S1022: Use the second feature recognition network to perform user emotion recognition on the target text data to obtain the second target feature data.

[0045] The second objective feature data is used to characterize user emotions. It is understandable that by acquiring this second objective feature data, it is possible to better understand and respond to users' emotional needs, which helps to improve the quality of interaction between agents and users and increase user satisfaction.

[0046] For example, user emotions are defined as three types: "positive," "neutral," and "negative." Specifically, Table 1 also records the mapping relationship between each call status, each user's intention, and each user's emotion. Suppose that a user, while in a "connected" call status, speaks with impatience when saying, "I've been looking at car insurance from another company recently," then the user's emotion is determined to be "negative." Based on this emotion, the system generates corresponding second target feature data to characterize the user's "negative" emotion.

[0047] Sub-step S1023: Use a feature combination network to perform feature combination processing on the first target feature data and the second target feature data to obtain target feature data that represents the user's intention.

[0048] Understandably, by combining the first and second target feature data, a more comprehensive set of target feature data can be formed. This helps to achieve a more comprehensive and accurate analysis and generate response strategies for users when analyzing the target feature data in the future.

[0049] Step S103: Based on the preset mapping rule table and target feature data, determine the target response strategy to address the user's intentions. The mapping rule table records the mapping relationships between multiple target feature data and multiple response strategies. It can be understood that by utilizing the mapping rule table, the target feature data that accurately and comprehensively represents the user's intent can be used to quickly and precisely determine the target response strategies to help agents deal with users. For example, suppose the target feature data corresponding to user A represents their call state as "communicating about a car insurance claim," their intention as "requesting an introduction to health insurance," and their emotion as "positive." Referring to Table 1, the target response strategy suitable for agents dealing with user A is the "topic-entry strategy."

[0050] In one embodiment, in addition to recording the mapping relationship between multiple target feature data and multiple response strategies, the mapping rule table also records the mapping relationship between the cumulative number of times multiple target feature data are acquired and the response strategies; based on the preset mapping rule table, the target response strategy for dealing with the target combination feature data is determined, including: determining the cumulative number of times the target combination feature data is acquired; based on the target feature data, determining multiple first response strategies corresponding to it from the mapping rule table; based on the cumulative number of acquisitions, determining a corresponding second response strategy from the multiple first response strategies as the target response strategy.

[0051] Understandably, the cumulative number of acquisitions reflects the frequency of occurrence of target feature data, allowing the system to make more precise and flexible dynamic strategy adjustments based on this information. For example, some coping strategies may be suitable at the beginning of a call, but as the frequency of target feature data increases and reaches a certain threshold, the system will adjust to a coping strategy that is more appropriate for the current call context.

[0052] For example, referring to Table 1, during a call between an agent and a user, if the cumulative number of times the target combination feature data used to represent the call status as "communicating about million-dollar medical insurance", the intention as "dissatisfaction with the product", and any emotion is less than 3, the target response strategy for dealing with the target combination feature data is an objection facilitation strategy. However, if the cumulative number of times the target combination feature data is obtained is greater than or equal to 3, the target response strategy for dealing with the target combination feature data is adjusted to a product recommendation strategy.

[0053] Step S104: Input the target text data and target response strategy into the preset script generation model for processing to generate target script text suitable for responding to users.

[0054] The dialogue generation model, trained on a large amount of dialogue data (e.g., GPT, BERT, or T5), automatically generates context-relevant natural language responses tailored to user needs based on the input target text data and target response strategies. Essentially, by utilizing the dialogue generation model, target text data and target response strategies can be transformed into natural and fluent response content suitable for answering users. This target dialogue text not only accurately addresses user needs but also adjusts tone and wording based on user emotions and dialogue context, enabling agents to interact efficiently with users. This not only improves the overall service quality of agents but also enhances overall user satisfaction.

[0055] It should be noted that in some cases, such as business scenarios that require highly customized and professional judgment, the script generation model may not be used. Instead, the target response strategy can be sent directly to the agent. The agent will then organize their language and interact with the user via voice based on the general response direction provided by the target response strategy, combined with their personal experience and professional knowledge, thereby providing the user with more personalized services that meet their actual needs.

[0056] In one embodiment, such as Figure 3 As shown, step S104 includes sub-steps S1041 to S1042.

[0057] Sub-step S1041: Input the target text data and target response strategy into the preset speech generation model for analysis to determine the strategy category of the target response strategy.

[0058] Depending on the specific business category, there can be two or more strategy categories. For example, there might be two strategy categories: an ending strategy to instruct the agent to end the voice interaction with the user, and a continuing strategy to instruct the agent to continue the voice interaction with the user. In addition to the above two, strategy categories can also include guiding strategies applicable to situations where the customer has not yet clearly expressed their needs or requires further prompting, guiding the user into a deeper conversation, and so on. Determining the strategy category of the target response strategy using a script generation model helps to more accurately generate the target script text suitable for responding to the user based on the strategy category.

[0059] Sub-step S1042: Based on the strategy category and target text data of the target response strategy, generate target dialogue text for instructing the agent to end the voice interaction with the user or instructing the agent to continue the voice interaction with the user.

[0060] For example, suppose the strategy categories include ending strategies and continuing strategies. Accordingly, when the target response strategy is an ending strategy, a target dialogue text is generated based on the target response strategy and target text data to instruct the agent to end the voice interaction with the user. For example, the target dialogue text "I won't bother you anymore, have a nice day, goodbye" is generated. This not only ends the conversation but also expresses respect and good wishes to the user in a warm tone, helping to improve user goodwill and overall experience. When the target response strategy is a continuing strategy, a target dialogue text is generated based on the target response strategy and the dialogue process reflected by the target dialogue text. For example, the dialogue text "We have a comprehensive medical insurance plan that is very suitable for you..." is generated to continue guiding the conversation and provide the user with more service information or product recommendations.

[0061] By generating more targeted and relevant scripts that better reflect the user's true intentions or needs through the strategy categories of target response strategies, the smooth and targeted dialogue between agents and users can be ensured, which helps improve customer service efficiency and user satisfaction.

[0062] In one embodiment, the target response strategies include position-keeping strategies, topic-entry strategies, product Q&A strategies, topic objection strategies, and product recommendation strategies. Position-keeping strategies can be used to provide closing statements when the agent is unable to advance the communication process. Topic-entry strategies can be used to find opportune moments to introduce topics during the communication process and provide recommended statements. Product Q&A strategies can be used to provide professional answers to product-related questions raised by users. Topic objection strategies can be used to provide solutions and consensus-building statements for objections raised by users during the communication process. Product recommendation strategies can be used to recommend suitable products to users based on their specific needs.

[0063] In one embodiment, based on the strategy category of the target response strategy and the target text data, a target dialogue text is generated to instruct the agent to end the voice interaction with the user or to instruct the agent to continue the voice interaction with the user. This includes: if the category of the target response strategy is a placeholder strategy, then a target dialogue text is generated based on the target response strategy and the target text data to instruct the agent to end the voice interaction with the user; if the category of the target response strategy is a topic entry strategy, a product Q&A strategy, a topic objection strategy, or a product recommendation strategy, then a target dialogue text is generated based on the target response strategy and the target text data to instruct the agent to continue the voice interaction with the user.

[0064] Taking the insurance business recommendation scenario in the financial sector as an example, the specific uses of each strategy are as follows: The placeholder strategy is used to provide closing remarks when agents are unable to advance the communication process. For example, when a user shows no interest or inability to continue the conversation, the system generates an appropriate closing remark based on the placeholder strategy, such as "Thank you for your call. Please feel free to contact us if you have any questions. Have a nice day." The topic-entry strategy is used to find the right time to introduce the topic of comprehensive medical insurance during car insurance renewal communication and provides recommended scripts. For example, when a user inquires about car insurance renewal information, the agent can use the topic-entry strategy to steer the conversation towards comprehensive medical insurance. The system generates a recommended script: "In addition to car insurance, our comprehensive medical insurance is also very popular. It provides comprehensive medical coverage. Are you interested in learning more?" The product Q&A strategy is used to provide professional answers to questions raised by users regarding information about the comprehensive medical insurance product, underwriting, and claims processing. For example, when a user raises questions about the product, underwriting, and claims processing, the system generates a precise answer based on the product Q&A strategy: "Our comprehensive medical insurance provides coverage including hospitalization and surgery costs. The underwriting process usually requires a simple health questionnaire, and the specific terms can be customized according to your health condition." The topic-based objection strategy is used to provide solutions and consensus-building scripts for addressing objections that users may have during discussions about comprehensive medical insurance. For example, if a user has objections to the price, coverage, or other aspects of comprehensive medical insurance, the system can generate the following script based on the topic-based objection strategy: "I understand your concerns about the price, but our comprehensive medical insurance offers coverage far exceeding that of general medical insurance on the market, and our premiums are very competitive. Would you consider adjusting the coverage?" Product recommendation strategies are used to tailor suitable comprehensive medical insurance products to meet specific customer needs. For example, if a user indicates specific requirements regarding premiums and coverage amounts, the system generates a script based on the product recommendation strategy, such as, "Based on your needs, I recommend one of our comprehensive medical insurance plans suitable for your family. It includes hospitalization reimbursement regardless of age, offers comprehensive coverage, and has a reasonable premium, making it very suitable for your situation." Understandably, through these specific application scenarios and corresponding strategies, the system can flexibly respond to different situations in its interactions with users, ensuring that agents have appropriate strategies to provide high-quality service in various dialogue scenarios. This helps agents smoothly guide the dialogue process, resolve user concerns, and ultimately facilitate transactions or improve customer satisfaction.

[0065] In one embodiment, the script generation model includes a strategy category recognition network, a topic end script generation network, a topic entry script generation network, a product Q&A script generation network, a topic objection script generation network, and a product recommendation script generation network.

[0066] Specifically, the strategy category recognition network is used to train the topic entry dialogue generation network based on topic entry strategies, multiple sample speech and text data, and multiple sample user profile data. The product question-and-answer dialogue generation network is trained based on product question-and-answer strategies, multiple sample speech and text data, and multiple product data. The topic objection dialogue generation network is trained based on topic objection strategies, multiple sample speech and text data, and multiple sample user profile data. The product recommendation dialogue generation network is trained based on product recommendation strategies, multiple sample speech and text data, multiple sample user profile data, and multiple product data.

[0067] In one embodiment, the target text data and the target response strategy are jointly input into a strategy category recognition network for analysis to obtain the category of the target response strategy. This includes: inputting the target text data and the target response strategy into the strategy category recognition network for analysis to obtain the category of the target response strategy; if the category of the target response strategy is a placeholder strategy, then using a topic ending dialogue generation network to process the target text data and the target response strategy to obtain a topic ending dialogue; if the category of the target response strategy is a topic entry strategy, then obtaining the user's target profile data from a preset user profile database, and using a topic entry dialogue generation network to process the target profile data, target text data, and target response strategy to obtain a topic entry dialogue text; if the target response strategy... If the category is product Q&A strategy, the target text data and target response strategy are processed using a product Q&A script generation network to obtain product Q&A script text. If the category of the target response strategy is topic objection strategy, the target text data and target response strategy are processed using a topic objection script generation network to obtain topic objection script text. If the category of the target response strategy is product recommendation strategy, the target profile data of the user is obtained from the user profile database, and the target text data and target response strategy are processed using a product recommendation script generation network to obtain product recommendation script text. The topic ending script, topic entry script text, product Q&A script text, topic objection script text, or product recommendation script text are used as the target script text suitable for replying to the user.

[0068] Specifically, the topic entry dialogue generation network can accurately select the appropriate time to enter a new topic in a conversation and generate highly guiding language by analyzing the user's interests and needs; the product Q&A dialogue generation network can generate professional answers based on the user's product questions, ensuring that the information is accurate and meets business needs; the topic objection dialogue generation network can generate appropriate responses when the user has objections to the product or service, helping agents eliminate user concerns and promote consensus; and the product recommendation dialogue generation network can recommend the most suitable products or services based on the user's specific needs and interests, helping agents provide personalized suggestions.

[0069] By combining target text data and target response strategies, and processing each response strategy using a pre-defined script generation model that can generate different types of scripts, the generated target script text is made highly consistent with the user's intentions, thereby improving customer satisfaction and business conversion capabilities.

[0070] Step S105: Send the target script text to the agent so that the agent can interact with the user via voice based on the target script text.

[0071] Taking an insurance product recommendation scenario as an example, the system generates the following target script: "Based on your needs, I recommend our family version of the comprehensive medical insurance. It not only provides comprehensive coverage but also offers higher coverage amounts for family members, making it very suitable for your situation. Are you interested in learning more details?" This target script is then displayed on the computer terminal next to the agent. The agent then interacts with the user via voice based on this script, further elaborating on the recommendations, answering the user's questions, and encouraging the user to learn more.

[0072] Understandably, by sending targeted scripts that closely match user intent to agents, agents can interact with users more effectively based on the scripts, ensuring smooth and targeted dialogue, and guiding users to learn more about products or services at appropriate times, thereby improving customer satisfaction.

[0073] The agent dialogue recommendation method provided in the above embodiments involves inputting target text data obtained by text conversion of voice data between agents and users into a preset intention analysis model for analysis, thereby determining target feature data that can more comprehensively and accurately represent user intentions; based on a preset mapping rule table and target feature data, and through the one-to-one mapping relationship between target feature data and target response strategies in the mapping rule table, quickly and accurately determining the target response strategies for responding to user intentions; inputting the target text data and target response strategies together into a preset dialogue generation model for processing to generate target dialogue text suitable for responding to users; and sending the target dialogue text to agents so that agents can interact with users through voice based on the target dialogue text, thereby improving the matching degree between agent dialogue and user intentions, and thus improving the accuracy of agent dialogue recommendations, achieving a simultaneous improvement in agent service quality and user satisfaction.

[0074] Please refer to Figure 4 , Figure 4 This is a schematic block diagram of a call center script recommendation device provided in an embodiment of this application.

[0075] like Figure 4 As shown, the agent script recommendation device 200 includes: The acquisition module 201 is used to acquire target text data, which is obtained by text conversion based on the voice data of the agent and the user; Analysis module 202 is used to input the target text data into a preset intention analysis model for analysis, so as to determine the target feature data used to represent the user's intention; The mapping module 203 is used to determine a target response strategy for responding to the user's intention based on a preset mapping rule table and the target feature data; wherein, the mapping rule table records the mapping relationship between multiple target feature data and multiple response strategies; The generation module 204 is used to input the target text data and the target response strategy into a preset script generation model for processing, so as to generate target script text suitable for responding to the user; The sending module 205 is used to send the target script text to the agent, so that the agent can interact with the user by voice based on the target script text.

[0076] In one embodiment, the acquisition module 201 is further configured to: Acquire voice data from the conversation between the agent and the user; perform text conversion on the voice data to obtain the voice text data to be analyzed; determine the context voice text data containing the user's voice text data from the voice text data to be analyzed, and use the context voice text data as the target text data.

[0077] In one embodiment, such as Figure 5 As shown, the analysis module 202 includes: The first feature recognition submodule 2021 is used to perform call status recognition on the target text data using the first feature recognition network to obtain first target feature data, which is used to characterize the call status of the user.

[0078] The second feature recognition submodule 2022 is used to perform user emotion recognition on target text data using the second feature recognition network to obtain second target feature data, which is used to represent user emotions.

[0079] The feature combination submodule 2023 is used to perform feature combination processing on the first target feature data and the second target feature data using a feature combination network to obtain target feature data that represents the user's intention.

[0080] In one embodiment, the first feature recognition network includes a first feature recognition subnetwork, a call state analysis subnetwork, a second feature recognition subnetwork, and a feature fusion subnetwork. The first feature recognition submodule 2021 is further configured to: The first feature extraction subnetwork is used to identify the call status of the target text data to obtain the first sub-target feature data that represents the call status of the user. The call status analysis subnetwork is used to analyze the feature data of the first sub-target to determine the call status category corresponding to the feature data of the first sub-target. The call status categories include the first category of call status and the second category of call status. If the call status category corresponding to the first sub-target feature data is the first category of call status, then the first sub-target feature data used to characterize the call status of the user will be used as the first target feature data. If the call status category corresponding to the first candidate feature data is the first category of call status, then the second feature extraction sub-network is used to identify the user's intention in the target text data to obtain the second sub-target feature data used to represent the user's intention; and the feature fusion sub-network is used to perform feature fusion processing on the first target feature data and the second sub-target feature data to obtain the first target feature data used to represent the user's call status and user intention.

[0081] In one embodiment, the mapping rule table also records the mapping relationship between the cumulative number of times multiple target feature data are acquired and the corresponding strategies. The mapping module 203 is further used for: Determine the cumulative number of times the target combination feature data has been acquired; Based on the target feature data, multiple corresponding first response strategies are determined from the mapping rule table; Based on the cumulative number of acquisitions, the corresponding second response strategy is determined from multiple first response strategies as the target response strategy.

[0082] In one embodiment, such as Figure 6 As shown, the sending module 205 includes: The strategy analysis submodule 2051 is used to input the target text data and the target response strategy into a preset speech generation model for analysis in order to determine the strategy category of the target response strategy.

[0083] The script generation submodule 2052 is used to generate target script text for instructing agents to end voice interaction with users or instructing agents to continue voice interaction with users, based on the strategy category of the target response strategy and the target text data.

[0084] In one embodiment, the dialogue generation model includes a topic-ending dialogue generation network for generating a first text indicating the end of the voice interaction, and a topic-entry dialogue generation network, a product question-and-answer dialogue generation network, a topic objection dialogue generation network, and a product recommendation dialogue network for generating text indicating the continuation of the voice interaction. The sending module 205 is further configured to: If the target response strategy is a placeholder strategy, then the topic ending phrase generation network is used to process the target text data and the target response strategy to obtain the topic ending phrase. If the target response strategy is a topic-based approach strategy, then the target profile data of the user is obtained from the preset user profile database, and the target profile data, target text data and target response strategy are processed by the topic-based approach speech generation network to obtain the topic-based approach speech text; wherein, the topic-based approach speech generation network is trained based on the topic-based approach strategy, multiple sample speech text data and multiple sample user profile data. If the target response strategy is a product question-and-answer strategy, then the product question-and-answer script generation network is used to process the target text data and the target response strategy to obtain the product question-and-answer script text; wherein, the product question-and-answer script generation network is trained based on the product question-and-answer strategy, multiple sample speech and text data and multiple product data; If the target response strategy is a topic objection strategy, then the topic objection speech generation network is used to process the target text data and the target response strategy to obtain the topic objection speech text; wherein, the topic objection speech generation network is trained based on the topic objection strategy, multiple sample speech text data and multiple sample user profile data. If the target response strategy is a product recommendation strategy, then the target profile data of the user is obtained from the user profile database, and the target text data and the target response strategy are processed by the product recommendation script generation network to obtain the product recommendation script text; wherein, the product recommendation script generation network is trained based on the product recommendation strategy, multiple sample voice text data, multiple sample user profile data and multiple product data; Use the following texts as target texts for responding to users: topic closing statements, topic entry statements, product Q&A statements, topic objection statements, or product recommendation statements.

[0085] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the device and each module and unit described above can be referred to the corresponding processes in the aforementioned embodiments of the agent dialogue recommendation method, and will not be repeated here.

[0086] The apparatus provided in the above embodiments can be implemented as a computer program, which can be used in, for example... Figure 7 It runs on the computer device shown.

[0087] Please see Figure 7 , Figure 7 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0088] like Figure 7As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The memory may include a storage medium and internal memory, and the storage medium may be non-volatile or volatile.

[0089] The storage medium may store the operating system and computer programs. These computer programs include program instructions that, when executed, cause the processor to perform any agent script recommendation method.

[0090] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0091] Internal memory provides an environment for the execution of computer programs stored in the storage medium. When these computer programs are executed by the processor, the processor can execute any method of recommending agent scripts.

[0092] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0093] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0094] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: The target text data is obtained by converting the voice data of the agent and the user into text. The target text data is input into a preset intention analysis model for analysis in order to determine the target feature data used to represent user intentions; Based on a pre-defined mapping rule table and target feature data, target response strategies for addressing user intentions are determined; the mapping rule table records the mapping relationships between multiple target feature data and multiple response strategies. The target text data and target response strategy are input together into a preset script generation model for processing to generate target script text suitable for responding to users; The target script text is sent to the agent so that the agent can interact with the user via voice based on the target script text.

[0095] In one embodiment, when the processor acquires the target text data, it is used to: Acquire voice data from the conversation between the agent and the user; The speech data is converted into text to obtain the speech-text data to be analyzed. Identify the contextual speech-text data containing the user's speech-text data from the speech-text data to be analyzed, and use the contextual speech-text data as the target text data.

[0096] In one embodiment, the intent analysis model includes a first feature recognition network, a second feature recognition network, and a feature combination network. When the processor inputs target text data into the preset intent analysis model for analysis to determine target feature data representing user intent, it performs the following: The first feature recognition network is used to identify the call status of the target text data to obtain the first target feature data, which is used to characterize the call status of the user. The second feature recognition network is used to perform user emotion recognition on the target text data to obtain second target feature data, which is used to represent user emotions. The first target feature data and the second target feature data are combined using a feature combination network to obtain target feature data that represents the user's intention.

[0097] In one embodiment, the first feature recognition network includes a first feature recognition subnetwork, a call state analysis subnetwork, a second feature recognition subnetwork, and a feature fusion subnetwork. When the processor analyzes the target text data using the first feature recognition network to obtain first target feature data, it is configured to: The first feature extraction subnetwork is used to identify the call status of the target text data to obtain the first sub-target feature data that represents the call status of the user. The call status analysis subnetwork is used to analyze the feature data of the first sub-target to determine the call status category corresponding to the feature data of the first sub-target. The call status categories include the first category of call status and the second category of call status. If the call status category corresponding to the first sub-target feature data is the first category of call status, then the first sub-target feature data used to characterize the call status of the user will be used as the first target feature data. If the call status category corresponding to the first candidate feature data is the first category of call status, then the second feature extraction sub-network is used to identify the user's intention in the target text data to obtain the second sub-target feature data used to represent the user's intention; and the feature fusion sub-network is used to perform feature fusion processing on the first target feature data and the second sub-target feature data to obtain the first target feature data used to represent the user's call status and user intention.

[0098] In one embodiment, the mapping rule table also records the mapping relationship between the cumulative number of times multiple target feature data are acquired and the corresponding strategies. When the processor determines the target response strategy for dealing with the target combination feature data based on the preset mapping rule table, it is used to implement: Determine the cumulative number of times the target combination feature data has been acquired; Based on the target feature data, multiple corresponding first response strategies are determined from the mapping rule table; Based on the cumulative number of acquisitions, the corresponding second response strategy is determined from multiple first response strategies as the target response strategy.

[0099] In one embodiment, when the processor inputs target text data and target response strategies into a preset script generation model for processing to generate target script text suitable for responding to users, it is used to: The target text data and target response strategies are input together into a pre-defined script generation model for analysis to determine the strategy category of the target response strategy; and Based on the strategy category and target text data of the target response strategy, generate target script text to instruct the agent to end the voice interaction with the user or to instruct the agent to continue the voice interaction with the user.

[0100] In one embodiment, the dialogue generation model includes a topic-ending dialogue generation network for generating a first text indicating the end of voice interaction, and a topic-entry dialogue generation network, a product question-and-answer dialogue generation network, a topic objection dialogue generation network, and a product recommendation dialogue network for generating a first text indicating the end of voice interaction. The processor, when inputting target text data and target response strategies into the preset dialogue generation model for processing to generate target dialogue text suitable for responding to users, is used to: If the target response strategy is a placeholder strategy, then the topic ending phrase generation network is used to process the target text data and the target response strategy to obtain the topic ending phrase. If the target response strategy is a topic-based approach strategy, then the target profile data of the user is obtained from the preset user profile database, and the target profile data, target text data and target response strategy are processed by the topic-based approach speech generation network to obtain the topic-based approach speech text; wherein, the topic-based approach speech generation network is trained based on the topic-based approach strategy, multiple sample speech text data and multiple sample user profile data. If the target response strategy is a product question-and-answer strategy, then the product question-and-answer script generation network is used to process the target text data and the target response strategy to obtain the product question-and-answer script text; wherein, the product question-and-answer script generation network is trained based on the product question-and-answer strategy, multiple sample speech and text data and multiple product data; If the target response strategy is a topic objection strategy, then the topic objection speech generation network is used to process the target text data and the target response strategy to obtain the topic objection speech text; wherein, the topic objection speech generation network is trained based on the topic objection strategy, multiple sample speech text data and multiple sample user profile data. If the target response strategy is a product recommendation strategy, then the target profile data of the user is obtained from the user profile database, and the target text data and the target response strategy are processed by the product recommendation script generation network to obtain the product recommendation script text; wherein, the product recommendation script generation network is trained based on the product recommendation strategy, multiple sample voice text data, multiple sample user profile data and multiple product data; Use the following texts as target texts for responding to users: topic closing statements, topic entry statements, product Q&A statements, topic objection statements, or product recommendation statements.

[0101] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the computer equipment described above can be referred to the corresponding process in the aforementioned embodiment of the agent script recommendation method, and will not be repeated here.

[0102] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0103] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, and the method implemented when the program instructions are executed can refer to various embodiments of the agent script recommendation method of this application.

[0104] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0105] Furthermore, the computer's usable storage medium may primarily include a stored program area and a stored data area. The stored program area may store the operating system, applications required for at least one function, etc.; the stored data area may store data created based on the use of blockchain nodes, etc. The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. A blockchain is essentially a decentralized database, a chain of data blocks linked using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain may include a blockchain underlying platform, a platform product service layer, and an application service layer, etc.

[0106] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0107] It should also be understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. It should be noted that, herein, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0108] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. The above descriptions are merely specific implementations of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for recommending sales scripts for agents, characterized in that, include: Acquire target text data, which is obtained by text conversion based on voice data from agents and users; The target text data is input into a preset intention analysis model for analysis in order to determine the target feature data used to characterize the user's intention; Based on a preset mapping rule table and the target feature data, a target response strategy for responding to the user's intention is determined; wherein, the mapping rule table records the mapping relationship between multiple target feature data and multiple response strategies; The target text data and the target response strategy are input together into a preset script generation model for processing, so as to generate target script text suitable for responding to the user; The target script text is sent to the agent so that the agent can interact with the user via voice based on the target script text.

2. The agent script recommendation method as described in claim 1, characterized in that, The intention analysis model includes a first feature recognition network, a second feature recognition network, and a feature combination network; The step of inputting the target text data into a preset intention analysis model for analysis to determine target feature data for characterizing user intention includes: The first feature recognition network is used to identify the call status of the target text data to obtain first target feature data, which is used to characterize at least the call status of the user. The second feature recognition network is used to perform user emotion recognition on the target text data to obtain second target feature data, which is used to characterize user emotions. The feature combination network is used to perform feature combination processing on the first target feature data and the second target feature data to obtain target feature data that represents user intention.

3. The agent script recommendation method as described in claim 2, characterized in that, The first feature recognition network includes a first feature recognition subnetwork, a call status analysis subnetwork, a second feature recognition subnetwork, and a feature fusion subnetwork; The step of analyzing the target text data using the first feature recognition network to obtain first target feature data includes: The first feature extraction sub-network is used to identify the call status of the target text data to obtain first sub-target feature data that characterizes the call status of the user. The call status analysis sub-network is used to analyze the first sub-target feature data to determine the call status category corresponding to the first sub-target feature data. The call status category includes a first category of call status and a second category of call status. If the call status category corresponding to the first sub-target feature data is the first category of call status, then the first sub-target feature data used to characterize the call status of the user will be used as the first target feature data. If the call status category corresponding to the first candidate feature data is the first category of call status, then the second feature extraction sub-network is used to perform user intention recognition on the target text data to obtain second sub-target feature data for representing user intention; and the feature fusion sub-network is used to perform feature fusion processing on the first target feature data and the second sub-target feature data to obtain first target feature data for representing the user's call status and user intention.

4. The method for recommending sales scripts as described in claim 1, characterized in that, The step of inputting the target text data and the target response strategy into a preset script generation model for processing to generate target script text suitable for responding to the user includes: The target text data and the target response strategy are input together into a preset speech generation model for analysis to determine the strategy category of the target response strategy; and Based on the strategy category of the target response strategy and the target text data, a target script text is generated to instruct the agent to end the voice interaction with the user or to instruct the agent to continue the voice interaction with the user.

5. The agent script recommendation method as described in claim 1, characterized in that, The script generation model includes a strategy category recognition network, a topic end script generation network, a topic entry script generation network, a product Q&A script generation network, a topic objection script generation network, and a product recommendation script generation network. The step of inputting the target text data and the target response strategy into a preset script generation model for processing to generate target script text suitable for responding to the user includes: The target text data and the target response strategy are input together into the strategy category recognition network for analysis to obtain the category of the target response strategy; If the target response strategy is classified as a placeholder strategy, then the topic ending phrase generation network is used to process the target text data and the target response strategy to obtain the topic ending phrase. If the target response strategy is classified as a topic-based approach strategy, then the target profile data of the user is obtained from a preset user profile database, and the target profile data, the target text data, and the target response strategy are processed by the topic-based approach speech generation network to obtain the topic-based approach speech text; wherein, the topic-based approach speech generation network is trained based on the topic-based approach strategy, multiple sample speech text data, and multiple sample user profile data. If the target response strategy is a product question-and-answer strategy, then the product question-and-answer script generation network is used to process the target text data and the target response strategy to obtain product question-and-answer script text; wherein, the product question-and-answer script generation network is trained based on the product question-and-answer strategy, multiple sample speech text data and multiple product data; If the target response strategy is classified as a topic objection strategy, then the topic objection speech generation network is used to process the target text data and the target response strategy to obtain the topic objection speech text; wherein, the topic objection speech generation network is trained based on the topic objection strategy, multiple sample speech text data and multiple sample user profile data; If the target response strategy is a product recommendation strategy, then the target profile data of the user is obtained from the user profile database, and the target text data and the target response strategy are processed by the product recommendation script generation network to obtain the product recommendation script text; wherein, the product recommendation script generation network is trained based on the product recommendation strategy, multiple sample voice text data, multiple sample user profile data and multiple product data. Use the topic ending phrase, the topic entry phrase, the product Q&A phrase, the topic objection phrase, or the product recommendation phrase as the target phrase for responding to the user.

6. The agent script recommendation method as described in any one of claims 1-5, characterized in that, The mapping rule table also records the mapping relationship between the cumulative number of times multiple target feature data are acquired and the corresponding strategies. The method for determining a target response strategy based on a preset mapping rule table to address the target combination feature data includes: Determine the cumulative number of times the target combination feature data has been acquired; Based on the target feature data, multiple corresponding first response strategies are determined from the mapping rule table; Based on the cumulative number of acquisitions, a corresponding second response strategy is determined from the plurality of first response strategies as the target response strategy.

7. The agent script recommendation method as described in any one of claims 1-5, characterized in that, The acquisition of target text data includes: Acquire voice data from the conversation between the agent and the user; The speech data is processed by text conversion to obtain the speech-text data to be analyzed; Contextual speech text data containing the user's speech text data is determined from the speech text data to be analyzed, and the contextual speech text data is used as the target text data.

8. A call center script recommendation device, characterized in that, The agent script recommendation device includes: The acquisition module is used to acquire target text data, which is obtained by text conversion based on the voice data of the agent and the user; The analysis module is used to input the target text data into a preset intention analysis model for analysis in order to determine the target feature data used to characterize the user's intention. The mapping module is used to determine a target response strategy for responding to the user's intention based on a preset mapping rule table and the target feature data; wherein, the mapping rule table records the mapping relationship between multiple target feature data and multiple response strategies; The generation module is used to input the target text data and the target response strategy into a preset script generation model for processing, so as to generate target script text suitable for responding to the user; The sending module is used to send the target script text to the agent, so that the agent can interact with the user via voice based on the target script text.

9. A computer device, characterized in that, The computer device includes a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the agent script recommendation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the agent script recommendation method as described in any one of claims 1 to 7.

Images