Talk skill recommendation method and device, equipment and medium

Abstract

The invention belongs to the field of big data, and relates to a verbal skill recommendation method, comprising: receiving interaction information between a client terminal and an agent terminal, the interaction information comprising consultation information sent by the client terminal and reply information returned by the agent terminal based on the consultation information; based on the consultation information, acquiring multi-source behavior data of a target client corresponding to the client terminal from a multi-source data platform by adopting a preset point burying technology; constructing a customer portrait of the target customer according to the multi-source behavior data; based on the consultation information and the reply information, obtaining a basic verbal skill template from a preset verbal skill library; and based on the customer portrait, adjusting the basic verbal skill template, generating a target verbal skill, and recommending the target verbal skill to the proxy terminal, so that the proxy terminal interacts with the target customer based on the target verbal skill. The invention further provides a device, equipment and a medium. The method can be applied to the business fields of financial insurance and the like, and the verbal skill recommendation accuracy can be improved.

Description

Technical Field

[0001] This application relates to the field of big data technology and is applied to online processing business scenarios such as finance and insurance. In particular, it relates to a method, device, equipment, and medium for recommending sales scripts. Background Technology

[0002] In the financial and insurance sector, digital sales support systems have become a key infrastructure for enhancing agent professional efficiency and driving business growth. Currently, leading insurance institutions have achieved a high level of coverage for digital tools; however, the increase in sales conversion rates has not met expectations, highlighting the deep-seated shortcomings of traditional technology architectures.

[0003] From a technical architecture perspective, existing solutions generally adopt a three-tier architecture consisting of a front-end interactive interface, a mid-level rule engine, and a back-end static knowledge base. This architecture essentially builds a keyword-matching-based retrieval question-answering system. Its operating mechanism involves retrieving and returning relevant information from a static knowledge base using pre-defined keyword rules. Taking the Salesforce Einstein platform as an example, although it utilizes natural language processing technology to achieve semantic expansion to some extent, its core still heavily relies on manually annotated frequently asked questions (FAQs). This reliance leads to multi-dimensional mismatches between the system's generated response strategies and actual business scenarios, failing to accurately meet the diverse and personalized needs of customers.

[0004] More importantly, existing systems have extremely weak support capabilities in complex negotiation scenarios. In actual sales processes, agents often face various complex negotiation situations, such as customers questioning the price of insurance products or requesting detailed comparisons with competitors. However, traditional systems can only provide standardized sales script templates, which often lack specificity and flexibility, and cannot be dynamically adjusted according to the specific characteristics of customers, the specific negotiation context, and real-time market dynamics. This significantly reduces the accuracy of script recommendations, thus hindering sales conversion rates. In the increasingly competitive financial insurance market, this inefficient sales support method severely restricts business development. Therefore, there is an urgent need for a system that can effectively improve the accuracy of sales script recommendations to solve the problem of low accuracy in current digital sales support systems in complex negotiation scenarios, which in turn hinders sales conversion rates. Summary of the Invention

[0005] The purpose of this application is to provide a sales script recommendation method, apparatus, computer equipment, and storage medium to solve the problem that existing digital sales support systems have low accuracy in recommending sales scripts in complex negotiation scenarios, which in turn makes it difficult to improve sales conversion rates.

[0006] Firstly, a script recommendation method is provided, which adopts the following technical solution:

[0007] The system receives interaction information between client terminals and agent terminals. This interaction information includes inquiry information sent by client terminals and response information returned by agent terminals based on the inquiry information. Based on the inquiry information, it uses preset tracking technology to obtain multi-source behavioral data of the target customer corresponding to the client terminal from a multi-source data platform. Based on the multi-source behavioral data, it constructs a customer profile of the target customer. Based on the inquiry and response information, it obtains a basic script template from a preset script library. Based on the customer profile, it adjusts the basic script template to generate target scripts and recommends the target scripts to the agent terminals so that the agent terminals can interact with the target customers based on the target scripts.

[0008] Secondly, a script recommendation device is provided, which adopts the following technical solution:

[0009] The receiving module is used to receive the interaction information between the client terminal and the agent terminal. The interaction information includes the inquiry information sent by the client terminal and the reply information returned by the agent terminal based on the inquiry information.

[0010] The data acquisition module is used to acquire multi-source behavioral data of the target customer corresponding to the customer terminal from the multi-source data platform based on consultation information and using preset tracking technology.

[0011] The building module is used to construct customer profiles for target customers based on multi-source behavioral data;

[0012] The template acquisition module is used to retrieve basic script templates from a preset script library based on consultation and response information;

[0013] The adjustment module is used to adjust the basic script template based on customer profiles, generate target scripts, and recommend the target scripts to the agent terminals so that the agent terminals can interact with the target customers based on the target scripts.

[0014] Thirdly, a computer device is provided, which adopts the following technical solution:

[0015] The system receives interaction information between client terminals and agent terminals. This interaction information includes inquiry information sent by client terminals and response information returned by agent terminals based on the inquiry information. Based on the inquiry information, it uses preset tracking technology to obtain multi-source behavioral data of the target customer corresponding to the client terminal from a multi-source data platform. Based on the multi-source behavioral data, it constructs a customer profile of the target customer. Based on the inquiry and response information, it obtains a basic script template from a preset script library. Based on the customer profile, it adjusts the basic script template to generate target scripts and recommends the target scripts to the agent terminals so that the agent terminals can interact with the target customers based on the target scripts.

[0016] Fourthly, a computer-readable storage medium is provided, which adopts the following technical solution:

[0017] The system receives interaction information between client terminals and agent terminals. This interaction information includes inquiry information sent by client terminals and response information returned by agent terminals based on the inquiry information. Based on the inquiry information, it uses preset tracking technology to obtain multi-source behavioral data of the target customer corresponding to the client terminal from a multi-source data platform. Based on the multi-source behavioral data, it constructs a customer profile of the target customer. Based on the inquiry and response information, it obtains a basic script template from a preset script library. Based on the customer profile, it adjusts the basic script template to generate target scripts and recommends the target scripts to the agent terminals so that the agent terminals can interact with the target customers based on the target scripts.

[0018] Compared with existing technologies, the embodiments of this application have the following main advantages: By using data tracking technology to obtain multi-source behavioral data of target customers from a multi-source data platform and constructing a profile, the personalized characteristics of customers, such as family structure and asset status, are accurately captured, overcoming the problem that traditional systems rely solely on static knowledge bases and cannot parse unstructured data, resulting in incomplete customer profiles. Based on received consultation and response information, basic script templates are obtained from a preset script library and dynamically adjusted according to the customer profile to generate target scripts. This makes the scripts no longer limited to standardized templates but can be combined with specific customer characteristics. In complex negotiation scenarios, more targeted and flexible scripts can be provided, improving the accuracy of script recommendations, accurately meeting diverse customer needs, and effectively solving the problems of weak support capabilities and difficulty in improving sales conversion rates in existing systems under complex negotiation scenarios. Attached Figure Description

[0019] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying 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.

[0020] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0021] Figure 2 A flowchart of an embodiment of the speech recommendation method according to this application;

[0022] Figure 3 This is a schematic diagram of a structure of an embodiment of the speech recommendation device according to this application;

[0023] Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0025] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0026] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0027] like Figure 1 As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables.

[0028] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0029] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptop computer 1011, tablet computer 1012 or mobile phone 1013, terminal device 101 can also be e-book reader, MP3 player (Moving Picture Experts Group Audio Layer III), MP4 player (Moving Picture Experts Group Audio Layer IV), laptop computer and desktop computer, etc.

[0030] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0031] It should be noted that the script recommendation method provided in this application embodiment is generally executed by a server / terminal device, and correspondingly, the script recommendation device is generally set in the server / terminal device.

[0032] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0033] Continue to refer to Figure 2 The flowchart illustrates an embodiment of the conversation recommendation method according to this application. The conversation recommendation method includes the following steps:

[0034] Step S201: Receive interaction information between the client terminal and the agent terminal. The interaction information includes inquiry information sent by the client terminal and reply information returned by the agent terminal based on the inquiry information.

[0035] The customer terminal is the device through which customers interact with the sales system. It serves as the entry point for customers to initiate inquiries, receive information, and perform other operations, representing the point of entry for customer interaction with the system. During the financial and insurance sales process, customers use this terminal to consult with agents, obtain product information, and so on.

[0036] The agent terminal is the device used by the agent to handle sales operations. It serves as the terminal for the agent to receive customer inquiries, respond to customers, and manage the sales process, representing the entry point for interaction between the agent and the system and customers. During the sales process, the agent uses this terminal to receive customer information, generate sales scripts, and respond to customers. For example, an agent might use a computer as their agent terminal for business operations.

[0037] Interactive information refers to the data transmitted between the customer terminal and the agent terminal. It includes customer inquiries and agent responses, representing the content of communication between the two parties during the sales process. This information reflects the communication dynamics of the sales process and provides analytical support for the sales script recommendation engine.

[0038] Inquiry information refers to questions sent from the customer's terminal to the agent's terminal regarding insurance products or services. These are questions proactively raised by the customer during the financial insurance sales process, indicating the customer's focus on the product or service. They are used by agents to understand customer needs and provide a basis for subsequent responses and sales pitches. For example, a customer inquiring about the coverage of a critical illness insurance policy is an inquiry.

[0039] The response information is the answer returned by the agent terminal to the customer's inquiry. It is the agent's response after receiving a customer's inquiry, representing the agent's answer to the customer's question. It is used to resolve customer inquiries and advance the sales process.

[0040] Step S202: Based on the consultation information, use preset tracking technology to obtain multi-source behavioral data of the target customer corresponding to the customer terminal from the multi-source data platform.

[0041] Among them, event tracking technology is a technical means used to collect data. It collects user operation and behavior data by setting up tracking points at specific locations within a system, representing a technology that can accurately obtain user behavior information. For example, event tracking points can be set up on an insurance sales website to collect data on customer page browsing and product clicks.

[0042] A multi-source data platform is a system for storing and managing various types of data. It integrates data from different channels, such as basic customer information, transaction records, and social media data, representing a storage system that can provide comprehensive data support.

[0043] In this context, target customers are the individuals or groups that agents hope to close deals with during the financial insurance sales process. These are customers who have a need for or potential intention to purchase insurance, representing the primary service recipients of the insurance sales business. For example, a middle-aged customer with family protection needs is a target customer.

[0044] Multi-source behavioral data refers to the behavioral records of target customers across different platforms and scenarios. It includes customer behavior data from multiple channels such as financial platforms, social media, and e-commerce, representing the diversity of customer behavior. Examples include browsing history on insurance apps and spending records on e-commerce platforms.

[0045] Step S203: Construct a customer profile of the target customer based on multi-source behavioral data.

[0046] A customer profile is a model that comprehensively describes the characteristics and behaviors of a target customer. It is built upon multi-source behavioral data and includes information such as basic customer information, consumption habits, and risk preferences, representing a complete picture of the customer in the financial and insurance field. For example, a profile might include information such as the customer's age, income, and the insurance products they have purchased.

[0047] Step S204: Based on the consultation information and response information, obtain the basic script template from the preset script library.

[0048] The script library is a collection of pre-set scripts. It contains standardized script templates for different sales scenarios, customer types, and questions, representing a data set that provides rich script resources. For example, it includes scripts for scenarios such as customer price inquiries and competitor comparisons.

[0049] Among them, the basic sales script templates are general-purpose script formats preset in the script library for specific sales scenarios. They are standardized script frameworks developed based on common sales questions and scenarios, representing the initial form of the script. For example, a basic sales script template for a customer inquiring about the advantages of an insurance product includes framework content such as product feature introductions.

[0050] Step S205: Based on the customer profile, adjust the basic script template to generate the target script, and recommend the target script to the agent terminal so that the agent terminal can interact with the target customer based on the target script.

[0051] The adjustment process involves customizing the basic script template. Based on customer profiles and other information, it modifies the content, tone, and emphasis of the basic script template, demonstrating the ability to make the script more tailored to customer needs. This is used to generate more targeted and flexible target scripts.

[0052] Among them, the target script is the script ultimately generated and recommended to the agent terminal, which is tailored to the specific characteristics of the customer, the negotiation context, and market dynamics. The representative script is the script ultimately used by the agent to respond to the customer. It is used directly by the agent during the sales process to improve sales conversion rates. For example, in response to a customer's question about the price of critical illness insurance, a response script is generated that includes product value-added services and comprehensive protection clauses.

[0053] This application's embodiments utilize tracking technology to acquire multi-source behavioral data of target customers from a multi-source data platform and construct customer profiles. This accurately captures personalized customer characteristics, such as family structure and asset status, overcoming the problem of traditional systems relying solely on static knowledge bases and being unable to parse unstructured data, resulting in incomplete customer profiles. Based on received consultation and response information, a basic script template is retrieved from a pre-set script library and dynamically adjusted according to the customer profile to generate target scripts. This allows scripts to move beyond standardized templates and incorporate specific customer characteristics. In complex negotiation scenarios, this provides more targeted and flexible scripts, improving script recommendation accuracy, precisely meeting diverse customer needs, and effectively solving the problems of weak support capabilities and difficulty in improving sales conversion rates in existing systems during complex negotiation scenarios.

[0054] In some optional implementations of this embodiment, step 202, based on the consultation information, uses a preset data tracking technique to obtain multi-source behavioral data of the target customer corresponding to the customer terminal from the multi-source data platform, specifically including the following steps:

[0055] Behavioral analysis is performed on consultation information to identify key data elements associated with the behavior of target customers; based on the key data elements, a multi-source data platform for data collection is determined; and pre-set tracking technology is used to collect multi-source behavioral data of target customers corresponding to customer terminals from the multi-source data platform.

[0056] Key element data, derived from consultation information, is crucial data closely related to the behavior of target customers. It reflects the core behavioral characteristics and needs of target customers during the financial and insurance sales process. A multi-source data platform is a system for storing and managing data from different channels and of different types.

[0057] In one example, within the financial insurance sector, this embodiment is illustrated using an insurance company's auto insurance sales as an example. When a target customer inquires about auto insurance through a customer terminal, behavioral analysis can be performed on the received inquiry information. For instance, if a customer inquires about "the price and coverage of auto insurance for the first year of a new car," key data elements can be analyzed, including vehicle type (e.g., SUV), vehicle age (new car), and purchase price. Based on these elements, data is collected from car sales platforms (to obtain detailed vehicle information), traffic management department data platforms (to obtain potential risk data such as historical traffic violations), and the company's internal historical insurance data platform. Pre-set tracking technology is used to collect multi-source behavioral data from these platforms, such as the customer's browsing history of auto insurance prices for different car models on car sales platforms.

[0058] This application embodiment can accurately locate key element data related to the behavior of target customers by analyzing the behavior of consultation information, thereby determining a multi-source data platform and collecting multi-source behavioral data, which provides a rich and targeted data foundation for building a comprehensive and accurate customer profile.

[0059] In some optional implementations, step S203, which involves constructing a customer profile of the target customer based on multi-source behavioral data, specifically includes the following steps:

[0060] Using a pre-defined integration method, multi-source behavioral data is fused to obtain an initial customer profile of the target customer. Using a pre-defined graph representation learning technique, the relationships between the features in the initial customer profile are modeled and analyzed to generate multi-level labels. Based on historical evaluation records, the multi-level labels are evaluated to generate a customer profile of the target customer based on the evaluation results.

[0061] The integration method is a technical means of fusing multi-source behavioral data. It uses specific algorithms and rules to clean, transform, and merge data from different platforms and formats, eliminating redundancy and conflicts between data.

[0062] The initial customer profile is a preliminary description of customer characteristics obtained by integrating multi-source behavioral data. It is a preliminary model that includes basic customer information, behavioral preferences, consumption habits, and other characteristics, formed after preliminary processing and analysis of the collected multi-source behavioral data.

[0063] Among them, graph representation learning is a technique used by the profile optimization unit to analyze the relationships between features in the initial customer profile. It maps the features and relationships in the customer profile to a low-dimensional vector space to uncover the potential connections and structural information between features.

[0064] Among them, modeling analysis is the process of using graph representation learning techniques to process the relationships between features in the initial customer profile.

[0065] Among them, multi-level tags are a set of tags generated based on modeling analysis to describe the characteristics of target customers. It describes customer characteristics from different levels and perspectives, such as basic attribute tags, behavioral preference tags, and spending power tags, representing a data set that can comprehensively and meticulously describe customer characteristics.

[0066] Historical evaluation records serve as a reference for evaluating multi-level tags. They include information such as past evaluation results, feedback, and actual application effectiveness of customer profile tags.

[0067] In one example, within the financial insurance sector, this implementation is illustrated using an insurance company's auto insurance sales as an example. After collecting multi-source behavioral data from customers on various platforms using pre-defined tracking techniques, such as browsing records of different car insurance prices on car sales platforms, a pre-defined integration method can be used to fuse the collected multi-source behavioral data. For example, data on customer attention to additional car insurance services from car sales platforms, traffic accident rate data from traffic management department data platforms, and insurance plan data for similar car models from the company's internal historical insurance data platform can be cleaned, matched, and integrated to obtain an initial customer profile of the target customer, including information such as customer sensitivity to car insurance prices and their demand for additional services. Then, graph embedding technology is used to model and analyze the relationships between various features in the initial customer profile (such as vehicle type, customer age, and historical insurance records) to generate multi-level labels, such as "young high-risk SUV owner" and "focuses on a balance between price and basic coverage." Based on historical assessment records, these labels are evaluated. If it is found that the accuracy of the assessment of "high risk" needs to be improved, optimization is carried out by combining new data and feedback, and finally an accurate customer profile is generated.

[0068] This application embodiment can effectively integrate data from different sources through a preset integration method to obtain an initial customer profile, eliminating differences and conflicts between data and making the profile more complete and consistent. Graph representation learning technology is used to model and analyze the relationships between features in the initial customer profile, generating multi-level labels, and optimizing the labels based on historical evaluation records, resulting in a highly accurate customer profile. Based on this, the subsequent real-time script recommendation engine can dynamically recommend more targeted and flexible scripts in complex negotiation scenarios based on the accurate customer profile, improving the accuracy of script recommendations and thus effectively increasing sales conversion rates.

[0069] In some optional implementations, the step "using a preset integration method to fuse multi-source behavioral data to obtain an initial customer profile of the target customer" specifically includes the following steps:

[0070] Based on the key indicators of different data sources in the multi-source data platform, a preset integration method is used to assign weights to the multi-source behavioral data and determine the weights of the multi-source behavioral data; based on the weights, the multi-source behavioral data is weighted and summed to generate an initial customer profile of the target customer.

[0071] Key performance indicators (KPIs) are quantitative standards used to measure the importance of different data sources in a multi-source data platform. They reflect the value of data sources in building customer profiles through a series of technical indicators (such as data accuracy, completeness, timeliness, and relevance to business).

[0072] The weight is a parameter used to quantify the importance of multi-source behavioral data in building the initial customer profile. It is calculated based on key indicators from different data sources within the multi-source data platform.

[0073] In one example, within the financial insurance field, this implementation is illustrated using a life insurance sales example from an insurance company. For target customers inquiring about life insurance, behavioral analysis is performed on the inquiry information. For instance, if a customer inquires about the "coverage period and returns of whole life insurance," key data elements are identified, such as the customer's age, income level, and family structure. Based on these elements, data is collected from the customer's local social security data platform (to obtain basic medical insurance information), financial management platform (to obtain asset allocation information), and the company's internal historical insurance data platform. Pre-defined tracking techniques are used to collect multi-source behavioral data, such as records of customer interest in different financial products on the financial management platform. Based on key indicators from different data sources (such as the accuracy of the social security data platform, the timeliness of the financial management platform data, and the relevance of the company's internal historical insurance data to business operations), an analytic hierarchy process (AHP) is used to integrate and calculate the weights of the multi-source behavioral data. For example, the weight of the social security data platform is 0.3, the weight of the financial management platform is 0.4, and the weight of the company's internal historical insurance data platform is 0.3. Based on these weights, the multi-source behavioral data is weighted and summed. For example, by weighting and integrating customers' medical insurance information from the social security data platform, customers' asset return information from the financial management platform, and insurance plan information of similar customers from the company's internal historical insurance data platform, an initial customer profile of the target customer can be generated, which includes information such as the customer's expectations for insurance returns and risk tolerance.

[0074] This application embodiment can reasonably allocate weights to multi-source behavioral data from different data sources based on key indicators such as data accuracy, timeliness, and business relevance using a preset integration method. This allows for highlighting the contributions of important data sources and avoiding interference from irrelevant or low-quality data when constructing the initial customer profile. By performing weighted summation on the multi-source behavioral data based on weights, data from different platforms and of different types is effectively integrated to generate a more comprehensive and accurate initial customer profile.

[0075] In some optional implementations, step S205, based on the customer profile, adjusts the basic script template to generate the target script, specifically including the following steps:

[0076] Based on customer profiles and consultation information, the Monte Carlo simulation method is used to generate quotation information corresponding to the consultation information; the quotation information is embedded into the basic script template to obtain the optimized basic script template; the target script style is determined according to the behavioral tendencies in the customer profile; the optimized basic script template is customized and modified according to the target script style and the attribute characteristics in the customer profile to generate the script in the target script style, and the script in the target script style is determined as the target script.

[0077] Monte Carlo simulation is a mathematical and statistical method used to generate price quotes. It can fully consider various uncertainties and generate more reasonable and adaptive price quotes.

[0078] The pricing information is generated based on customer profiles and consultation information, using methods such as Monte Carlo simulation, and includes cost or price-related information corresponding to the consultation information. The target sales pitch style refers to the stylistic characteristics that the target sales pitch should possess, determined based on the behavioral tendencies identified in the customer profile.

[0079] Among these, attribute features are important data information used in customer profiling to describe the characteristics of a customer. They encompass various aspects such as the customer's age, gender, occupation, income, spending habits, and interests.

[0080] In one example, within the financial insurance sector, suppose a customer inquires about critical illness insurance products with an insurance agent. The inquiry includes the information: "30-year-old male, interested in cancer coverage, limited budget." Simultaneously, the system receives preliminary responses from the agent, such as mentioning common cancer coverage terms in the market. The system first integrates the inquiry and response information. For example, using natural language processing, key information such as "30-year-old male," "cancer coverage," and "limited budget" is extracted to generate template search information. Then, a pre-set critical illness insurance script library is searched, matching a basic script template containing these key elements, such as, "For 30-year-old male customers, we offer several critical illness insurance policies with specific cancer coverage; premiums can be flexibly adjusted according to your budget." Based on the constructed customer profile and inquiry information, Monte Carlo simulation is used, considering factors such as market competition, claim probability, and company profits, to simulate different pricing schemes and generate a suitable price for the customer, such as, "An annual premium of 5,000 yuan provides 500,000 yuan in cancer coverage." This pricing information is then embedded into the basic script template to obtain an optimized template. Based on the behavioral tendencies in the customer profile (assuming a conservative customer based on historical consultation records), the target sales pitch style is determined to be "conservative". Combining the customer's attributes (30-year-old male, limited budget), the optimized template is customized to generate the target sales pitch: "Dear customer, we recommend a critical illness insurance policy specifically for 30-year-old men, providing comprehensive and stable cancer protection. For only 5,000 yuan in annual premiums, you can obtain 500,000 yuan in coverage, safeguarding your health."

[0081] This application's embodiments utilize Monte Carlo simulation based on customer profiles and consultation information to generate pricing information and embed it into a basic sales script template. This ensures the script not only includes product information but also provides reasonable price suggestions, enhancing its commercial applicability and persuasiveness. By determining the target sales script style based on behavioral tendencies in the customer profile and customizing the optimized basic sales script template in conjunction with attribute characteristics, different styles of sales scripts can be generated to meet the communication preferences of different customers. This significantly improves the customer experience, increases customer acceptance of the sales script, and ultimately helps improve business conversion rates and customer satisfaction.

[0082] In some optional implementations, before receiving the interaction information between the client terminal and the agent terminal in step S201, the following steps are further included:

[0083] The system collects historical interaction data from multiple real customers and historical processing terminals across multiple dimensions. It then uses a pre-defined generative adversarial network (GAN) to train the historical interaction data and generate virtual customer data. Pre-defined personality setting parameters are obtained, and virtual customers are generated based on the virtual customer data and these parameters. Multiple interaction scenarios and corresponding target narrative templates are constructed based on the historical interaction data. Based on a pre-defined pressure configuration strategy, corresponding pressure parameters are set for each interaction scenario. When a selection operation by the agent terminal in response to a target interaction scenario is detected, corresponding interference variables are triggered based on the target pressure parameters corresponding to the target interaction scenario. Based on the interference variables and the target narrative template corresponding to the target interaction scenario, the virtual customer sends target interaction information to the agent terminal for simulated interaction training.

[0084] The historical processing terminal is the actual business terminal device or system that interacted with real customers during historical interactions. It is used to provide historical interaction data with real customers, providing a data foundation for virtual customer creation and stress testing.

[0085] Historical interaction data comprises various types of data generated during historical interactions between multiple real customers and historical processing terminals. This includes information such as customer inquiry content, terminal responses, interaction time, and interaction results.

[0086] Generative Adversarial Networks (GANs) are used to generate pre-defined algorithmic models for virtual customer data. They consist of a generator and a discriminator, which continuously optimize the quality of the generated data through a game-like interaction.

[0087] Virtual customer data refers to data generated by training historical interaction data using a generative adversarial network. It includes information such as the virtual customer's consultation content, preference characteristics, and interaction patterns. This data is used to generate virtual customers in conjunction with personality-defined parameters, providing a basis for simulating customer behavior in interactive simulation training.

[0088] Among them, the personality setting parameters are preset parameters used to define the personality and behavioral characteristics of virtual customers. They specify the characteristics of virtual customers in terms of attitude, tone, reaction speed, etc. during the interaction process.

[0089] Virtual customers refer to entities generated based on virtual customer data and personality setting parameters that can simulate real customer interaction behavior. They possess similar consultation needs, behavioral patterns, and personality traits to real customers. They are used to replace real customers in simulated interaction training, providing diverse interaction scenarios and challenges for agent terminals.

[0090] The interaction scenario refers to a specific context in which a simulated customer interacts with an agent terminal, constructed based on historical interaction data. It provides a concrete contextual framework for simulation interaction training, enabling the agent terminal to be trained in a near-realistic environment.

[0091] The target narrative template refers to a pre-defined template for each interaction scenario used to describe the interaction process and content. It specifies the logical order of the dialogue between the two parties, key questions, and key points for responses. It is used in simulated interaction training to provide a basic dialogue framework and content guidance for the interaction between virtual clients and agent terminals.

[0092] The stress configuration strategy refers to the preset rules and methods for setting corresponding stress parameters for each interaction scenario. This is used to reasonably set stress parameters so that simulation interaction training can simulate real-world business scenarios with different stress levels.

[0093] The stress parameter refers to the metric used to measure the stress level for each interaction scenario, based on a stress configuration strategy. It can be a specific value or range such as a time limit, question difficulty, or the intensity of customer emotions. In simulation interaction training, it is used to trigger interference variables based on the target stress parameter corresponding to the target interaction scenario, thereby increasing the complexity and difficulty of the interaction.

[0094] Among these, confounding variables refer to factors that trigger in the target interaction scenario to increase the complexity and uncertainty of the interaction. These can include sudden changes in customer requirements, additional questions raised, or technical malfunctions.

[0095] Among them, target interaction information refers to the information sent from the virtual client to the agent terminal after the agent terminal selects the target interaction scenario, based on the interference variables and the target narrative template corresponding to the target interaction scenario.

[0096] Simulated interaction training refers to the training process that simulates real-world interactions with agent terminals using virtual customers, interaction scenarios, target narrative templates, stress parameters, and interference variables. This is used to identify problems in agent terminals' use of scripts, business processing, and stress management, and through continuous training and optimization, improve their performance in actual business operations.

[0097] In one example, within the financial insurance sector, data can be collected from multiple real customers' historical interactions with insurance sales terminals. This includes information such as the customer's age, health status, budget, key concerns (coverage, claims conditions, etc.) when inquiring about critical illness or accident insurance, as well as the salesperson's responses. A pre-defined generative adversarial network is then used to train this historical interaction data, generating virtual customer data containing the virtual customer's potential consultation needs and characteristics. Simultaneously, pre-defined personality parameters are acquired, such as "cautious," "impulsive," and "professional." Based on the virtual customer data and personality parameters, virtual customers with different personality traits are generated. For example, a "cautious" virtual customer might pay particular attention to the details of the insurance terms, repeatedly inquiring about claims conditions and exclusions. Multiple interaction scenarios are constructed based on the historical interaction data, such as a customer suddenly changing their insurance budget or raising new coverage needs. A corresponding target narrative template is developed for each scenario, clarifying the interaction process and key issues. According to a pre-defined stress configuration strategy, stress parameters are set for each interaction scenario, such as time limits and question difficulty. When a target interaction scenario is selected by the agent terminal (i.e., the terminal used by insurance sales personnel), a disturbance variable is triggered based on the target stress parameter corresponding to the target interaction scenario. For example, in a scenario where a customer inquires about critical illness insurance, the disturbance variable "the customer becomes emotionally agitated and demands a quick protection plan after learning that someone they know has a critical illness" is randomly triggered. Combined with the target narrative template corresponding to the target interaction scenario, a virtual customer sends the target interaction information to the agent terminal. Through simulation interaction training, insurance sales personnel can familiarize themselves with various complex and stressful scenarios in advance, improving their coping abilities.

[0098] In one embodiment, multiple business scenarios can be determined based on historical interaction data. Multiple interaction scenarios are then constructed based on these business scenarios. A real customer profile is built for each real customer based on the historical interaction data. A correlation is established between the real customer profile and the multiple interaction scenarios. A narrative template is selected from a preset narrative template library based on the real customer profile. Based on the correlation, a narrative template corresponding to each interaction scenario is determined. The narrative template corresponding to each interaction scenario is customized to obtain a target narrative template for each interaction scenario. Based on a preset pressure configuration strategy, corresponding pressure parameters are set for each interaction scenario. Current market behavior data is obtained. When a selection operation of a target interaction scenario among multiple interaction scenarios is detected by the agent terminal, a corresponding interference variable is triggered based on the pressure parameter corresponding to the target interaction scenario. The interference variable is adjusted based on the market behavior data to obtain an adjusted interference variable. Based on the adjusted interference variable and the target narrative template corresponding to the target interaction scenario, target interaction information is sent to the agent terminal through a virtual customer to conduct simulation interaction training for the agent terminal.

[0099] This application embodiment can collect historical interaction data between real customers and historical processing terminals from multiple dimensions, and use generative adversarial networks to train and generate virtual customer data, then combine this with personality setting parameters to generate virtual customers. This process makes the behavior and characteristics of the generated virtual customers highly similar to real customers, providing rich and realistic simulation objects for simulation interaction training, ensuring that the training scenario is highly consistent with actual business. Interaction scenarios and target narrative templates are constructed based on historical interaction data, and pressure parameters are set according to a pressure configuration strategy. When the agent terminal selects an interaction scenario, interference variables are triggered, and target interaction information is sent through the virtual customer. This method can simulate complex and ever-changing actual business pressure scenarios, comprehensively testing the agent terminal's ability to respond to different pressure situations and handle business. Through this simulation training, the agent terminal can continuously accept challenges and training in a virtual environment, effectively improving its business level and adaptability, thereby increasing customer satisfaction and business conversion rate in actual business.

[0100] In some optional implementations, after step S206, based on the customer profile, the basic script template is adjusted to generate the target script, and the target script is recommended to the agent terminal so that the agent terminal can interact with the target customer based on the target script, the following steps are also included:

[0101] Receives dialogue records and customer feedback information between client terminals and agent terminals; uses a pre-set causal reasoning model to analyze the interaction results of the dialogue records and customer feedback information to obtain relationship analysis results; and generates interaction improvement suggestions based on the relationship analysis results.

[0102] Among them, the dialogue record is a record of all dialogue content generated during the interaction between the client terminal and the agent terminal.

[0103] Customer feedback information refers to the customer's evaluation and opinions on the interaction process and results after the interaction has ended.

[0104] Among them, the causal reasoning model refers to a pre-set algorithm or model used to analyze the causal relationship between dialogue records and customer feedback information. It is used to conduct in-depth analysis of dialogue records and customer feedback information to determine the causal relationship between factors such as dialogue and service and results such as customer satisfaction and decision-making, providing support for generating relationship analysis results and interaction improvement suggestions.

[0105] The relationship analysis results refer to the conclusions drawn from analyzing dialogue records and customer feedback information using a causal reasoning model.

[0106] Among them, the interaction improvement suggestions are based on the relationship analysis results and are used to optimize the interaction process and dialogue of the agent terminal.

[0107] In one example, within the financial insurance business, a critical illness insurance sales scenario is used. After the customer's terminal (the mobile app or computer used by the customer) and the agent's terminal (the business system used by the insurance agent) complete their interaction, the system can receive the dialogue records between the two parties. This includes the customer's description of their health status, family financial situation, and insurance budget, as well as the agent's introduction to the critical illness insurance coverage, premiums, and claims process. Simultaneously, customer feedback information is obtained, such as the customer's evaluation of the clarity of the agent's explanation and their satisfaction with the insurance product. A pre-set causal reasoning model is used to analyze the dialogue records and customer feedback. For example, the model finds that when the agent explains the types of diseases covered by critical illness insurance in detail with specific examples, the customer's understanding of the insurance product is higher, and their satisfaction increases accordingly. Conversely, when the agent overemphasizes premium discounts while neglecting coverage details, the customer is more likely to have doubts, leading to a decrease in satisfaction. Based on this, the analysis unit obtains relationship analysis results, clarifying the causal relationship between the way the communication is presented, the focus of the content, and the customer's level of understanding and satisfaction. Based on the relationship analysis results, suggestions for interaction improvement are generated. It is recommended that agents, when introducing critical illness insurance, add real-life cases to illustrate the scope of coverage, while balancing the ratio of premium to coverage information.

[0108] This application embodiment can comprehensively understand the actual situation during the interaction process by receiving dialogue records and customer feedback information after the interaction ends. Through in-depth data analysis using a pre-set causal reasoning model, it can accurately identify the causal relationships between factors such as the use of communication techniques and service methods, and results such as customer satisfaction and purchasing decisions, obtaining accurate relationship analysis results. The interaction improvement suggestions generated based on these results are highly targeted.

[0109] In one embodiment, after step S206, based on the customer profile, the basic script template is adjusted to generate the target script, and the target script is recommended to the agent terminal so that the agent terminal can interact with the target customer based on the target script, the following steps may be further included:

[0110] Receive historical interaction case data, perform experience mining on the historical interaction case data to obtain key experience data, and organize the key experience data in a structured manner to obtain a target practice model library.

[0111] Among them, historical interaction case data is a record of the interaction process and results between the client terminal and the agent terminal in the past.

[0112] Among them, key experience data refers to experience information that is of great guiding and reference value to the business, which is extracted from historical interaction case data.

[0113] The Target Practice Pattern Library is a database containing various successful practice patterns, obtained by the experience sharing unit after structuring key experience data. It allows agent terminals to quickly query and refer to practice patterns in the Target Practice Pattern Library when facing similar business situations, improving interaction and business processing capabilities.

[0114] In one example, within the financial insurance business, taking an insurance company as an example, the company receives a large amount of historical interaction case data. This data covers sales and service interaction records for different customer types (such as young office workers, middle-aged and elderly groups, etc.) and different insurance products (such as life insurance, health insurance, accident insurance, etc.). First, experience mining is performed on this historical interaction case data. For example, in processing a health insurance sales case for a young office worker, it was found that when the agent proactively mentioned that health insurance could complement medical insurance, covering medical expenses not covered by medical insurance, and explained the coverage corresponding to different sum insured amounts, the customer was more likely to understand the product's value, and their willingness to purchase significantly increased. For middle-aged and elderly groups, if life insurance is introduced in conjunction with scenarios such as children's education and retirement planning, emphasizing the importance of life insurance for family financial security, customers are more interested. Through similar analysis, numerous key experience data points are mined. Next, the key experience data is structured and organized. Successful sales techniques, communication skills, and problem-solving methods for different customer types and different insurance products are categorized and summarized to form a target practice model library. For example, for health insurance sales to young working professionals, a communication model of "complementary medical insurance + coverage amount guarantee" has been developed; for life insurance sales to middle-aged and elderly people, a communication model of "family economic protection + scenario integration" has been developed.

[0115] This application's embodiments can comprehensively collect various communication records and processing situations from past business by receiving historical interaction case data, providing rich material for experience mining. Through in-depth mining of this data, key experience data can be accurately extracted. After these key experience data are structured and organized into a target practice model library, insurance sales personnel can quickly query and refer to them. New employees can use this model library to quickly master effective communication methods, shorten the training cycle, and adapt to their work quickly. Experienced sales personnel can also gain new ideas from it, optimizing their own communication skills and service methods. This helps to unify team service standards, improve overall business performance, and enhance customer satisfaction.

[0116] It should be emphasized that, in order to further ensure the privacy and security of the aforementioned interactive information, multi-source behavioral data, basic dialogue templates, and target dialogue, the aforementioned interactive information, multi-source behavioral data, basic dialogue templates, and target dialogue can also be stored in a blockchain node.

[0117] 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. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together 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 can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0118] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0119] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0120] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of a script recommendation device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0121] like Figure 3 As shown, the script recommendation device 400 in this embodiment includes: a receiving module 401, a data acquisition module 402, a construction module 403, a template acquisition module 404, and an adjustment module 405. Wherein:

[0122] The receiving module 401 is used to receive the interaction information between the client terminal and the agent terminal. The interaction information includes the inquiry information sent by the client terminal and the reply information returned by the agent terminal based on the inquiry information.

[0123] The data acquisition module 402 is used to acquire multi-source behavioral data of the target customer corresponding to the customer terminal from the multi-source data platform based on the consultation information and using preset tracking technology.

[0124] Module 403 is used to build customer profiles of target customers based on multi-source behavioral data;

[0125] The template acquisition module 404 is used to acquire basic script templates from a preset script library based on consultation and response information.

[0126] The adjustment module 405 is used to adjust the basic script template based on the customer profile, generate the target script, and recommend the target script to the agent terminal so that the agent terminal can interact with the target customer based on the target script.

[0127] In this embodiment, multi-source behavioral data of target customers can be obtained from a multi-source data platform using data tracking technology to build customer profiles. This accurately captures personalized customer characteristics, such as family structure and asset status, overcoming the problem of incomplete customer profiles caused by traditional systems relying solely on static knowledge bases and being unable to parse unstructured data. Based on received consultation and response information, basic script templates are retrieved from a pre-set script library and dynamically adjusted according to the customer profile to generate target scripts. This allows scripts to move beyond standardized templates and incorporate specific customer characteristics. In complex negotiation scenarios, this provides more targeted and flexible scripts, improving script recommendation accuracy, accurately meeting diverse customer needs, and effectively solving the problems of weak support capabilities and difficulty in improving sales conversion rates in existing systems during complex negotiation scenarios.

[0128] In one embodiment, the data acquisition module 402 includes:

[0129] The analysis submodule is used to perform behavioral analysis on consultation information to identify key data elements that are associated with the behavior of target customers.

[0130] The platform identifies a sub-module used to determine the multi-source data platform for data collection based on key element data.

[0131] The data collection submodule is used to collect multi-source behavioral data of the target customer corresponding to the client terminal from the multi-source data platform using preset data collection techniques.

[0132] This application embodiment can accurately locate key element data related to the behavior of target customers by analyzing the behavior of consultation information, thereby determining a multi-source data platform and collecting multi-source behavioral data, which provides a rich and targeted data foundation for building a comprehensive and accurate customer profile.

[0133] In one embodiment, the construction module 403 includes:

[0134] The fusion submodule is used to fuse multi-source behavioral data using a preset fusion method to obtain an initial customer profile of the target customer.

[0135] The modeling submodule is used to model and analyze the relationships between features in the initial customer profile using a preset graph representation learning technique, and generate multi-level labels.

[0136] The evaluation submodule is used to evaluate multi-level tags based on historical evaluation records, and generate customer profiles of target customers based on the evaluation results.

[0137] This application embodiment can effectively integrate data from different sources through a preset integration method to obtain an initial customer profile, eliminating differences and conflicts between data and making the profile more complete and consistent. Graph representation learning technology is used to model and analyze the relationships between features in the initial customer profile, generating multi-level labels, and optimizing the labels based on historical evaluation records, resulting in a highly accurate customer profile. Based on this, the subsequent real-time script recommendation engine can dynamically recommend more targeted and flexible scripts in complex negotiation scenarios based on the accurate customer profile, improving the accuracy of script recommendations and thus effectively increasing sales conversion rates.

[0138] In one embodiment, the fusion submodule is further configured to assign weights to multi-source behavioral data based on key indicators of different data sources in the multi-source data platform using a preset integration method, and determine the weights of the multi-source behavioral data; and to generate an initial customer profile of the target customer by performing a weighted summation of the multi-source behavioral data based on the weights.

[0139] This application embodiment can reasonably allocate weights to multi-source behavioral data from different data sources based on key indicators such as data accuracy, timeliness, and business relevance using a preset integration method. This allows for highlighting the contributions of important data sources and avoiding interference from irrelevant or low-quality data when constructing the initial customer profile. By performing weighted summation on the multi-source behavioral data based on weights, data from different platforms and of different types is effectively integrated to generate a more comprehensive and accurate initial customer profile.

[0140] In one embodiment, the adjustment module 405 includes:

[0141] The generation submodule is used to generate quotation information corresponding to the consultation information based on customer profiles and consultation information, using the Monte Carlo simulation method;

[0142] The embedding submodule is used to embed the quotation information into the basic script template to obtain an optimized basic script template;

[0143] The style determination submodule is used to determine the target communication style based on the behavioral tendencies in the customer profile;

[0144] The modification submodule is used to customize the optimized basic script template based on the target script style and the attribute characteristics in the customer profile, generate the script in the target script style, and determine the script in the target script style as the target script.

[0145] This application's embodiments utilize Monte Carlo simulation based on customer profiles and consultation information to generate pricing information and embed it into a basic sales script template. This ensures the script not only includes product information but also provides reasonable price suggestions, enhancing its commercial applicability and persuasiveness. By determining the target sales script style based on behavioral tendencies in the customer profile and customizing the optimized basic sales script template in conjunction with attribute characteristics, different styles of sales scripts can be generated to meet the communication preferences of different customers. This significantly improves the customer experience, increases customer acceptance of the sales script, and ultimately helps improve business conversion rates and customer satisfaction.

[0146] In one embodiment, the script recommendation device 400 further includes:

[0147] The data acquisition module is used to collect historical interaction data between multiple real customers and historical processing terminals in a multi-dimensional manner.

[0148] The training module is used to train on historical interaction data using a pre-defined generative adversarial network to generate virtual customer data.

[0149] The generation module is used to obtain preset personality setting parameters and generate virtual customers based on virtual customer data and personality setting parameters;

[0150] The template building module is used to build multiple interaction scenarios and a target narrative template for each interaction scenario based on historical interaction data.

[0151] The settings module is used to set corresponding pressure parameters for each interaction scenario based on a preset pressure configuration strategy.

[0152] The triggering module is used to trigger the corresponding interference variable based on the target pressure parameter corresponding to the target interaction scenario when the agent terminal detects a selection operation of the target interaction scenario among multiple interaction scenarios.

[0153] The sending module is used to send target interaction information to the agent terminal through a virtual client based on the interference variables and the target narrative template corresponding to the target interaction scenario, so as to conduct simulation interaction training of the agent terminal.

[0154] This application embodiment can collect historical interaction data between real customers and historical processing terminals from multiple dimensions, and use generative adversarial networks to train and generate virtual customer data, then combine this with personality setting parameters to generate virtual customers. This process makes the behavior and characteristics of the generated virtual customers highly similar to real customers, providing rich and realistic simulation objects for simulation interaction training, ensuring that the training scenario is highly consistent with actual business. Interaction scenarios and target narrative templates are constructed based on historical interaction data, and pressure parameters are set according to a pressure configuration strategy. When the agent terminal selects an interaction scenario, interference variables are triggered, and target interaction information is sent through the virtual customer. This method can simulate complex and ever-changing actual business pressure scenarios, comprehensively testing the agent terminal's ability to respond to different pressure situations and handle business. Through this simulation training, the agent terminal can continuously accept challenges and training in a virtual environment, effectively improving its business level and adaptability, thereby increasing customer satisfaction and business conversion rate in actual business.

[0155] In one embodiment, the script recommendation device 400 further includes:

[0156] The recording and receiving module is used to receive the dialogue records and customer feedback information between the client terminal and the agent terminal;

[0157] The analysis module is used to analyze the interaction results of dialogue records and customer feedback information using a preset causal reasoning model to obtain relationship analysis results.

[0158] The suggestion generation module is used to generate interactive improvement suggestions based on the relationship analysis results.

[0159] This application embodiment can comprehensively understand the actual situation during the interaction process by receiving dialogue records and customer feedback information after the interaction ends. Through in-depth data analysis using a pre-set causal reasoning model, it can accurately identify the causal relationships between factors such as the use of communication techniques and service methods, and results such as customer satisfaction and purchasing decisions, obtaining accurate relationship analysis results. The interaction improvement suggestions generated based on these results are highly targeted.

[0160] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0161] Computer device 6 includes a memory 61, a processor 62, and a network interface 63 that are interconnected via a system bus. It should be noted that only computer device 6 with memory 61, processor 62, and network interface 63 is shown in the figure; however, it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described herein is a device capable of automatically performing 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 (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0162] Computer devices can include desktop computers, laptops, handheld computers, and cloud servers. These devices allow for human-computer interaction with users through methods such as keyboards, mice, remote controls, touchpads, or voice-activated devices.

[0163] The memory 61 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), 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, magnetic disk, optical disk, etc. In some embodiments, the memory 61 may be an internal storage unit of the computer device 6, such as the hard disk or memory of the computer device 6. In other embodiments, the memory 61 may also be an external storage device of the computer device 6, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 6. Of course, the memory 61 may include both the internal storage unit and the external storage device of the computer device 6. In this embodiment, the memory 61 is typically used to store the operating system and various application software installed on the computer device 6, such as computer-readable instructions for speech recommendation methods. In addition, the memory 61 may also be used to temporarily store various types of data that have been output or will be output.

[0164] In some embodiments, processor 62 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. Processor 62 is typically used to control the overall operation of computer device 6. In this embodiment, processor 62 is used to execute computer-readable instructions stored in memory 61 or to process data, such as computer-readable instructions for executing a speech recommendation method.

[0165] The network interface 63 may include a wireless network interface or a wired network interface, which is typically used to establish a communication connection between the computer device 6 and other electronic devices.

[0166] This application's embodiments utilize tracking technology to acquire multi-source behavioral data of target customers from a multi-source data platform and construct customer profiles. This accurately captures personalized customer characteristics, such as family structure and asset status, overcoming the problem of traditional systems relying solely on static knowledge bases and being unable to parse unstructured data, resulting in incomplete customer profiles. Based on received consultation and response information, a basic script template is retrieved from a pre-set script library and dynamically adjusted according to the customer profile to generate target scripts. This allows scripts to move beyond standardized templates and incorporate specific customer characteristics. In complex negotiation scenarios, this provides more targeted and flexible scripts, improving script recommendation accuracy, precisely meeting diverse customer needs, and effectively solving the problems of weak support capabilities and difficulty in improving sales conversion rates in existing systems during complex negotiation scenarios.

[0167] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the above-described speech recommendation method.

[0168] This application's embodiments utilize tracking technology to acquire multi-source behavioral data of target customers from a multi-source data platform and construct customer profiles. This accurately captures personalized customer characteristics, such as family structure and asset status, overcoming the problem of traditional systems relying solely on static knowledge bases and being unable to parse unstructured data, resulting in incomplete customer profiles. Based on received consultation and response information, a basic script template is retrieved from a pre-set script library and dynamically adjusted according to the customer profile to generate target scripts. This allows scripts to move beyond standardized templates and incorporate specific customer characteristics. In complex negotiation scenarios, this provides more targeted and flexible scripts, improving script recommendation accuracy, precisely meeting diverse customer needs, and effectively solving the problems of weak support capabilities and difficulty in improving sales conversion rates in existing systems during complex negotiation scenarios.

[0169] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of the various embodiments of this application.

[0170] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

[0171] The software tools or components not belonging to our company that appear in the embodiments of this application are merely examples and do not represent actual use.

Claims

1. A method for recommending sales scripts, characterized in that, Includes the following steps: Receive interaction information between a client terminal and an agent terminal, the interaction information including inquiry information sent by the client terminal and reply information returned by the agent terminal based on the inquiry information; Based on the consultation information, multi-source behavioral data of the target customer corresponding to the customer terminal is obtained from the multi-source data platform using preset data tracking technology. Based on the multi-source behavioral data, a customer profile of the target customer is constructed; Based on the consultation information and the response information, a basic script template is obtained from a preset script library; Based on the customer profile, the basic script template is adjusted to generate a target script, which is then recommended to the agent terminal so that the agent terminal can interact with the target customer based on the target script.

2. The method according to claim 1, characterized in that, The step of obtaining multi-source behavioral data of the target customer corresponding to the customer terminal from a multi-source data platform based on the consultation information using preset tracking technology specifically includes: Behavioral analysis is performed on the consultation information to identify key data elements associated with the behavior of the target customer; Based on the aforementioned key element data, a multi-source data platform for data collection is determined. Using a pre-defined data collection technique, multi-source behavioral data of the target customer corresponding to the customer terminal is collected from the multi-source data platform.

3. The method according to claim 1, characterized in that, The step of constructing the customer profile of the target customer based on the multi-source behavioral data specifically includes: Using a preset integration method, the multi-source behavioral data are fused to obtain an initial customer profile of the target customer; Using a pre-defined graph representation learning technique, the relationships between features in the initial customer profile are modeled and analyzed to generate multi-level labels; Based on historical evaluation records, the multi-level tags are evaluated to generate a customer profile of the target customer based on the evaluation results.

4. The method according to claim 3, characterized in that, The step of fusing the multi-source behavioral data using a preset integration method to obtain the initial customer profile of the target customer specifically includes: Based on the key indicators of different data sources in the multi-source data platform, a preset integration method is used to assign weights to the multi-source behavioral data and determine the weights of the multi-source behavioral data. The multi-source behavioral data is weighted and summed based on the weights to generate an initial customer profile of the target customer.

5. The method according to claim 1, characterized in that, The step of adjusting the basic script template based on the customer profile to generate the target script specifically includes: Based on the customer profile and the consultation information, the Monte Carlo simulation method is used to generate the quotation information corresponding to the consultation information; The quotation information is embedded into the basic script template to obtain the optimized basic script template; Based on the behavioral tendencies in the customer profile, determine the target communication style; Based on the target sales pitch style and the attribute characteristics in the customer profile, the optimized basic sales pitch template is customized and modified to generate a sales pitch in the target sales pitch style, and the sales pitch in the target sales pitch style is determined as the target sales pitch.

6. The method according to claim 1, characterized in that, Before the step of receiving the interaction information between the client terminal and the agent terminal, the method further includes: Collect historical interaction data from multiple real customers and historical processing terminals in historical interactions from multiple dimensions; A pre-defined generative adversarial network is used to train the historical interaction data to generate virtual customer data; Obtain preset personality setting parameters, and generate virtual customers based on the virtual customer data and the personality setting parameters; Based on the historical interaction data, multiple interaction scenarios and a target narrative template corresponding to each interaction scenario are constructed. Based on the preset pressure configuration strategy, set corresponding pressure parameters for each interaction scenario; When the agent terminal is detected to be selecting a target interaction scenario among the multiple interaction scenarios, the corresponding interference variable is triggered based on the target pressure parameter corresponding to the target interaction scenario. Based on the interference variables and the target narrative template corresponding to the target interaction scenario, the virtual client sends target interaction information to the agent terminal to conduct simulation interaction training for the agent terminal.

7. The method according to claim 1, characterized in that, After the steps of adjusting the basic script template based on the customer profile to generate a target script, and recommending the target script to the agent terminal so that the agent terminal can interact with the target customer based on the target script, the method further includes: Receive the dialogue records and customer feedback information between the client terminal and the agent terminal; Using a pre-defined causal reasoning model, the interaction results of the dialogue records and customer feedback information are analyzed to obtain relationship analysis results; Based on the relationship analysis results, suggestions for improving the interaction are generated.

8. A script recommendation device, characterized in that, include: The receiving module is used to receive interaction information between the client terminal and the agent terminal. The interaction information includes inquiry information sent by the client terminal and reply information returned by the agent terminal based on the inquiry information. The data acquisition module is used to acquire multi-source behavioral data of the target customer corresponding to the customer terminal from the multi-source data platform based on the consultation information and using preset tracking technology. A construction module is used to construct a customer profile of the target customer based on the multi-source behavioral data; The template acquisition module is used to acquire basic dialogue templates from a preset dialogue database based on the consultation information and the response information. The adjustment module is used to adjust the basic script template based on the customer profile, generate a target script, and recommend the target script to the agent terminal so that the agent terminal can interact with the target customer based on the target script.

9. A computer device, characterized in that, It includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the speech 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 computer-readable instructions that, when executed by a processor, implement the steps of the speech recommendation method as described in any one of claims 1 to 7.

Images