Conversation strategy generation method and device based on intention recognition, equipment and medium

Abstract

The invention relates to the technical field of artificial intelligence and the field of financial science and technology, and discloses a dialogue strategy generation method based on intention recognition, and the method comprises the steps: generating a user portrait through obtaining user basic information; and analyzing the service information of the user and the preliminary user portrait, and generating a preliminary dialogue strategy in combination with the to-be-handled service field. The user feedback information is processed through the intention recognition model, multiple service intentions of the client in the target service field are analyzed, confidence analysis is conducted on the service intentions, and effective target intentions are recognized. And according to the effective target intention, adjusting the preliminary dialogue strategy, generating a subsequent dialogue strategy, and gradually optimizing the user experience. Through identification and analysis of the user intention, the limitation that the existing call-out system cannot accurately respond to the customer demand is overcome, through identification and confidence analysis of the effective target intention, the priority demand of the customer can be accurately judged, the dialogue strategy can be adjusted based on the personalized demand of the customer, and the user experience is improved. Therefore, the effectiveness and the reach rate of the outbound call are improved.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence technology and financial technology, and in particular to a method, device, equipment and storage medium for generating a dialogue strategy based on intent recognition. Background Art

[0002] In the financial industry, especially the insurance industry, customer service plays a core role in maintaining customer relationships and sustainable business development. To ensure that customers receive timely information on claims, renewals, preservation and other services, insurance companies usually use outbound call systems to remind customers. However, existing outbound call systems have obvious deficiencies in terms of accuracy and effectiveness of customer service.

[0003] The current outbound call system mainly relies on pre-set fixed scripts. Regardless of the actual needs or feedback of the customer, the system will execute according to the established process. The main problems of this static strategy include the lack of in-depth understanding of customer intentions. When providing services, the existing outbound call system is difficult to accurately match the current needs and business scenarios of customers, which often leads to a disconnect between the communication content and customer needs.

[0004] Existing outbound call systems mainly rely on direct feedback from customers during conversations, and fail to deeply analyze customers' potential needs or business intentions. Existing technologies can only provide preset reminders based on the limited information provided by customers, lack in-depth mining of customer feedback data, and fail to capture customers' potential needs or priorities. Because existing technologies lack the ability to analyze customers' historical behaviors or preferences, outbound call content is usually standardized and fails to effectively adjust to customers' personalized needs.

[0005] Although the outbound call system can obtain customer feedback during the call, the use of this information is often limited to simple storage or recording, and it fails to play a role in subsequent communication. In addition, the existing system cannot adjust the content of the speech according to customer feedback, which means that even if the customer is dissatisfied with or does not understand the content of the outbound call, the system will still use the same fixed speech in subsequent communication, resulting in reduced communication efficiency.

[0006] Therefore, the existing outbound call system cannot optimize communication strategies based on customer feedback and changes in demand, which limits the response speed and flexibility of outbound call services and cannot effectively adapt to customer needs. Summary of the invention

[0007] The main purpose of the present invention is to provide a method, device, equipment and storage medium for generating a dialogue strategy based on intent recognition, aiming to solve the technical problems that the prior art cannot adjust the content of outbound calls based on the actual needs and feedback of customers, lacks in-depth mining and response to customer intentions, resulting in low outbound call efficiency and insufficient personalized service capabilities.

[0008] To achieve the above object, the present invention provides a method for generating a dialogue strategy based on intention recognition, comprising:

[0009] Obtaining basic information of the user, and generating a preliminary user profile based on the basic information;

[0010] Obtaining business information of users in the pending business area, analyzing the business information and preliminary user portraits, and generating a first conversation strategy for the pending business based on the analysis results;

[0011] Execute the first dialogue strategy and obtain user feedback information;

[0012] Analyzing the user feedback information through an intention recognition model to identify multiple business intentions of the user in the target business field;

[0013] Performing confidence analysis on the multiple business intentions, and taking the business intentions with confidence higher than a preset threshold as valid target intentions;

[0014] Based on the effective target intention, the initial dialogue strategy is adjusted to generate a subsequent dialogue strategy.

[0015] Furthermore, in order to achieve the above-mentioned object, the present invention provides a dialogue strategy generation device based on intention recognition, comprising:

[0016] A user portrait generation module, used to obtain basic information of the user and generate a preliminary user portrait based on the basic information;

[0017] A first conversation strategy generation module is used to obtain business information of users in the pending business field, analyze the business information and preliminary user portraits, and generate a first conversation strategy for the pending business based on the analysis results;

[0018] A feedback information collection module, used to execute the first dialogue strategy and obtain user feedback information;

[0019] A user intention recognition module, which analyzes the user feedback information through an intention recognition model to identify multiple business intentions of the user in the target business field;

[0020] A target intention determination module, used to perform confidence analysis on the multiple business intentions, and take the business intentions with confidence higher than a preset threshold as valid target intentions;

[0021] A subsequent dialogue strategy generation module is used to adjust the initial dialogue strategy based on the effective target intention to generate a subsequent dialogue strategy.

[0022] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a computer device, which includes a memory, a processor, and a dialogue strategy generation program based on intent recognition stored in the memory and executable on the processor, wherein the dialogue strategy generation program based on intent recognition, when executed by the processor, implements the steps of the dialogue strategy generation method based on intent recognition as described above.

[0023] Furthermore, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, on which is stored a dialogue strategy generation program based on intent recognition, and when the dialogue strategy generation program based on intent recognition is executed by a processor, the steps of the dialogue strategy generation method based on intent recognition as described above are implemented.

[0024] Beneficial effects: The present invention relates to the fields of artificial intelligence technology and financial technology, and discloses a method for generating a dialogue strategy based on intent recognition, which generates a user portrait by acquiring basic user information. By collecting the user's personal information, economic information and behavior patterns, the user's business information and preliminary user portrait are analyzed after processing, and a preliminary dialogue strategy is generated in combination with the pending business field. The user feedback information is processed by the intent recognition model, and the customer's multiple business intentions in the target business field are analyzed, and the business intentions are analyzed for confidence. The business intentions with a confidence higher than a preset threshold are identified as valid target intentions. According to the effective target intention, the preliminary dialogue strategy is adjusted, and the subsequent dialogue strategy is generated to gradually optimize the user experience. The present invention overcomes the limitation that the existing outbound call system cannot accurately respond to customer needs by identifying and analyzing user intentions. Compared with the traditional outbound call method that relies on fixed words, the dialogue strategy can be adjusted based on the customer's personalized needs, thereby improving the effectiveness and reach of the outbound call. Specifically, by deeply exploring the actual needs and preferences of customers, customer dissatisfaction caused by repeated outbound calls or irrelevant content can be reduced. In addition, through the identification and confidence analysis of effective target intentions, the customer's priority needs can be accurately judged, thereby improving the response speed and accuracy of outbound call services. It significantly reduced the frequency of ineffective communication, improved the overall efficiency of the outbound call system, and enhanced customer satisfaction with the service, effectively supporting the maintenance of customer relationships and the sustainable development of the business. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0026] Figure 1 A schematic diagram of an application environment of a method for generating a dialogue strategy based on intent recognition in an embodiment of the present invention;

[0027] Figure 2 A flowchart of an embodiment of a method for generating a dialogue strategy based on intent recognition according to the present invention;

[0028] Figure 3 It is a functional module diagram of a preferred embodiment of a device for generating a dialogue strategy based on intention recognition according to the present invention;

[0029] Figure 4 A schematic diagram of the structure of a computer device in one embodiment of the present invention;

[0030] Figure 5 FIG. 4 is another schematic diagram of the structure of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0031] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0032] The method for generating a dialogue strategy based on intention recognition provided by the embodiment of the present invention can be applied in the following aspects: Figure 1 In the application environment of, the user end communicates with the server end through the network. The server end can obtain the user's basic information through the user end and generate a user portrait. By collecting the user's personal information, economic information and behavior pattern, the user's business information and preliminary user portrait are analyzed after processing, and a preliminary dialogue strategy is generated in combination with the pending business field. The user feedback information is processed by the intention recognition model, and the customer's multiple business intentions in the target business field are analyzed, and the business intentions are analyzed with confidence. The business intentions with a confidence higher than the preset threshold are identified as valid target intentions. According to the effective target intention, the preliminary dialogue strategy is adjusted, and the subsequent dialogue strategy is generated to gradually optimize the user experience. The present invention overcomes the limitation that the existing outbound call system cannot accurately respond to customer needs by identifying and analyzing user intentions. Compared with the traditional outbound call method that relies on fixed words, the dialogue strategy can be adjusted based on the customer's personalized needs, thereby improving the effectiveness and reach of outbound calls. Specifically, by deeply exploring the actual needs and preferences of customers, customer dissatisfaction caused by repeated outbound calls or irrelevant content can be reduced. In addition, through the identification and confidence analysis of effective target intentions, the customer's priority needs can be accurately judged, thereby improving the response speed and accuracy of outbound call services. The frequency of ineffective communication is significantly reduced, the overall efficiency of the outbound call system is improved, and customer satisfaction with the service is enhanced, effectively supporting the maintenance of customer relationships and the sustainable development of the business. Among them, the user end can be but is not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices. The server end can be implemented with an independent server or a server cluster composed of multiple servers. The present invention is described in detail below through specific embodiments.

[0033] See also Figure 2 , Figure 2This is a flowchart of an embodiment of a method for generating a dialogue strategy based on intention recognition provided by the present invention. It should be noted that although a logical sequence is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0034] like Figure 2 As shown, the method for generating a dialogue strategy based on intention recognition proposed in the present invention includes the following steps:

[0035] S10, obtaining basic information of the user, and generating a preliminary user portrait based on the basic information;

[0036] In this embodiment, the basic information of the user generally includes the user's personal information, economic information and behavior pattern. Personal information includes basic information such as name, age, gender, occupation, etc.; economic information may include income, consumption capacity, investment behavior, etc.; behavior pattern involves the user's service usage frequency, purchase history, access channels, etc.

[0037] The system automatically collects this information and stores it in the data management platform by connecting to multiple data sources (such as user-filled forms, third-party data sources, historical transaction records, etc.). In order to ensure the accuracy of the data, the system can also verify the collected information to ensure the validity of the user-entered data.

[0038] Based on the user's basic information, the system generates a preliminary user profile. This profile is based on the user's personal information, economic status and behavior patterns to form an overall user profile to describe the customer's characteristics. These characteristics can be used for subsequent business needs analysis and customized services.

[0039] The system integrates and classifies various types of user information through machine learning or rule matching models to generate a preliminary user profile. The process of generating a user profile may include technical means such as data clustering, statistical analysis, and rule matching. Each user profile contains several attributes, such as the customer's financial ability, risk preference, behavior pattern, and other information. The data processing tool maps the user's basic data to a specific classification system and outputs graphical user characteristics.

[0040] Example: In the insurance field, a customer fills in relevant personal information, such as age, occupation, and income, when applying for insurance. The system can also obtain the customer's past claims records and the frequency of interaction with the insurance company from historical data. Based on this data, the system generates a preliminary user profile of the customer, indicating that the customer has applied for claims twice in the past three years, is a user with a low risk preference, and is in good financial condition, suitable for recommending long-term life insurance and financial management insurance products. Through these analyses, insurance companies can provide more accurate insurance product recommendations based on user profiles, reducing the risk of customer loss caused by blind sales.

[0041] By obtaining basic information about users and generating preliminary user portraits, the system can effectively identify customers’ personalized needs and behavioral preferences, providing important data support for subsequent business strategy generation. It can provide customized services based on the actual situation of customers, thereby improving customer satisfaction and the service efficiency of insurance companies.

[0042] S20, obtaining business information of users in the pending business field, analyzing the business information and the preliminary user portrait, and generating a first conversation strategy for the pending business based on the analysis result;

[0043] In this embodiment, business information related to the to-do items is extracted from the user's business records. Business information may include the user's historical transaction records, the current business process, and pending matters (such as insurance claims, renewals, preservation, etc.). Through these business data, the system can understand the user's current business needs.

[0044] The system obtains the user's business information by connecting to business systems (such as insurance business management systems, CRM systems, etc.). This data may include unfinished claims applications, policy renewal reminders, unprocessed complaint records, etc. Through the API interface, the system automatically aggregates this information from multiple business systems and stores and organizes it in a unified manner.

[0045] Based on the business information obtained, the system combines the business information with the preliminary user portrait for analysis. By comparing the user's current business situation with the personal characteristics in the user portrait, the system can identify the user's demand priorities and business preferences in order to generate a more targeted dialogue strategy.

[0046] Through data analysis tools (such as data mining models, rule matching models, etc.), the system can analyze the correlation between the user's business information and the preliminary user portrait. The analysis process may include analysis of the user's historical behavior, evaluation of the user's current business priority, etc. Based on the analysis results, the system identifies which business information the user is currently most concerned about, or which services can bring a better experience to the user.

[0047] After analyzing the user's business information and profile, the system will generate a first conversation strategy that matches the user's current needs. The strategy defines how to communicate with the user, which matters to prioritize, and how to guide the user to complete the relevant business process during the conversation.

[0048] The system generates a personalized conversation strategy through a preset speech strategy library combined with the user's business analysis results. According to the user's priority business needs, the system can first remind the user to deal with urgent matters (such as expiring insurance policies) in the conversation, and then deal with secondary matters. This strategy is automatically generated by the system and can be dynamically adjusted according to user needs.

[0049] Example: In the insurance field, a customer's policy is about to expire, and the customer has unprocessed claims. The system first obtains these to-do items from the business system, and combined with the customer's portrait analysis, determines that the customer's priority demand for policy renewal is higher than claim settlement. The first conversation strategy generated by the system will first remind the customer to renew, and then remind the claim progress after the renewal is successful. This not only ensures the timely processing of the customer's priority business, but also avoids customer dissatisfaction caused by multiple outbound calls. In this way, insurance companies can manage customer needs more effectively and provide more targeted services.

[0050] By combining the user's business information with the preliminary user portrait for analysis, the system can generate a more accurate first conversation strategy to ensure that the conversation content is highly matched with the user's actual needs. It can effectively avoid irrelevant reminders and information redundancy, and improve the customer's service experience and communication efficiency. At the same time, by giving priority to the business that users are concerned about, it can reduce the number of outbound calls, reduce customer antipathy, and improve customer satisfaction.

[0051] S30, executing the first dialogue strategy and obtaining user feedback information;

[0052] In this embodiment, executing the first conversation strategy means that the system communicates with the user based on the conversation content generated by the analysis. In the insurance business scenario, this may include the first interaction with the user through outbound calls, SMS, email, or online messages. The strategy is based on the user's business information and profile to ensure that the communication content is relevant to the customer's needs.

[0053] Based on the generated first conversation strategy, the system selects the appropriate communication channel (such as phone, SMS, etc.) and pushes the preset conversation content to the user. The system can perform outbound calls through the call center platform or customer management system, and pass the message to the user terminal through API or interface calls. The conversation content can be personalized according to the user's business scenario (such as claim reminder, renewal reminder).

[0054] After executing the first conversation strategy, the system will collect user feedback information. This feedback can be the customer's response to the first communication content, which may include the user's questions, requests, or other interactive content. Feedback information is crucial for subsequent strategy adjustments, and can help the system understand whether the customer is satisfied with the current communication content and understand the user's current demand status.

[0055] User feedback information is returned to the system through different interaction channels. Feedback from phone calls can be converted into text through voice recognition technology, or directly fed back through options entered by the customer. For SMS or email, feedback information is the user's text reply, which is automatically recorded and parsed by the system. Through text analysis or simple option processing, the system can obtain clear feedback from users.

[0056] Feedback here does not only refer to the user's explicit response in the conversation (such as "yes" or "no"), but can also be the content of the conversation itself. For example, the user's reply or response is also part of the conversation and can be interpreted by the system as feedback information, thereby helping the system adjust subsequent strategies.

[0057] The system directly extracts user feedback from the conversation through speech recognition technology or text analysis technology. For example, in a telephone conversation, the user's verbal response is converted into text and stored; while in SMS or email exchanges, the system directly extracts text feedback. This feedback information is not only used to record the user's current demand status, but also can be used as a basis for adjusting the content of subsequent conversations.

[0058] Example: During the insurance claims process, the user receives an outbound call from the system, which prompts the user to complete the claim application and upload the required materials. The user directly responds in the conversation that the materials have been uploaded, and the system uses the response as feedback information to automatically update the business process and avoid reminding the user to upload materials again. This improves the user experience and reduces the burden of repeated communication.

[0059] By executing the first conversation strategy and obtaining user feedback, the system can timely understand the actual needs of users and further adjust the service strategy based on the feedback information, thus improving the accuracy and timeliness of communication and avoiding the occurrence of ineffective communication.

[0060] S40, analyzing the user feedback information through an intention recognition model to identify multiple business intentions of the user in the target business field;

[0061] In this embodiment, after the system obtains user feedback information, the feedback content is analyzed through the intent recognition model. The intent recognition model can be part of the natural language processing (NLP) technology, which can extract the user's potential needs or intentions from the user's language or text. By analyzing the user's feedback, the system can understand the specific operations that the user wants to perform or the needs expressed.

[0062] The system first converts the user's feedback information into processable text data through speech recognition or text analysis technology. Then, the intent recognition model performs semantic analysis on these text data and identifies the user's business intent through technologies such as keyword matching, context understanding and grammatical analysis. The intent recognition model can be trained based on machine learning algorithms, such as using pre-labeled dialogue data sets, to improve the accuracy and precision of recognition.

[0063] After analyzing the user's feedback, the system can identify the user's current multiple business intentions. These intentions may involve multiple business needs, such as the user may submit a claim application, renewal reminder, preservation needs, etc. The system can extract the user's different intentions through the intention recognition model and process them according to priority.

[0064] Through the intent recognition model, the system can not only identify the user's main business intent, but also capture secondary or hidden intents. For example, a user may mention the need to process a claim while also mentioning the renewal of an insurance policy during a conversation. The intent recognition model will identify these different business intents based on context, keywords, and contextual relationships. The system will classify and sort each intent to ensure that it is handled by priority.

[0065] Example: In the insurance claims process, users interact with the system via phone or online chat, and request to check the claim status and inquire about policy renewal. The system uses the intent recognition model to identify the user's two main needs: one is to check the claim status, and the other is to remind the policy renewal. The system will handle these two needs at the same time, providing users with a one-stop solution to avoid users making calls or sending inquiry emails again, thereby improving the user experience.

[0066] By analyzing user feedback information through the intent recognition model, the system can accurately extract multiple business intentions of users from their conversations or text exchanges. This improves the system's ability to respond to complex customer needs and avoids the limitation of traditional outbound call systems that can only handle single needs in one direction. The recognition of multiple business intentions allows the system to handle multiple businesses at the same time, optimizes service efficiency, reduces the need for customers to communicate multiple times, and improves customer experience and satisfaction.

[0067] S50, performing confidence analysis on the multiple business intentions, and taking the business intentions with confidence higher than a preset threshold as valid target intentions;

[0068] In this embodiment, after identifying multiple business intentions of the user, the system performs confidence analysis on these intentions to determine the possibility or credibility of each business intention. Confidence analysis is based on the analysis results of the user feedback content by the intent recognition model to evaluate whether each business intention accurately reflects the user's real needs or preferences.

[0069] Confidence analysis can be performed using a machine learning model that calculates the confidence of each business intent based on multiple factors such as historical data, context, and user behavior patterns. The system matches each business intent with pre-set features or criteria and evaluates the credibility of the intent through confidence scoring. The model can be trained based on supervised learning methods, using large-scale annotated data sets to improve the accuracy of the analysis. The confidence score for each business intent is usually between 0 and 1. The higher the confidence, the more likely it is that the intent is actually a user need.

[0070] After the confidence analysis, the system scores each business intent and marks the business intent with a confidence level higher than the preset threshold as a valid target intent. The threshold is set based on business needs and actual application scenarios to ensure that only intents with higher credibility are considered to be the user's real needs.

[0071] The system sets a preset confidence threshold. Usually in business scenarios, the threshold can be adjusted according to the needs of specific business. For example, in an insurance scenario, it may be set to 0.8, which means that business intents with a confidence level of 80% or more will be marked as valid target intents. Intents below this threshold may be ignored or enter the subsequent further analysis process. Through this mechanism, the system can effectively filter out business intents with high uncertainty and improve the accuracy and efficiency of the system.

[0072] Example: In an insurance business scenario, a user reports the need for a claim and also mentions the need to renew the policy. The system extracts two business intents through the intent recognition model: claim and renewal. The system then performs confidence analysis on these two intents to assess the user's real needs. Assume that the system's confidence score for the claim intent is 0.9 and the score for the renewal intent is 0.7. In this scenario, the confidence threshold is set to 0.8, so the claim is marked as a valid target intent, while renewal is not a priority demand for the time being.

[0073] Through confidence analysis of multiple business intentions, the system can accurately identify the user's real needs and avoid communication errors or service delays caused by misjudgment of business intentions. The confidence analysis mechanism improves the accuracy of business processing and ensures that the system can prioritize the most important business needs, thereby optimizing the customer service experience and reducing unnecessary communication and duplication of services.

[0074] S60: Based on the effective target intention, adjust the initial dialogue strategy to generate a subsequent dialogue strategy.

[0075] In this embodiment, after identifying the user's effective target intent, the system adjusts the first generated dialogue strategy according to the intent. The purpose of the adjustment is to ensure that the subsequent dialogue content is more in line with the user's current needs and optimize the user's communication experience. The first dialogue strategy may be generated based on fixed preliminary information, but after user feedback and target intent are clear, the system will make the dialogue content more accurate through dynamic adjustments.

[0076] The system first identifies the user's core needs based on the effective target intent, such as claims, renewals, or preservation. Then, the system focuses on the business that the user cares about most by modifying the conversation content and structure. The adjustment process may include: deleting unnecessary information, adding instructions related to the target intent, or changing the order of the conversation. The implementation method can automatically generate or adjust the conversation strategy through a rule engine or a machine learning-based model. The system can adjust specific inquiry content, recommendations, etc. based on conversation templates, user portraits, and historical feedback data.

[0077] Based on the adjusted first conversation strategy, the system generates a conversation strategy suitable for subsequent interactions with the user. The focus of the subsequent conversation strategy is to further deepen the communication with the user, respond to customer feedback, or provide further services or suggestions based on customer needs. The subsequent strategy ensures that the conversation content is continuously optimized based on user feedback to avoid redundant and irrelevant information.

[0078] After the first dialogue strategy adjustment, the system will generate subsequent dialogue strategies based on the user's effective target intention and feedback. For example, after the user has confirmed the claim application, the system will automatically generate a dialogue strategy about the claim progress or other related services. The system can call different business modules (such as claims process module, customer service module, etc.), and combine business rules and user portrait data to generate more targeted dialogue content. This process can be based on a rule base, template generation, or dynamic generation.

[0079] Example description: In the customer service scenario of the insurance business, the main demand reported by users is policy renewal. After identifying the user's valid target intent, the system adjusts the first dialogue strategy to remove content that is not related to claims. Next, the system generates a subsequent dialogue strategy to provide users with detailed step-by-step instructions for policy renewal and the option of online renewal payment. Through this strategy adjustment and generation process, the system ensures that customer needs are accurately responded to, while improving service efficiency and customer experience.

[0080] By adjusting the initial conversation strategy and generating subsequent conversation strategies, the system can significantly improve the efficiency and accuracy of outbound calls or customer service communications. The adjusted conversation strategy ensures that the communication content is more focused on the user's real needs and avoids interference from irrelevant content. The generated subsequent conversation strategy can not only further optimize the service process, but also improve customer satisfaction through multiple interactions and enhance the personalization and flexibility of services.

[0081] The present invention relates to the fields of artificial intelligence technology and financial technology, and discloses a method for generating a dialogue strategy based on intent recognition, which generates a user portrait by acquiring basic user information. By collecting the user's personal information, economic information and behavior patterns, the user's business information and preliminary user portrait are analyzed after processing, and a preliminary dialogue strategy is generated in combination with the pending business field. The user feedback information is processed by the intent recognition model, and the customer's multiple business intentions in the target business field are analyzed, and the business intentions are analyzed for confidence. The business intentions with a confidence higher than a preset threshold are identified as valid target intentions. According to the valid target intentions, the preliminary dialogue strategy is adjusted, and the subsequent dialogue strategy is generated to gradually optimize the user experience. The present invention overcomes the limitation that the existing outbound call system cannot accurately respond to customer needs by identifying and analyzing user intentions. Compared with the traditional outbound call method that relies on fixed words, the dialogue strategy can be adjusted based on the customer's personalized needs, thereby improving the effectiveness and reach of the outbound call. Specifically, by deeply exploring the actual needs and preferences of customers, customer dissatisfaction caused by repeated outbound calls or irrelevant content can be reduced. In addition, through the identification and confidence analysis of effective target intentions, the customer's priority needs can be accurately judged, thereby improving the response speed and accuracy of the outbound call service. It significantly reduced the frequency of ineffective communication, improved the overall efficiency of the outbound call system, and enhanced customer satisfaction with the service, effectively supporting the maintenance of customer relationships and the sustainable development of the business.

[0082] In one embodiment, the above S50 includes:

[0083] S501, performing a preliminary analysis on each business intent, if the business intent contains positive deterministic words, assigning a preset high initial confidence; if the business intent contains uncertain words, assigning a preset low initial confidence;

[0084] S502, analyzing the user's historical behavior data, if the user's current business intention is consistent with the historical behavior, then increasing the confidence of the business intention; if the user's current business intention is inconsistent with the historical behavior, then decreasing the confidence of the business intention;

[0085] S503, if the business intention of the user in multiple rounds of conversations is consistent, the confidence of the business intention is increased; if the business intention of the user in multiple rounds of conversations is inconsistent or conflicting, the confidence of the business intention is reduced;

[0086] S504, if the user does not perform an operation that meets the business intention within a preset time after the feedback, the confidence of the business intention is reduced; if the user performs an operation that meets the business intention within a preset time after the feedback, the confidence of the business intention is increased;

[0087] S505, determining the final confidence level of each business intent through deterministic vocabulary matching, historical behavior data analysis, multi-round dialogue feedback, and compliance of operation behaviors;

[0088] S506, comparing the final confidence of each business intent with a preset threshold, and if the final confidence is higher than the threshold, marking the business intent as a valid target intent.

[0089] In this embodiment, when analyzing multiple business intentions of a user, the system first performs a preliminary analysis of each intention to determine its credibility. The initial score of credibility depends on the key words in the intention. If the feedback content contains clear, positive deterministic words (such as "completed", "confirmed"), a higher initial confidence is assigned to the intention. On the contrary, if the feedback content uses uncertain words (such as "maybe", "uncertain"), a lower initial confidence is assigned.

[0090] Through natural language processing (NLP) technology, the system first converts the user's feedback text into structured data and analyzes the key words in it. The keyword library is pre-defined according to business needs and contains positive deterministic words and uncertain words. The system uses a rule engine or a classifier based on the NLP model to assign an initial confidence level to each business intent based on these keywords.

[0091] The system further adjusts the confidence of each intent by comparing the user's current business intent with historical behavior data. If the user's current intent is consistent with his or her historical behavior, the confidence of the intent is increased; if not, the confidence is reduced. Historical behavior data can include the user's past operation records, transaction behaviors, preferences, etc.

[0092] The system queries the user's historical operation records and transaction behavior data by accessing the historical data management platform or user behavior database. The system analyzes the correlation between the user's current intention and his past behavior through rule matching or machine learning models. Based on similarity calculation or behavior consistency analysis, the system increases or decreases the confidence. For example, if the user has frequently renewed the policy before, the confidence will be increased when he currently proposes the intention to renew.

[0093] The system continuously monitors the user's business intent in multiple rounds of conversations. If the user maintains a consistent business intent in multiple rounds of conversations, the system will increase the confidence of the intent. On the contrary, if the user's intent is inconsistent or conflicting in multiple conversations, the system will reduce the confidence of the intent accordingly.

[0094] By analyzing multiple rounds of conversation records, the system tracks the user's feedback data in different rounds of conversation. The system uses the historical conversation data recorded by the conversation management platform to analyze whether the user's intentions in each round of conversation are consistent. Through the understanding of user semantics by NLP technology, the system can determine the intentions in multiple rounds of conversation. Figure 1 If the user's intention is inconsistent (e.g., initially mentioning renewal and later mentioning cancellation of renewal), the system will reduce the confidence of the intention.

[0095] If the user performs a business action within the preset time after the feedback Figure 1 If the user performs the relevant operation within the specified time (such as paying premiums, uploading materials, etc.), the system will increase the confidence of the business intention. If the user does not perform the relevant operation within the specified time, the system will reduce the confidence.

[0096] The system monitors the user's actual operation behavior (such as clicking a link, uploading a file, completing a payment, etc.) to determine whether the user has performed an operation that matches the intention. The system compares these operation behaviors with the business intention and decides whether to adjust the confidence level through behavior compliance analysis. For example, if the user completes the payment after receiving the policy renewal reminder, the system will further increase the confidence level of the renewal intention.

[0097] The system calculates the confidence of each business intent through deterministic vocabulary matching, historical behavior data, multi-round dialogue feedback, and operational behavior compliance. The system then compares the confidence of each intent with the preset threshold. If the confidence is higher than the set threshold, the system marks the business intent as a valid target intent and enters the subsequent processing flow.

[0098] The system will compare the confidence of each intent calculated by the above technical steps with the preset threshold. The preset threshold can be adjusted according to the needs of the business scenario, usually 0.7 or 0.8. Intents with confidence higher than the threshold are considered valid target intents, and the system will further generate new business policies based on the intent.

[0099] Example: In a text message interaction, the user expressed the desire to process a claims settlement, but used words such as "may need" and "uncertain", so the system assigned a low initial confidence to the intent. The system checked the user's history and found that the user had mentioned claims settlement many times before, but failed to submit materials, so the confidence of the intent was lowered. In multiple rounds of conversations, the user's intent was repeatedly uncertain, and no action was taken within the preset time, so the system eventually marked the intent as invalid.

[0100] In this embodiment, by analyzing the confidence of multiple business intentions, the system can accurately identify the real needs of users and ensure that high-confidence business intentions can be processed first. This effectively improves communication efficiency and avoids invalid operations caused by misjudgment of user needs. Finally, by giving priority to business intentions with confidence levels higher than the preset threshold, the system improves service response speed and customer satisfaction.

[0101] In one embodiment, in the above S20, the business information and the preliminary user portrait are analyzed, and a first conversation strategy for the pending business is generated based on the analysis result, including:

[0102] S201, the business information includes claim records, renewal performance and complaint records, and the claim frequency, claim amount and claim processing results are analyzed based on the claim records to generate a user claim score;

[0103] S202, analyzing the timeliness and frequency of renewal based on the renewal performance, and generating a user renewal score;

[0104] S203, analyzing the number of complaints and complaint handling results based on the complaint records to generate a user complaint score;

[0105] S204, performing weighted scoring based on the user's claim score, user renewal score, and user complaint score to generate a user's business scoring result;

[0106] S205: Generate a first conversation strategy based on the business scoring result and the preliminary user portrait in combination with a preset speech strategy library.

[0107] In this embodiment, the system generates a user's claim score by analyzing the user's claim record, including claim frequency, claim amount, and claim processing results. Claim record is an important indicator in the insurance business, reflecting the user's insurance usage behavior and the degree of dependence on claim services.

[0108] The system accesses the claims data in the insurance business management system. First, it counts the number of claims applications submitted by users. Users with higher frequency may represent higher risks. Next, the system analyzes the amount of each claim and the result of the processing (such as whether it was successfully approved or rejected). Through a weighted method, a claims score is generated for the user based on the frequency, amount, and processing result. For example, if a user frequently makes high-value claims and the processing results are smooth, he or she will receive a higher claims score, indicating that he or she is a high-frequency claims user.

[0109] The system generates a renewal score for users based on their policy renewal performance, mainly the timeliness and frequency of renewal. Renewal performance reflects the user's loyalty to insurance products and their sensitivity to renewal timeliness.

[0110] The system accesses the user's renewal data and analyzes the renewal time of the policy. If the user renews on time, the system assigns a higher renewal score to the user; if the user often pays late or renews infrequently, the system assigns a lower renewal score to the user. The renewal frequency mainly depends on the policy type and payment cycle selected by the user, such as annual renewal or quarterly renewal. Users who renew on time show higher loyalty and their scores will increase accordingly.

[0111] The system generates a complaint score for users by analyzing their complaint records, including the number of complaints and the results of complaint handling. Users' complaint behavior can reflect their dissatisfaction with the service and the severity of potential problems.

[0112] The system is connected to the complaint management system. First, the number of user complaints is counted. Frequent complaints may mean that the user is less satisfied with the service. Secondly, the system analyzes the handling results of each complaint, including whether it is resolved in a timely manner and the satisfaction of the resolution. For users who have made multiple complaints and are not satisfied with the handling, the system will give a lower complaint score, prompting the system that the user may need special attention and processing.

[0113] The system weights the user's claim score, renewal score and complaint score to generate the user's comprehensive business score. The weighting method is set according to the different business focuses of the insurance company, and different weights can be used to reflect the importance of each score in the comprehensive score.

[0114] The system combines the scores of claims, renewals, and complaints according to the weights through a preset weighted model. For example, the weight of the claims score can be set to 50%, the renewal score to 30%, and the complaint score to 20%. After weighting, the system generates the user's business score results. Users with higher scores may be high-value customers or have higher service needs. The system determines the subsequent dialogue strategy based on the score.

[0115] After generating the user's business rating results, the system combines the user's preliminary profile and generates the first dialogue strategy through the preset dialogue strategy library. The first dialogue strategy will be personalized based on the user's comprehensive business rating and profile.

[0116] The system combines the user's preliminary user profile, such as the user's age, gender, occupation and other characteristics, with the business scoring results to determine the focus of communication. The system calls the corresponding dialogue template from the preset speech strategy library. For example, for users with high scores and frequent claims, the system may focus on recommending new high-insurance products; while for users with frequent complaints, the first dialogue strategy may be more inclined to appeasement and problem solving. The system generates personalized dialogue content by matching user profiles and business scores.

[0117] Example description: In the insurance business scenario, the user's claim record shows a low claim success rate, and the complaint record shows that the user's multiple complaints have not been satisfactorily resolved. The system assigns a low complaint score and claim score to the user. Combined with the user's profile, the first conversation strategy generated by the system focuses on resolving user complaints and adopts soothing words to prioritize solving the user's current problems.

[0118] In this embodiment, by analyzing the claims records, renewal performance and complaint records, the system can generate a more comprehensive and accurate user business score. It can better reflect the user's behavioral characteristics and business needs, and generate a personalized first conversation strategy based on these scores and user portraits. The pertinence of the conversation strategy is improved, which can significantly improve customer satisfaction and business processing efficiency.

[0119] In one embodiment, the above S60 includes:

[0120] S601, analyzing the effective target intention, identifying the type of user demand, and classifying the type of user demand into completed demand or new business demand;

[0121] S602, if the type of the user demand is a completed demand, identifying the content in the first dialogue strategy that does not match the effective target intent, and removing the unmatched content from the first dialogue strategy;

[0122] S603, if the type of the user demand is a new service demand, select prompt information matching the user demand from the service prompt library, and add the prompt information to the first conversation strategy;

[0123] S604: Generate the subsequent dialogue strategy based on the adjusted first dialogue strategy.

[0124] In this embodiment, the system analyzes the user's demand type based on the identified valid target intention. The user's demand type can generally be divided into two categories: completed demand and new business demand. Completed demand refers to the transaction that the user has completed and no longer needs to be processed further; new business demand refers to the new transaction or new request that the user currently wants to process.

[0125] The system analyzes the intent of user feedback and combines historical data, business processes and other information to determine whether the demand has been completed. For example, if the user has completed the claim application and received feedback, the demand is marked as completed; if the user makes a new business request (such as a new policy purchase or modification), the system identifies it as a new business demand. The system classifies through rule matching or machine learning models to ensure the accuracy of demand classification.

[0126] If the user's demand type is a completed demand, the system will adjust the first dialogue strategy to identify and remove content that does not match the valid target intent. This step can prevent users from receiving redundant or irrelevant content again on completed demands.

[0127] The system analyzes the content of the first conversation strategy and identifies parts that are not relevant to the valid target intent. For example, if the first conversation strategy includes a follow-up on a claim application, but the user has already completed the application, the system will remove the claim-related prompt information. The removal process can analyze and filter out unmatched conversation content through preset business logic or natural language processing technology. The system then updates the first conversation strategy to ensure that the user receives only content related to his or her current needs.

[0128] If the user's demand type is a new business demand, the system will select prompt information that matches the user's demand from the business prompt library to ensure that the conversation content is highly relevant to the user's actual needs. This feature ensures that the system can flexibly respond to changes in user needs.

[0129] The system searches the business prompt library to retrieve prompt information that matches the user's new business needs. The business prompt library contains pre-defined standard prompts, such as instructions for new business processes, step prompts, and precautions. By matching the user's need type and business scenario, the system can select the most appropriate prompt information and dynamically add it to the first conversation strategy. The system can use keyword matching or content-based recommendations to find the prompt that best meets the user's current needs.

[0130] Based on the adjusted first conversation strategy, the system generates a subsequent conversation strategy. The subsequent conversation strategy means that the system further optimizes the conversation content according to the user's current needs and feedback to ensure that subsequent interactions can continue to respond to the user's personalized needs.

[0131] Based on the adjusted first conversation strategy, the system combines the user's feedback data and business information to generate a subsequent conversation strategy suitable for the next interaction. The generation process of the subsequent conversation strategy may include personalized recommendations, process tracking, and prompts for other related services. The system can dynamically adjust the subsequent conversation strategy based on the specific needs of the user through a rule engine or machine learning model. For example, after completing the first claim communication, the system will generate a subsequent conversation strategy to help users track the progress of the claim or provide related follow-up services.

[0132] Example description: In the insurance scenario, the user's effective target intent is that the claim application has been completed, and the system recognizes that this requirement is a completed requirement. The system removes the claim reminder content in the first dialogue strategy to avoid sending irrelevant prompt information to the user. Then, based on the user's policy renewal request, the system selects renewal-related prompt information from the business prompt library and dynamically adds it to the dialogue strategy. Based on these adjustments, the system generates a subsequent dialogue strategy for the user's next renewal operation.

[0133] By adjusting the first conversation strategy based on effective target intent, this embodiment can significantly reduce repetitive and irrelevant information push and ensure the accuracy of the conversation content. The system can adjust the communication content according to the actual needs of the user to avoid unnecessary information interference. At the same time, the system can select the most appropriate prompt information from the business prompt library in the new business demand scenario, realize the personalized customization of the conversation content, and improve the user experience and service efficiency. By generating subsequent conversation strategies, the system can ensure that the continuous interaction with the user meets their latest needs, and improve the flexibility and response speed of the service.

[0134] In one embodiment, after the above S50, the method further includes:

[0135] S507, optimizing the preliminary user portrait based on the effective target intent, wherein the optimization includes adjusting the user's behavior attributes and business preferences;

[0136] S508, when a new business domain is identified, obtaining business information of the user in the new business domain;

[0137] S509, combining the optimized user portrait with the business information in the new business field to generate a first conversation strategy in the new business field.

[0138] In this embodiment, after determining the valid target intent, the system will optimize the user portrait according to these intents. This optimization mainly involves the user's behavioral attributes and business preferences. Behavioral attributes include the user's behavior patterns, operating habits, etc., and business preferences reflect the user's needs and preference changes in specific business areas. Through this optimization, the system can more accurately reflect the user's current characteristics.

[0139] The system records the user's business interactions and feedback to analyze changes in their behavior attributes (such as frequent claims or multiple policy modification applications). In addition, the system also infers the user's business preferences based on the type of product or service selected by the user. The system integrates these changing data into the user's preliminary profile and optimizes it. The updated user profile will reflect the user's behavior changes more dynamically and in real time.

[0140] When the system recognizes that the user has new business needs (such as switching from claims to insurance purchase), the system will obtain user information related to the business field. The new business field may be a business that the user has never been involved in, and the system needs to re-plan the dialogue strategy based on the new business field data.

[0141] The system retrieves the user's information in the new business field through the business management system or related business database. For example, if the user enters the policy renewal field from the claims field, the system will obtain the user's relevant information in the policy renewal, such as the policy expiration date, past renewal behavior, etc. The system connects to different business systems through the API interface and retrieves the user's data in the new business field in real time to ensure timely adjustment of the dialogue strategy.

[0142] After acquiring new business domain information, the system combines it with the optimized user profile to generate a first conversation strategy in the new business domain. This strategy will not only be based on the user's business history, but will also take into account the user's current behavioral attributes and business preferences to ensure the accuracy and relevance of the first conversation content.

[0143] The system dynamically generates the first conversation strategy by connecting to the preset conversation strategy library, combining the optimization data in the user portrait and the information in the new business field. For example, when a user enters the insurance purchase business field, the system will generate recommended insurance products based on their past claims records, renewal records, etc., and generate customized conversation strategies based on the user's current needs. The system uses algorithms to conduct in-depth analysis of user portraits and new business data to ensure that the conversation content matches user needs.

[0144] Example: In the insurance business, the user's effective goal intention is to complete the claim, and the system recognizes that the user has a new business need, namely, policy modification. The system optimizes the user's user profile based on the user's behavioral attributes and preferences, marking him as a customer who "prefers fast claims and efficient processing procedures." Subsequently, the system obtains the user's policy information, and combined with the updated user profile, generates a first dialogue strategy for policy modification, provides the user with simplified modification steps, and recommends that he choose appropriate new policy terms.

[0145] By optimizing user portraits based on effective target intent, this embodiment enables the system to dynamically and accurately reflect changes in user behavior and preferences, ensuring the real-time and accuracy of the portraits. Combined with new business domain information, the first conversation strategy generated by the system can better match user needs, improving the accuracy and personalization of the conversation. The entire process reduces the interference of redundant information, optimizes the user experience, and improves the response speed of the service and user satisfaction.

[0146] In one embodiment, the above S10 includes:

[0147] S101, collect the user's personal information, economic information and behavior patterns to form basic user data;

[0148] S102, cleaning and standardizing the basic user information;

[0149] S103, generating a preliminary user portrait based on the processed information, wherein the preliminary user portrait includes the user's personal characteristics, economic status, and behavioral habits;

[0150] S104: storing the preliminary user portrait in a user portrait database.

[0151] In this embodiment, the system comprehensively collects basic user data, including personal information (such as age, gender, occupation), economic information (such as income level, consumption habits) and behavior patterns (such as purchasing behavior, service usage frequency, etc.) of the user. This information forms the basis of the user portrait.

[0152] The system obtains basic information about users through a variety of data sources. Personal information can be obtained from the registration form filled out by the user, and economic information can be obtained by analyzing the user's consumption records, transaction history, etc. Behavior patterns can be obtained through the user's historical operation records, website access paths, or the use of insurance products. The system connects with multiple business systems and external databases through API interfaces to form unified basic user data.

[0153] Since basic user information may come from multiple systems or channels, the data formats are often not uniform. Therefore, the system needs to clean and standardize the collected data to ensure the integrity and consistency of the data. The purpose of cleaning is to remove redundant, incomplete or erroneous data, and the purpose of standardization is to unify the data format to make subsequent analysis and processing more accurate.

[0154] The system processes user information through data cleaning tools, including identifying and removing duplicates and correcting inconsistent data formats (such as date format, currency unit, etc.). At the same time, the system completes incomplete fields (such as querying third-party databases to complete missing personal information). After standardization, all data fields are converted to a unified format for subsequent analysis and storage.

[0155] The cleaned and standardized user basic data is used to generate preliminary user portraits. User portraits are multi-dimensional descriptions of user characteristics, covering personal characteristics, economic status, and behavioral habits. Preliminary user portraits can reflect the basic characteristics of users and provide a basis for subsequent business analysis.

[0156] The system uses data analysis tools or machine learning algorithms to classify and cluster basic user information according to preset models to generate a profile of each user. The user's personal characteristics include age, gender, occupation, etc., economic status is evaluated based on income, expenditure and financial behavior, and behavioral habits are generated based on historical operation records, usage habits and other data. The system can use feature weighting to generate the final preliminary user profile and use this information for subsequent business processing.

[0157] The generated preliminary user portrait needs to be stored in a dedicated user portrait database so that the system can call it in subsequent business processes. This database is a centralized management system for storing all user portrait information, supporting efficient access operations and multi-dimensional data queries.

[0158] The system stores the generated preliminary user portraits in a distributed database or a dedicated user portrait management system. Each user record in the database is associated with its unique user ID for quick query. The stored procedure supports data compression and encryption to ensure the security and privacy of user data. The system improves query efficiency through an index mechanism to ensure that the business system can quickly retrieve relevant user portrait data.

[0159] By collecting the user's personal information, economic information and behavior patterns, the system can build a preliminary user portrait, laying the foundation for subsequent personalized services and business recommendations. Data cleaning and standardization ensure the accuracy and consistency of the data and optimize the data quality. The generated preliminary user portrait can accurately reflect the basic characteristics of the user. After being stored in the user portrait database, the system can quickly retrieve and analyze and apply it in combination with subsequent business needs, improving the system's response speed and personalized service level.

[0160] In one embodiment, in the above S20, after the first dialogue strategy of the pending business is generated based on the analysis result, the following step is further included:

[0161] S206, comparing the first dialogue strategies of multiple pending services through text similarity analysis, and identifying similar content with a similarity higher than a preset threshold;

[0162] S207, merging similar contents in multiple first dialogue strategies into a common dialogue segment;

[0163] S208: Generate a comprehensive dialogue strategy covering multiple pending services based on the common dialogue segment.

[0164] In this embodiment, the text similarity between different strategies is evaluated by analyzing the first dialogue strategies of multiple pending services. Text similarity analysis is a natural language processing (NLP) technology used to determine the similarity between two texts, with the purpose of identifying dialogue contents with high similarity to avoid repeatedly generating multiple similar dialogue strategies.

[0165] The system uses text similarity algorithms (such as cosine similarity, Jaccard similarity, BERT model, etc.) to compare the first conversation strategies of different businesses. The system first converts the text into a vectorized representation, and then calculates the similarity score between the texts. Preset thresholds are used to judge similar content. Usually, texts with a similarity score of a certain percentage (such as 0.8 or 80%) are considered similar. The similarity algorithm can be adjusted according to the business scenario to ensure that similar texts can be accurately identified.

[0166] After identifying the dialogue policy contents with high similarity, the system merges these similar parts to generate common dialogue segments. The purpose of this merging process is to reduce repeated dialogue content and optimize the efficiency of dialogue policy generation.

[0167] The system merges the highly similar content identified by similarity analysis. The merging process involves simplifying the repeated content and removing the redundant parts. The system uses a preset rule library to determine how to merge these fragments to ensure that the merged content is concise and complete. For example, "Please upload materials as soon as possible" that appears repeatedly in multiple business scenarios can be merged into a common dialogue fragment to cover all dialogue strategies involving uploading materials. After merging, the system stores the common dialogue fragment in the strategy library for subsequent use when generating comprehensive dialogue strategies.

[0168] After merging similar conversation segments, the system generates a comprehensive conversation strategy covering multiple pending services based on these common conversation segments. The purpose of the comprehensive conversation strategy is to provide users with unified communication content, avoid repeated outbound calls or notifications for multiple pending services, and improve the user experience.

[0169] The system generates a comprehensive dialogue strategy that includes multiple business scenarios by calling the merged common dialogue segments. The system will dynamically combine different common dialogue segments based on the user's actual to-do items to ensure that the comprehensive dialogue strategy can cover multiple user needs. For example, the system can merge services such as claim reminders and renewal reminders into one-time notifications to reduce the number of outbound calls. The generated comprehensive dialogue strategy will be personalized according to the business needs of different users and pushed to users through appropriate channels.

[0170] Example: In the insurance business, users may have both claim applications and policy renewal reminders. By analyzing the first dialogue strategies for claims and renewals, the system found that the two are similar in the part where they prompt users to upload materials. Through text similarity analysis, the system identified the content "Please upload materials as soon as possible" as a highly similar segment and merged it into a common dialogue segment. Based on this common segment, the system generated a comprehensive dialogue strategy that includes claims and renewals, avoiding users from receiving repeated reminders to upload materials in the two businesses.

[0171] This embodiment analyzes the text similarity of the first dialogue strategies of multiple services, so that the system can effectively identify similar content and merge them, avoiding repeated dialogue generation processes. The efficiency of dialogue strategy generation is improved, the frequency of users receiving repeated content is reduced, and the user experience is optimized. The generation of comprehensive dialogue strategies enables the system to cover multiple pending services, reduce the number of outbound calls and notification redundancy, and ensure the accuracy and relevance of communication content.

[0172] In one embodiment, a device for generating a dialogue strategy based on intention recognition is provided, and the device for generating a dialogue strategy based on intention recognition corresponds one-to-one to the method for generating a dialogue strategy based on intention recognition in the above embodiment. Figure 3 , Figure 3 This is a functional module diagram of a preferred embodiment of a conversation strategy generation device based on intention recognition of the present invention. User portrait generation module 10, first conversation strategy generation module 20, feedback information collection module 30, user intention recognition module 40, target intention determination module 50 and subsequent conversation strategy generation module 60. The functional modules are described in detail as follows:

[0173] A user portrait generation module 10, used to obtain basic information of the user and generate a preliminary user portrait based on the basic information;

[0174] The first conversation strategy generation module 20 is used to obtain the business information of users in the pending business field, analyze the business information and the preliminary user portrait, and generate the first conversation strategy of the pending business based on the analysis result;

[0175] A feedback information collection module 30, used to execute the first dialogue strategy and obtain user feedback information;

[0176] A user intention recognition module 40, configured to analyze the user feedback information through an intention recognition model to recognize multiple business intentions of the user in the target business field;

[0177] A target intention determination module 50 is used to perform confidence analysis on the multiple business intentions, and to take the business intentions with confidence higher than a preset threshold as valid target intentions;

[0178] The subsequent dialogue strategy generation module 60 is used to adjust the initial dialogue strategy based on the effective target intention to generate a subsequent dialogue strategy.

[0179] In one embodiment, the target intention determination module 50 is specifically configured to:

[0180] Perform a preliminary analysis on each business intent. If the business intent contains positive deterministic words, a preset high initial confidence is assigned; if the business intent contains uncertain words, a preset low initial confidence is assigned;

[0181] Analyze the user's historical behavior data. If the user's current business intention is consistent with the historical behavior, the confidence of the business intention is increased; if the user's current business intention is inconsistent with the historical behavior, the confidence of the business intention is reduced;

[0182] If the user's business intentions in multiple rounds of conversations are consistent, the confidence of the business intentions is increased; if the user's business intentions in multiple rounds of conversations are inconsistent or conflicting, the confidence of the business intentions is reduced;

[0183] If the user does not perform an operation that is consistent with the business intent within the preset time after the feedback, the confidence in the business intent is reduced; if the user performs an operation that is consistent with the business intent within the preset time after the feedback, the confidence in the business intent is increased;

[0184] Determine the final confidence level of each business intent through deterministic vocabulary matching, historical behavior data analysis, multi-round dialogue feedback, and operational behavior compliance;

[0185] The final confidence of each business intent is compared with the preset threshold. If the final confidence is higher than the threshold, the business intent is marked as a valid target intent.

[0186] In one embodiment, the first dialogue strategy generation module 20 is specifically used to:

[0187] The business information includes claim records, renewal performance and complaint records. Based on the claim records, the claim frequency, claim amount and claim processing results are analyzed to generate a user claim score;

[0188] Based on the renewal performance, analyze the timeliness and frequency of renewals to generate a user renewal score;

[0189] Based on the complaint records, the number of complaints and complaint handling results are analyzed to generate a user complaint score;

[0190] Perform weighted scoring based on the user's claim score, user renewal score, and user complaint score to generate a user's business score result;

[0191] Based on the business scoring result and the preliminary user portrait, a first conversation strategy is generated in combination with a preset speech strategy library.

[0192] In one embodiment, the subsequent dialogue strategy generation module 60 is specifically used to:

[0193] Analyze the effective target intention, identify the type of user needs, and classify the type of user needs into completed needs or new business needs;

[0194] If the type of the user demand is a completed demand, identifying content in the first dialogue strategy that does not match the effective target intent, and removing the unmatched content from the first dialogue strategy;

[0195] If the type of the user demand is a new business demand, select prompt information matching the user demand from the business prompt library, and add the prompt information to the first conversation strategy;

[0196] The subsequent dialogue strategy is generated based on the adjusted first dialogue strategy.

[0197] In one embodiment, the target intention determination module 50 is specifically configured to:

[0198] Optimizing the preliminary user profile based on the effective target intent, wherein the optimization includes adjusting the user's behavioral attributes and business preferences;

[0199] When a new business domain is identified, business information of the user in the new business domain is obtained;

[0200] Combine the optimized user portrait with the business information in the new business field to generate the first conversation strategy in the new business field.

[0201] In one embodiment, the user portrait generation module 10 is specifically used to:

[0202] Collect users’ personal information, economic information and behavior patterns to form basic user data;

[0203] Cleaning and standardizing the basic user information;

[0204] Generate a preliminary user profile based on the processed information, wherein the preliminary user profile includes the user's personal characteristics, economic status, and behavioral habits;

[0205] The preliminary user portrait is stored in a user portrait database.

[0206] In one embodiment, the first dialogue strategy generation module 20 is specifically used to:

[0207] By analyzing text similarity, the first conversation strategies of multiple pending tasks are compared to identify similar content with a similarity higher than a preset threshold.

[0208] Merge similar content from multiple first-conversation strategies into common conversation snippets;

[0209] Based on the common dialogue fragment, a comprehensive dialogue strategy covering multiple pending businesses is generated.

[0210] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 4As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external user terminal through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the server side of a dialogue strategy generation method based on intent recognition.

[0211] In one embodiment, a computer device is provided. The computer device may be a user terminal, and its internal structure diagram may be as follows: Figure 5 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a user-side method for generating a dialogue strategy based on intention recognition.

[0212] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0213] Obtaining basic information of the user, and generating a preliminary user profile based on the basic information;

[0214] Obtaining business information of users in the pending business area, analyzing the business information and preliminary user portraits, and generating a first conversation strategy for the pending business based on the analysis results;

[0215] Execute the first dialogue strategy and obtain user feedback information;

[0216] Analyzing the user feedback information through an intention recognition model to identify multiple business intentions of the user in the target business field;

[0217] Performing confidence analysis on the multiple business intentions, and taking the business intentions with confidence higher than a preset threshold as valid target intentions;

[0218] Based on the effective target intention, the initial dialogue strategy is adjusted to generate a subsequent dialogue strategy.

[0219] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0220] Obtaining basic information of the user, and generating a preliminary user profile based on the basic information;

[0221] Obtaining business information of users in the pending business area, analyzing the business information and preliminary user portraits, and generating a first conversation strategy for the pending business based on the analysis results;

[0222] Execute the first dialogue strategy and obtain user feedback information;

[0223] Analyzing the user feedback information through an intention recognition model to identify multiple business intentions of the user in the target business field;

[0224] Performing confidence analysis on the multiple business intentions, and taking the business intentions with confidence higher than a preset threshold as valid target intentions;

[0225] Based on the effective target intention, the initial dialogue strategy is adjusted to generate a subsequent dialogue strategy.

[0226] It should be noted that the above functions or steps that can be implemented by the computer-readable storage medium or computer device can refer to the relevant descriptions on the server side and the user side in the aforementioned method embodiment. To avoid repetition, they will not be described one by one here.

[0227] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0228] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0229] It should be noted that if software tools or components other than those of the Company appear in the embodiments of the present application, they are only used for illustration and do not represent actual use. The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the above-mentioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the above-mentioned embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents; and these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for generating a dialogue strategy based on intention recognition, characterized in that: The following steps are involved: Obtaining basic information of the user, and generating a preliminary user profile based on the basic information; Obtaining business information of users in the pending business area, analyzing the business information and preliminary user portraits, and generating a first conversation strategy for the pending business based on the analysis results; Execute the first dialogue strategy and obtain user feedback information; Analyzing the user feedback information through an intention recognition model to identify multiple business intentions of the user in the target business field; Performing confidence analysis on the multiple business intentions, and taking the business intentions with confidence higher than a preset threshold as valid target intentions; Based on the effective target intention, the initial dialogue strategy is adjusted to generate a subsequent dialogue strategy.

2. The method for generating a dialogue strategy based on intention recognition according to claim 1, characterized in that: Confidence analysis is performed on the multiple business intentions, and business intentions with confidence levels higher than a preset threshold are taken as valid target intentions, including: Perform a preliminary analysis on each business intent. If the business intent contains positive deterministic words, a preset high initial confidence is assigned; if the business intent contains uncertain words, a preset low initial confidence is assigned; Analyze the user's historical behavior data. If the user's current business intention is consistent with the historical behavior, the confidence of the business intention is increased; if the user's current business intention is inconsistent with the historical behavior, the confidence of the business intention is reduced; If the user's business intentions in multiple rounds of conversations are consistent, the confidence of the business intentions is increased; if the user's business intentions in multiple rounds of conversations are inconsistent or conflicting, the confidence of the business intentions is reduced; If the user does not perform an operation that is consistent with the business intent within the preset time after the feedback, the confidence in the business intent is reduced; if the user performs an operation that is consistent with the business intent within the preset time after the feedback, the confidence in the business intent is increased; Determine the final confidence level of each business intent through deterministic vocabulary matching, historical behavior data analysis, multi-round dialogue feedback, and operational behavior compliance; The final confidence of each business intent is compared with the preset threshold. If the final confidence is higher than the threshold, the business intent is marked as a valid target intent.

3. The method for generating a dialogue strategy based on intention recognition according to claim 1, characterized in that: Analyze the business information and preliminary user portrait, and generate a first conversation strategy for the pending business based on the analysis results, including: The business information includes claim records, renewal performance and complaint records. Based on the claim records, the claim frequency, claim amount and claim processing results are analyzed to generate a user claim score; Based on the renewal performance, analyze the timeliness and frequency of renewals to generate a user renewal score; Based on the complaint records, the number of complaints and complaint handling results are analyzed to generate a user complaint score; Perform weighted scoring based on the user's claim score, user renewal score, and user complaint score to generate a user's business score result; Based on the business scoring result and the preliminary user portrait, a first conversation strategy is generated in combination with a preset speech strategy library.

4. The method for generating a dialogue strategy based on intention recognition as claimed in claim 1, characterized in that: Based on the effective target intention, the first dialogue strategy is adjusted to generate a subsequent dialogue strategy, including: Analyze the effective target intention, identify the type of user needs, and classify the type of user needs into completed needs or new business needs; If the type of the user demand is a completed demand, identifying content in the first dialogue strategy that does not match the effective target intent, and removing the unmatched content from the first dialogue strategy; If the type of the user demand is a new business demand, select prompt information matching the user demand from the business prompt library, and add the prompt information to the first conversation strategy; The subsequent dialogue strategy is generated based on the adjusted first dialogue strategy.

5. The method for generating a dialogue strategy based on intention recognition according to claim 1, characterized in that: After performing confidence analysis on the multiple business intentions and taking the business intentions with confidence higher than a preset threshold as valid target intentions, the method further includes: Optimizing the preliminary user profile based on the effective target intent, wherein the optimization includes adjusting the user's behavioral attributes and business preferences; When a new business domain is identified, business information of the user in the new business domain is obtained; Combine the optimized user portrait with the business information in the new business field to generate the first conversation strategy in the new business field.

6. The method for generating a dialogue strategy based on intention recognition according to claim 1, characterized in that: Obtaining basic information of the user and generating a preliminary user profile based on the basic information, including: Collect users’ personal information, economic information and behavior patterns to form basic user data; Cleaning and standardizing the basic user information; Generate a preliminary user profile based on the processed information, wherein the preliminary user profile includes the user's personal characteristics, economic status, and behavioral habits; The preliminary user portrait is stored in a user portrait database.

7. The method for generating a dialogue strategy based on intention recognition according to claim 1, characterized in that: After generating the first conversation strategy for pending business based on the analysis results, it also includes: By analyzing text similarity, the first conversation strategies of multiple pending tasks are compared to identify similar content with a similarity higher than a preset threshold. Merge similar content from multiple first-conversation strategies into common conversation snippets; Based on the common dialogue fragment, a comprehensive dialogue strategy covering multiple pending businesses is generated.

8. A dialogue strategy generation device based on intention recognition, characterized in that: The dialogue strategy generation device based on intention recognition includes: A user portrait generation module, used to obtain basic information of the user and generate a preliminary user portrait based on the basic information; A first conversation strategy generation module is used to obtain business information of users in the pending business field, analyze the business information and preliminary user portraits, and generate a first conversation strategy for the pending business based on the analysis results; A feedback information collection module, used to execute the first dialogue strategy and obtain user feedback information; A user intention recognition module, used to analyze the user feedback information through an intention recognition model to identify multiple business intentions of the user in the target business field; A target intention determination module, used to perform confidence analysis on the multiple business intentions, and take the business intentions with confidence higher than a preset threshold as valid target intentions; A subsequent dialogue strategy generation module is used to adjust the initial dialogue strategy based on the effective target intention to generate a subsequent dialogue strategy.

9. A computer device, characterized in that: The computer device includes a memory, a processor, and a dialogue strategy generation program based on intent recognition stored in the memory and executable on the processor. When the dialogue strategy generation program based on intent recognition is executed by the processor, the steps of the dialogue strategy generation method based on intent recognition as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The storage medium stores a dialogue strategy generation program based on intent recognition, and when the dialogue strategy generation program based on intent recognition is executed by the processor, the steps of the dialogue strategy generation method based on intent recognition as described in any one of claims 1-7 are implemented.

Images