Intelligent customer service training research platform based on high-precision semantic understanding technology
Through the intelligent customer service training platform based on high-precision semantic understanding technology, intelligent customer service is trained to simulate real-person responses, which solves the problem of untimely transfer of real-person customer service, improves the response speed and adaptability of intelligent customer service, and reduces the workload of real-person customer service.
Patent Information
- Application Number
- CN202510969875.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-15
- Publication Date
- 2025-10-28
Smart Images
Figure CN120849558A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent customer service training technology, and in particular to an intelligent customer service training research platform based on high-precision semantic understanding technology. Background Technology
[0002] Intelligent customer service is an industry-oriented technology developed on the basis of large-scale knowledge processing. It is based on (large-scale knowledge processing technology, natural language understanding technology, knowledge management technology, automatic question answering system, reasoning technology, etc.) and has industry universality. It not only provides enterprises with fine-grained knowledge management technology, but also establishes a fast and effective technical means for communication between enterprises and a large number of users based on natural language. At the same time, it can also provide enterprises with the statistical analysis information needed for refined management.
[0003] Semantic understanding technology refers to the process of using computer technology to understand text and answer questions related to it. Semantic understanding focuses more on understanding the context and controlling the accuracy of the answers. Currently, semantic understanding technology plays an important role in fields such as intelligent customer service and automated product question answering.
[0004] When using intelligent customer service, some people may request a real person to reply when the intelligent customer service responds. However, the number of real customer service representatives is limited, and they are busy and cannot transfer calls and provide services in a timely manner, resulting in long waiting times for the other party and reduced communication efficiency. Summary of the Invention
[0005] This invention discloses an intelligent customer service training and research platform based on high-precision semantic understanding technology. It aims to solve the technical problem that some people may request a real person to reply when the intelligent customer service responds, but the number of real customer service representatives is limited, they are busy, and they cannot transfer the call and provide services in a timely manner, resulting in long waiting times and reduced communication efficiency.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] A high-precision semantic understanding technology-based intelligent customer service training research platform includes a data filling unit, a database, a learning unit, a training unit, a training result analysis unit, and a correction unit. The data filling unit is connected to the database, and the training unit is also connected to the database. The database is a repository for storing all data. The data filling unit is used to fill new data into the database. The learning unit is used for the intelligent customer service to learn. The training unit is used to train the intelligent customer service after it has learned. The training result analysis unit is used to perform an overall analysis of the training results. The correction unit corrects the intelligent customer service's responses based on the results analyzed from the training results and the content in the database.
[0008] The learning unit includes a response logic setting module, a response logic import module, a response feature setting module, a human simulation response module, and a database update prompt module, wherein the database update prompt module is connected to the database;
[0009] The reply feature setting module includes a classification tone feature setting module, an attached emoji setting module, and a single reply quantity setting module;
[0010] The real-person simulation reply module includes a tone feature matching module, an attached emoji matching module, a reply logic range analysis module, and a quick transfer module;
[0011] The training unit includes a multi-window hiding training module, a classification training module, a question matching rate analysis module, a question solving rate analysis module, and a typical question selection module;
[0012] The training result analysis unit includes a typical question set module, a wrong question set module, and a wrong question sorting and analysis module. The typical question set module is connected to the typical question selection module, and the wrong question set module is connected to the classification training module.
[0013] By incorporating a training unit, the intelligent customer service system addresses the common issue of requests for human responses from users. While some users may request a real human to reply, the limited number of real human agents and their heavy workload prevent timely transfers and service provision. The training unit's human simulation response module allows the intelligent customer service system to learn and practice human-like responses. Furthermore, the tone feature matching module and the accompanying emoji setting module add interjections and corresponding emojis to the responses, thus mitigating the impersonal and mechanical nature of the intelligent customer service system. This provides sufficient buffer time for transferring users to real human agents, reduces their workload, and effectively improves the responsiveness of the intelligent customer service system in emergency situations.
[0014] In a preferred embodiment, the response logic setting module is used to set the response logic range limit data for the intelligent customer service, including setting logic range threshold data; the response logic import module is used to import response logic and response templates from the database; the response feature setting module is used to set the features in the intelligent customer service response; the human simulation response module is used for the robot to learn human simulation responses on the platform; the database update prompt module is used to prompt the database update dynamics; the classification tone feature setting module is used to set the tone features of the intelligent customer service in a specific environment after classification when in human simulation response state; the attached emoji setting module is used to set the attached emoji data of the intelligent customer service in human simulation response state; and the single response quantity setting module is used to set the maximum number of data responses that the intelligent customer service can send in a single response under normal response conditions.
[0015] The system includes a response feature setting module, where the single response quantity setting module can be used to set the maximum number of data responses the intelligent customer service can provide in a single response under normal circumstances. This means that when responding, the system can automatically match multiple types of data related to the question raised. The single response quantity setting module allows you to directly set the number of responses, enabling the intelligent customer service to better adapt to different contextual needs.
[0016] In a preferred embodiment, the tone feature matching module is used to match corresponding tone features in the simulated human response state of the intelligent customer service; the accompanying emoji matching module is used to match corresponding emojis in the simulated human response state of the intelligent customer service; the response logic range analysis module is used to train the intelligent customer service to recognize uncertain response states in the simulated human response state; the quick transfer module is used to train the intelligent customer service to quickly transfer to human customer service; the multi-window hiding training module is used to open multiple windows for classification and grouping training during the intelligent customer service training process, and hide multiple windows used for training; the classification training module is used to classify and train the intelligent customer service according to different contextual needs and different work areas; and the question matching rate analysis module is used to analyze the question matching of the intelligent customer service based on the training results. The problem-solving rate analysis module is used to analyze the problem-solving rate of the intelligent customer service based on the training results. The typical question selection module filters out atypical questions based on the training results and selects the typical questions that are answered. The classification training module includes a question bank classification search module, a question type import training module, an intelligent reply training module, and a high-precision semantic understanding module. The question bank classification search module is connected to a database and is used to search for question banks with different contextual needs and different work fields in the database. The question type import training module is used to import the searched question banks and import them into the platform for training. The intelligent reply training module is used for intelligent customer service to conduct dialogue training based on the question bank. The high-precision semantic understanding module trains the semantic understanding function of intelligent customer service based on high-precision semantic understanding technology during the dialogue training process.
[0017] By setting up a categorized training module, the intelligent customer service system is trained on different types of questions in different work fields and contexts, thereby improving the overall response performance of the intelligent customer service system. This allows the intelligent customer service system trained through this platform to directly adapt to different work environments and expand its scope of use.
[0018] In a preferred embodiment, the typical question set module is used to collect and statistically analyze typical questions and their responses during the training process; the incorrect question set module is used to collect and statistically analyze incorrect questions during training; and the incorrect question sorting and analysis module is used to sort and analyze incorrect questions according to their categories.
[0019] By setting up a training unit, which opens multiple windows for training during the training process through a multi-window hidden training module, the training speed is improved, allowing the intelligent customer service to conduct a large number of training sessions simultaneously. At the same time, by hiding the windows and quickly identifying and analyzing the training dialogues, problems can be detected directly through observation and analysis results without real-time monitoring of the intelligent customer service, which is conducive to improving the training and optimization capabilities of the intelligent customer service.
[0020] As shown above, an intelligent customer service training research platform based on high-precision semantic understanding technology includes a data filling unit, a database, a learning unit, a training unit, a training result analysis unit, and a correction unit. The data filling unit is connected to the database, and the training unit is also connected to the database. The database is a repository for storing all data. The data filling unit is used to fill new data into the database. The learning unit is used for the intelligent customer service to learn. The training unit is used to train the intelligent customer service after it has learned. The training result analysis unit is used to perform an overall analysis of the training results. The correction unit corrects the intelligent customer service's responses based on the results analyzed from the training results and the content in the database.
[0021] The learning unit includes a response logic setting module, a response logic import module, a response feature setting module, a human simulation response module, and a database update prompt module, wherein the database update prompt module is connected to the database;
[0022] The reply feature setting module includes a classification tone feature setting module, an attached emoji setting module, and a single reply quantity setting module;
[0023] The real-person simulation reply module includes a tone feature matching module, an attached emoji matching module, a reply logic range analysis module, and a quick transfer module;
[0024] The training unit includes a multi-window hiding training module, a classification training module, a question matching rate analysis module, a question solving rate analysis module, and a typical question selection module;
[0025] The training result analysis unit includes a typical question set module, a wrong question set module, and a wrong question organization and analysis module. The typical question set module is connected to the typical question selection module, and the wrong question set module is connected to the classification training module. The intelligent customer service training research platform based on high-precision semantic understanding technology provided by this invention has the technical effect of allowing intelligent customer service to conduct simulated human response training, providing sufficient buffer time for transferring to human customer service, and reducing the workload of human customer service personnel. Attached Figure Description
[0026] Figure 1This is a schematic diagram of the overall structure of an intelligent customer service training and research platform based on high-precision semantic understanding technology proposed in this invention.
[0027] Figure 2 This is a schematic diagram of the learning unit structure of an intelligent customer service training research platform based on high-precision semantic understanding technology proposed in this invention.
[0028] Figure 3 This is a schematic diagram of the response feature setting module of an intelligent customer service training and research platform based on high-precision semantic understanding technology proposed in this invention.
[0029] Figure 4 This is a schematic diagram of the human simulation response module structure of an intelligent customer service training and research platform based on high-precision semantic understanding technology proposed in this invention.
[0030] Figure 5 This is a schematic diagram of the training unit structure of an intelligent customer service training research platform based on high-precision semantic understanding technology proposed in this invention.
[0031] Figure 6 This is a schematic diagram of the classification training module structure of an intelligent customer service training research platform based on high-precision semantic understanding technology proposed in this invention.
[0032] Figure 7 This is a schematic diagram of the training result analysis unit structure of an intelligent customer service training research platform based on high-precision semantic understanding technology proposed in this invention. Detailed Implementation
[0033] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.
[0034] The intelligent customer service training research platform based on high-precision semantic understanding technology disclosed in this invention is mainly applied to intelligent customer service training scenarios.
[0035] Reference Figures 1-7 A high-precision semantic understanding technology-based intelligent customer service training research platform includes a data filling unit, a database, a learning unit, a training unit, a training result analysis unit, and a correction unit. The data filling unit is connected to the database, and the training unit is also connected to the database. The database is a repository for storing all data. The data filling unit is used to fill new data into the database. The learning unit is used for the intelligent customer service to learn. The training unit is used to train the intelligent customer service after it has learned. The training result analysis unit is used to perform an overall analysis of the training results. The correction unit corrects the intelligent customer service's responses based on the results analyzed from the training results and the content in the database.
[0036] The learning unit includes a response logic setting module, a response logic import module, a response feature setting module, a human simulation response module, and a database update prompt module. The database update prompt module is connected to the database.
[0037] The reply feature setting module includes a category tone feature setting module, an attached emoji setting module, and a single reply quantity setting module;
[0038] The real-person simulated reply module includes a tone feature matching module, an attached emoji matching module, a reply logic range analysis module, and a quick transfer module;
[0039] The training unit includes a multi-window hiding training module, a classification training module, a question matching rate analysis module, a question solving rate analysis module, and a typical question selection module;
[0040] The training results analysis unit includes a typical question collection module, a wrong question collection module, and a wrong question organization and analysis module. The typical question collection module is connected to the typical question selection module, and the wrong question collection module is connected to the classification training module. During the use of the intelligent customer service, some people may request a real person to reply when the intelligent customer service responds. However, the number of real customer service representatives is limited, and they are busy and cannot transfer the call in time to provide services. The real person simulation reply module in the learning unit allows the intelligent customer service to learn and train to respond in a real person's manner. Through the tone feature matching module and the attached emoji setting module, tone particles and corresponding emojis are added to the reply process, thereby compensating for the relatively cold and mechanical defects of the intelligent customer service. When the other party requests a real customer service reply, the reply method can be changed to simulate a real person's reply, providing sufficient buffer time for transferring to a real customer service representative and reducing the workload of the real customer service representatives. Through the reply logic range analysis module and the fast transfer module, the intelligent customer service can be effectively trained to identify the uncertainty of the reply raised by the other party and quickly transfer the call to the intelligent customer service representative after identification, effectively improving the response speed of the intelligent customer service in emergency response environments.
[0041] Reference Figure 2 In a preferred embodiment, the response logic setting module is used to set the response logic range limit data of the intelligent customer service, including setting logic range threshold data; the response logic import module is used to import response logic and response templates from the database; the response feature setting module is used to set the features in the intelligent customer service response; the human simulation response module is used for the robot to learn human simulation responses on the platform; and the database update prompt module is used to prompt the database update dynamics.
[0042] Reference Figure 3In a preferred embodiment, the classification tone feature setting module is used to set the tone features of the intelligent customer service in a specific environment after classification when responding to a real person in a simulated response state. The attached emoji setting module is used to set the attached emoji data of the intelligent customer service in a simulated response state. The single reply quantity setting module is used to set the maximum number of data replies that the intelligent customer service can send in a single response under normal response conditions. The single reply quantity setting module can be used to set the maximum number of data replies that the intelligent customer service can send in a single response under normal response conditions. That is, when responding, it can automatically match multiple types of data related to the question. The single reply quantity setting module can directly set the number of replies. For example, when using intelligent customer service in a business processing scenario, setting a higher reply quantity can match multiple relevant data and links during the response process, including business processing procedures, avoiding the other party's secondary questioning. When using intelligent customer service in a merchant's response, setting a lower reply quantity can achieve the other party's question, and a precise single reply can be sent based on the question, avoiding the other party's impatience caused by receiving too many replies at once, so that the intelligent customer service can better adapt to different contextual needs.
[0043] Reference Figure 4 In a preferred embodiment, the tone feature matching module is used to match corresponding tone features in the intelligent customer service's simulated human response state, the accompanying emoji matching module is used to match corresponding emojis in the intelligent customer service's simulated human response state, the response logic range analysis module is used to train the intelligent customer service to recognize uncertain response states in the simulated human response state, and the quick transfer module is used to train the intelligent customer service to quickly transfer to human customer service.
[0044] Reference Figure 5 In a preferred embodiment, the multi-window hiding training module is used to open multiple windows for categorized and grouped training during the training process of intelligent customer service, and hide multiple windows used for training. The classification training module is used to classify and train intelligent customer service according to different contextual needs and different work fields. The question matching rate analysis module is used to analyze the question matching rate of intelligent customer service based on the training results. The question resolution rate analysis module is used to analyze the question resolution rate of intelligent customer service based on the training results. The typical question selection module is used to filter out atypical questions based on the training results and select the typical questions that are answered.
[0045] Reference Figure 6 In a preferred embodiment, the classification training module includes a question bank classification search module, a question type import training module, an intelligent response training module, and a high-precision semantic understanding module, with the question bank classification search module connected to the database.
[0046] Reference Figure 6In a preferred embodiment, the question bank classification search module is used to search the database for question banks with different contextual needs and different work fields. The question type import training module is used to import the searched question banks and import them into the platform for training. The intelligent response training module is used for intelligent customer service to conduct dialogue training based on the question banks. The high-precision semantic understanding module trains the semantic understanding function of intelligent customer service during the dialogue training process based on high-precision semantic understanding technology. By setting up the classification training module, intelligent customer service is trained on different types of questions in different work fields and different contextual environments, thereby improving the overall response performance of intelligent customer service. This allows intelligent customer service trained through this platform to directly adapt to different work environments and expand its scope of use.
[0047] Reference Figure 7 In a preferred embodiment, the typical question set module is used to collect and statistically analyze typical questions and their responses during the training process; the incorrect question set module is used to collect and statistically analyze incorrect questions during training; the incorrect question sorting and analysis module sorts and analyzes incorrect questions according to their categories; during the training process, the training unit opens multiple windows through the multi-window hiding training module to improve training speed, enabling the intelligent customer service to conduct a large number of training sessions simultaneously. At the same time, by hiding the windows, the hidden training dialogues are quickly identified and analyzed directly through the question matching analysis module, the question resolution rate analysis module, and the training analysis unit. This allows for the detection of problems through observation and analysis results without real-time monitoring of the intelligent customer service, which is beneficial for improving the training and optimization capabilities of the intelligent customer service.
[0048] Working Principle: During use, some users may request a human response from the intelligent customer service system. However, the limited number of human agents and their heavy workload prevent timely transfers and service provision. The learning unit's human simulation response module trains the intelligent customer service system to learn and simulate human responses. The tone feature matching module and the accompanying emoji setting module add interjections and corresponding emojis to the responses, compensating for the somewhat cold and mechanical nature of the intelligent customer service system. When a human response is requested, the system can simulate a human response by changing the response method, providing sufficient buffer time for transfer and reducing the workload of human agents. The response logic range analysis module and the rapid transfer module effectively train the intelligent customer service system to identify uncertainties in responses and quickly transfer the user to the intelligent customer service system, significantly improving its response speed in emergency situations. During training, the classification training module trains the intelligent customer service system on different types of questions in different work domains and contexts, improving its overall responsiveness. This allows the trained intelligent customer service system to adapt to different work environments, expanding its application scope. The response volume setting module allows you to set the maximum number of data responses the intelligent customer service can provide in a single response under normal conditions. This means that during a response, it can automatically match multiple types of data related to the question. The module allows you to directly set the number of responses. For example, in a business transaction scenario, setting a higher response volume allows for matching multiple relevant data and links during the response process, including business transaction procedures, avoiding the need for secondary inquiries. In a merchant scenario, setting a lower response volume allows for precise, single-response responses to the question, preventing the customer from becoming impatient due to receiving too many responses at once. This enables the intelligent customer service to better adapt to different contexts. During training simulation, a multi-window hidden training module allows for opening multiple windows for training, increasing training speed and enabling the intelligent customer service to perform a large number of training sessions simultaneously. By hiding the windows, the question matching analysis module, question resolution rate analysis module, and training analysis unit can quickly identify and analyze the hidden training dialogues. This allows for the observation and analysis of problems without real-time monitoring of the intelligent customer service, improving its training and optimization capabilities.
[0049] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.
Claims
1. A training and research platform for intelligent customer service based on high-precision semantic understanding technology, comprising a data filling unit, a database, a learning unit, a training unit, a training result analysis unit, and a correction unit, characterized in that, The data filling unit is connected to the database, and the training unit is also connected to the database. The database is a repository for storing all data. The data filling unit is used to fill new data into the database. The learning unit is used for the intelligent customer service to learn. The training unit is used to train the intelligent customer service after it has learned. The training result analysis unit is used to perform an overall analysis of the training results. The correction unit corrects the intelligent customer service's responses based on the results analyzed from the training results and the content in the database. The learning unit includes a response logic setting module, a response logic import module, a response feature setting module, a human simulation response module, and a database update prompt module, wherein the database update prompt module is connected to the database; The reply feature setting module includes a classification tone feature setting module, an attached emoji setting module, and a single reply quantity setting module; The real-person simulation reply module includes a tone feature matching module, an attached emoji matching module, a reply logic range analysis module, and a quick transfer module; The training unit includes a multi-window hiding training module, a classification training module, a question matching rate analysis module, a question solving rate analysis module, and a typical question selection module; The training result analysis unit includes a typical question set module, a wrong question set module, and a wrong question sorting and analysis module. The typical question set module is connected to the typical question selection module, and the wrong question set module is connected to the classification training module.
2. The intelligent customer service training and research platform based on high-precision semantic understanding technology according to claim 1, characterized in that, The response logic setting module is used to set the response logic range limit data for the intelligent customer service, including setting logic range threshold data. The response logic import module is used to import response logic and response templates from the database. The response feature setting module is used to set the features in the intelligent customer service response. The human simulation response module is used for the robot to learn human simulation responses on the platform. The database update prompt module is used to prompt the database update status.
3. The intelligent customer service training and research platform based on high-precision semantic understanding technology according to claim 2, characterized in that, The classification tone feature setting module is used to set the tone features of the intelligent customer service in a specific environment after classification when the intelligent customer service is in a simulated human response state. The attached emoji setting module is used to set the attached emoji data of the intelligent customer service in a simulated human response state. The single reply quantity setting module is used to set the maximum number of data replies that the intelligent customer service can send in a single reply under normal reply conditions.
4. The intelligent customer service training and research platform based on high-precision semantic understanding technology according to claim 1, characterized in that, The tone feature matching module is used to match corresponding tone features in the intelligent customer service's simulated human response state. The attached emoji matching module is used to match corresponding emojis in the intelligent customer service's simulated human response state. The response logic range analysis module is used to train the intelligent customer service to recognize uncertain response states in the simulated human response state. The quick transfer module is used to train the intelligent customer service to quickly transfer to human customer service.
5. The intelligent customer service training and research platform based on high-precision semantic understanding technology according to claim 1, characterized in that, The multi-window hiding training module is used to open multiple windows for categorized and grouped training during the intelligent customer service training process, and hide multiple windows used for training. The classification training module is used to classify and train the intelligent customer service according to different contextual needs and different work fields. The question matching rate analysis module is used to analyze the question matching rate of the intelligent customer service based on the training results. The question resolution rate analysis module is used to analyze the question resolution rate of the intelligent customer service based on the training results. The typical question selection module filters out atypical questions based on the training results and selects the typical questions that are answered.
6. The intelligent customer service training and research platform based on high-precision semantic understanding technology according to claim 5, characterized in that, The classification training module includes a question bank classification search module, a question type import training module, an intelligent response training module, and a high-precision semantic understanding module. The question bank classification search module is connected to the database.
7. The intelligent customer service training and research platform based on high-precision semantic understanding technology according to claim 6, characterized in that, The question bank classification and search module is used to search the database for question banks with different contextual needs and different work fields. The question type import and training module is used to import the searched question banks and import them into the platform for training. The intelligent reply training module is used for intelligent customer service to conduct dialogue training based on the question bank. The high-precision semantic understanding module is based on high-precision semantic understanding technology to train the semantic understanding function of intelligent customer service during the dialogue training process.
8. The intelligent customer service training and research platform based on high-precision semantic understanding technology according to claim 7, characterized in that, The typical question collection module is used to collect and statistically analyze typical questions and their responses during the training process. The incorrect question collection module is used to collect and statistically analyze incorrect questions during training. The incorrect question sorting and analysis module is used to sort and analyze incorrect questions according to their categories.