Customer balance management system based on intelligent robot conversation

Abstract

The invention discloses a customer balance management system based on intelligent robot dialogue, which relates to the technical field of intelligent customer service and management and comprises an intelligent robot dialogue platform module, a balance management module, a payment module and a data analysis module. The intelligent robot dialogue platform module is used as a main interface for interaction between a user and the system and realizes dialogue communication with the user through a natural language processing technology. According to the customer balance management system, dialogue management is carried out through the intelligent robot, the system can automatically process balance query, recharging and conventional reminding operation of a large number of customers, the customer service response speed and processing efficiency are remarkably improved, the workload of manual customer service is reduced, 24-hour uninterrupted service is provided by the intelligent robot, and the customer service efficiency is improved. The customer can inquire and manage the balance at any time and any place, and does not need to wait for the working time of manual customer service, thereby greatly improving the satisfaction and convenience of the customer.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent customer service and management, and in particular to a customer balance management system based on intelligent robot dialogue. Background Art

[0002] With the rapid development of the logistics industry and the strengthening of the trend of online and offline integration, logistics customers have increasingly higher requirements for service efficiency, accuracy and convenience. Traditional customer balance management methods often rely on manual inquiries or simple self-service systems, which have problems such as low query efficiency, complex operations, and untimely information updates. In addition, with the increase in the number of customers and the increase in transaction frequency, how to quickly and accurately manage customer balances and provide personalized service experience has become a technical problem that the logistics industry needs to solve urgently;

[0003] In the existing technology, although some logistics systems have begun to try to introduce intelligent customer service functions, most of them focus on simple information query and feedback, and lack deep integration and optimization for customer balance management. Therefore, developing a system that can realize automatic management and query of customer balances based on intelligent robot dialogue is of great significance to improving the service quality and operational efficiency of the logistics industry. Summary of the invention

[0004] The purpose of the present invention is to provide a customer balance management system based on intelligent robot dialogue to solve the problems raised in the above background technology.

[0005] A customer balance management system based on intelligent robot dialogue, comprising: an intelligent robot dialogue platform module, a balance management module, a payment module and a data analysis module;

[0006] The intelligent robot dialogue platform module is used as the main interface for users to interact with the system. It uses natural language processing technology to achieve dialogue and communication with users and provide personalized service customization;

[0007] The balance management module is used to process business logic related to customer balances. It supports the intelligent recharge function for branches, automatically recharges customer accounts according to demand, sets a balance warning threshold, and automatically triggers the recharge process when the customer account balance is lower than this threshold, and sends a recharge reminder to the customer via SMS;

[0008] The payment module is used to be responsible for the actual payment process when the intelligent robot dialogue platform module receives the user's recharge request and it is reviewed by the balance management module;

[0009] The data analysis module is used to process and analyze the data collected by the system. It uses system data to conduct statistical analysis on the average daily order volume of customers, optimize service strategies and customer management, monitor system operation status, detect abnormal situations in real time, and trigger corresponding exception handling processes to ensure the stability and reliability of the system.

[0010] Preferably, the intelligent robot dialogue platform module specifically includes:

[0011] Rasa3.6 is selected as the framework of the dialogue system;

[0012] Through big data and offline interviews, we collected the balance management needs of logistics industry personnel, including balance query, recharge request and balance warning. We used JiebaTokenizer to segment the text entered by users into Chinese words, and applied LanguageModelFeaturizer combined with BERT word segmenter to generate language feature representation.

[0013] Determine the conversation intent that the intelligent robot needs to recognize based on the needs of the logistics industry;

[0014] Define different dialogue scenarios and user behavior paths, and write stories for each scenario, describing the interaction process between the user and the robot to construct the dialogue story;

[0015] For scenarios where multiple pieces of information need to be collected, design forms and define form fields and validation rules;

[0016] Use Rasa's natural language understanding component to perform intent recognition and entity extraction on the text entered by the user, and improve the accuracy of intent recognition and entity extraction through training data and machine learning algorithms;

[0017] Determine the robot’s response strategy based on the identified intent and extracted entities;

[0018] When you need to collect user information, start the form and guide the user to fill in the fields of the form step by step, and ensure the accuracy of the information by asking questions and confirming them.

[0019] Preferably, the intent recognition and entity extraction specifically include:

[0020] DIETClassifier is used as the intent recognition model to train and learn the intent of user input;

[0021] Use RegexEntityExtractor to quickly match and extract key entities;

[0022] Introduce EntitySynonymMapper to build a synonym library to solve the recognition problem caused by the diversity of user input.

[0023] Preferably, DIETClassifier is used as the intent recognition model, and the training and learning of the intent of user input specifically includes:

[0024] Collect a large amount of user input text related to balance management in the logistics industry, manually annotate the collected text, and clarify the intent category of each text;

[0025] Preprocess the data to remove impurities in the text that affect model training, including special characters and punctuation marks, and convert the text into a unified format;

[0026] Set the parameters of DIETClassifier according to specific needs and input the preprocessed annotated text data into DIETClassifier;

[0027] DIETClassifier first encodes the input text, converts the text into a numerical vector representation, and performs feature extraction and intent classification through a multi-layer neural network. During the training process, the weight parameters of the model are continuously adjusted to minimize the loss function.

[0028] Among them, the calculation formula of the loss function is:

[0029]

[0030] In the formula, L represents the loss value, i represents the index of summation, and y i represents the true intent category label, p i Represents the probability distribution of intent categories predicted by the model;

[0031] Use precision, recall, and F1 value metrics to evaluate model performance;

[0032] The calculation formula for accuracy is:

[0033]

[0034] Among them, the calculation formula of recall rate Recall is:

[0035]

[0036] Among them, the F1 value calculation formula is:

[0037]

[0038] In the formula, Precision represents the accuracy, and the accuracy calculation formula is:

[0039]

[0040] According to the evaluation results, the problems existing in the model are analyzed and corresponding optimization is performed, which includes: adding more training data to improve the generalization ability of the model, adjusting the hyperparameters of the model, and trying different optimization algorithms.

[0041] Preferably, the intelligent robot dialogue platform module executes corresponding business logic operations according to the identified user intentions and extracted entity information. The balance management module implements the balance warning function, sets thresholds and monitors customer balances. When the balance is lower than the threshold, the recharge process is automatically triggered, and reminders are sent through SMS services. The automatic management logic of disabling customers at outlets is introduced. When a customer is disabled, the balance in the database is automatically withdrawn to avoid waste of resources. The data analysis module provides a customer average daily order analysis function, which is calculated and displayed based on historical data to provide data support for outlet operations.

[0042] Preferably, the intelligent robot dialogue platform module includes:

[0043] Data collection unit, which is used to collect the balance management needs of logistics industry related personnel through big data and offline visits;

[0044] Chinese word segmentation unit, which is used to perform Chinese word segmentation on the text input by the user using JiebaTokenizer;

[0045] The language feature generation unit is used to apply LanguageModelFeaturizer in combination with BERT word segmenter to generate language feature representation and convert the text input by the user into a numerical vector form that can be processed by the model;

[0046] Intent recognition unit: The intent recognition unit is used to use DIETClassifier as the intent recognition model to train and learn the intent of user input. By learning from a large amount of annotated text data, it can accurately identify the intent category of the user input text.

[0047] Entity extraction unit: The entity extraction unit is used to use RegexEntityExtractor to quickly match and extract key entities, extract important information related to balance management from the text entered by the user, introduce EntitySynonymMapper, build a synonym library, solve the recognition problem caused by the diversity of user input, and improve the accuracy and robustness of entity extraction;

[0048] A response strategy determination unit, the response strategy determination unit is used to determine the robot's response strategy according to the recognized intention and the extracted entity;

[0049] The form processing unit is used to design forms, define form fields and validation rules for scenarios where multiple pieces of information need to be collected.

[0050] Preferably, the balance management module includes:

[0051] Recharge function unit: The recharge function unit is used to support the intelligent recharge function of the outlets for customers, automatically recharge the customer's account according to demand, and meet the customer's fund replenishment needs when the balance is insufficient;

[0052] A threshold setting unit, which is used to set a balance warning threshold to provide a standard for determining whether a customer's balance is too low;

[0053] Monitoring unit: The monitoring unit is used to monitor customer balances and keep an eye on customer account funds in real time;

[0054] A recharge trigger unit is used to automatically trigger the recharge process and start the fund replenishment procedure when the customer account balance is lower than the warning threshold;

[0055] SMS reminder unit, which is used to send recharge reminders to customers via SMS, timely inform customers of their balance status and recharge requirements, and increase customers' attention to their balances;

[0056] The customer disabled management unit is used to introduce automatic management logic for disabling customers at outlets. When a customer is disabled, its balance in the library is automatically withdrawn to avoid waste of resources.

[0057] Preferably, the balance management module includes:

[0058] The audit receiving unit is used to receive user recharge requests from the intelligent robot dialogue platform and conduct audits in conjunction with the balance management module to ensure the legality and rationality of the recharge requests;

[0059] The payment execution unit is responsible for the actual payment process after the recharge request is reviewed and recharges the funds into the customer account safely and accurately.

[0060] Preferably, the balance management module includes:

[0061] Data processing unit, the data processing unit is used to process the data collected by the system, including cleaning, sorting and conversion;

[0062] Statistical analysis unit: The statistical analysis unit is used to use system data to conduct statistical analysis on the average daily order volume of customers and obtain the business rules and trends of customers;

[0063] Status monitoring unit, the status monitoring unit is used to monitor the system operation status and detect in real time whether the system is running normally.

[0064] Preferably, the system data is used to conduct statistical analysis on the average daily order volume of customers, and the business rules and trends of customers are obtained, including:

[0065] Collect data related to customer orders from the system, including order time, order amount, and customer information;

[0066] Preprocessing the collected data, including: data cleaning, data sorting and data conversion;

[0067] The average daily order volume is calculated by determining the time range and counting the total number of orders, where the calculation formula for the average daily order volume is:

[0068] DAOQ=TO / NDR

[0069] In the formula, DAOQ represents the average daily order volume, TO represents the total number of orders, and NDR represents the number of days in the time range;

[0070] Analyze business patterns and trends through time series analysis, seasonality analysis, customer behavior analysis, and association analysis;

[0071] The statistical analysis results are presented to branch operators in the form of charts.

[0072] By adopting the above technical solution, a customer balance management system for intelligent robot dialogue is realized.

[0073] Compared with the prior art, the beneficial effects of the present invention are: the customer balance management system,

[0074] Through intelligent robots for dialogue management, the system can automatically handle routine operations such as balance inquiries, recharges, and reminders for a large number of customers, significantly improving customer service response speed and processing efficiency, and reducing the workload of manual customer service. Intelligent robots provide 24-hour uninterrupted service, and customers can conduct balance-related inquiries and management anytime and anywhere without waiting for the working hours of manual customer service, thereby greatly improving customer satisfaction and convenience. Compared with the traditional manual customer service model, the system reduces dependence on human resources and reduces the company's operating costs. At the same time, through continuous learning and optimization, intelligent robots can understand customer needs more accurately, reduce misoperations and service errors, and further control costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] See also Figure 1 ,The present invention provides a technical solution: a customer balance management system based on intelligent robot dialogue, comprising: an intelligent robot dialogue platform module, a balance management module, a payment module and a data analysis module;

[0078] The intelligent robot dialogue platform module specifically includes:

[0079] Rasa3.6 is selected as the framework of the dialogue system;

[0080] Through big data and offline interviews, we collected the balance management needs of logistics industry personnel, including balance query, recharge request and balance warning. We used JiebaTokenizer to segment the text entered by users into Chinese words, and applied LanguageModelFeaturizer combined with BERT word segmenter to generate language feature representation.

[0081] Determine the conversation intent that the intelligent robot needs to recognize based on the needs of the logistics industry;

[0082] Define different dialogue scenarios and user behavior paths, and write stories for each scenario, describing the interaction process between the user and the robot to construct the dialogue story;

[0083] For scenarios where multiple pieces of information need to be collected, design forms and define form fields and validation rules;

[0084] Use Rasa's natural language understanding component to perform intent recognition and entity extraction on the text entered by the user, and improve the accuracy of intent recognition and entity extraction through training data and machine learning algorithms;

[0085] Determine the robot’s response strategy based on the identified intent and extracted entities;

[0086] When you need to collect user information, start the form and guide the user to fill in the fields of the form step by step, and ensure the accuracy of the information by asking questions and confirming them.

[0087] The intent recognition and entity extraction specifically include:

[0088] DIETClassifier is used as the intent recognition model to train and learn the intent of user input;

[0089] Use RegexEntityExtractor to quickly match and extract key entities;

[0090] Introduce EntitySynonymMapper to build a synonym library to solve the recognition problem caused by the diversity of user input.

[0091] Preferably, DIETClassifier is used as the intent recognition model, and the training and learning of the intent of user input specifically includes:

[0092] Collect a large amount of user input text related to balance management in the logistics industry, manually annotate the collected text, and clarify the intent category of each text;

[0093] Preprocess the data to remove impurities in the text that affect model training, including special characters and punctuation marks, and convert the text into a unified format;

[0094] Set the parameters of DIETClassifier according to specific needs and input the preprocessed annotated text data into DIETClassifier;

[0095] DIETClassifier first encodes the input text, converts the text into a numerical vector representation, and performs feature extraction and intent classification through a multi-layer neural network. During the training process, the weight parameters of the model are continuously adjusted to minimize the loss function.

[0096] Among them, the calculation formula of the loss function is:

[0097]

[0098] In the formula, L represents the loss value, i represents the index of summation, and y i represents the true intent category label, p i Represents the probability distribution of intent categories predicted by the model;

[0099] Use precision, recall, and F1 value metrics to evaluate model performance;

[0100] The calculation formula for accuracy is:

[0101]

[0102] Among them, the calculation formula of recall rate Recall is:

[0103]

[0104] Among them, the F1 value calculation formula is:

[0105]

[0106] In the formula, Precision represents the accuracy, and the accuracy calculation formula is:

[0107]

[0108] According to the evaluation results, analyze the problems existing in the model and perform corresponding optimization, which includes: adding more training data to improve the generalization ability of the model, adjusting the hyperparameters of the model, and trying different optimization algorithms;

[0109] The intelligent robot dialogue platform module has powerful functions, can accurately identify the user's intention, and efficiently extract entity information. On this basis, the module will strictly follow the corresponding business logic operations to perform tasks, whether it is answering users' questions, providing specific services, or guiding users to complete a series of complex processes, it can be done accurately, efficiently and conveniently;

[0110] The balance management module plays an important role. It can realize the balance warning function, carefully set reasonable thresholds, and monitor the customer's balance at all times. Once the customer's balance is lower than the set threshold, the module will automatically trigger the recharge process to ensure that the customer can replenish the balance in time without affecting the normal operation of the business. At the same time, it will also send reminders to customers through SMS services so that customers can know their balance status at the first time. In addition, the module also introduces the automatic management logic of disabled customers at the branch. When a customer is disabled, the balance in the library can be automatically withdrawn, effectively avoiding the waste of resources and improving the efficiency of resource utilization.

[0111] Furthermore, the data analysis module provides strong data support for branch operations. It provides the function of analyzing the average daily order volume of customers, can perform accurate calculations based on rich historical data, and display them in a clear and intuitive manner. Through the analysis of the average daily order volume of customers, branch operators can better understand customer needs and behavior patterns, thereby formulating more scientific and reasonable operation strategies and improving the overall operation efficiency and service quality of the branches.

[0112] Using system data, we conduct statistical analysis on the average daily order volume of customers, and obtain the business rules and trends of customers, including:

[0113] Collect data related to customer orders from the system, including order time, order amount, and customer information;

[0114] Preprocessing the collected data, including: data cleaning, data sorting and data conversion;

[0115] The average daily order volume is calculated by determining the time range and counting the total number of orders, where the calculation formula for the average daily order volume is:

[0116] DAOQ=TO / NDR

[0117] In the formula, DAOQ represents the average daily order volume, TO represents the total number of orders, and NDR represents the number of days in the time range;

[0118] Analyze business patterns and trends through time series analysis, seasonal analysis, customer behavior analysis and correlation analysis; present statistical analysis results to branch operators in the form of charts.

Claims

1. A customer balance management system based on intelligent robot dialogue, characterized in that: Includes: intelligent robot dialogue platform module, balance management module, payment module and data analysis module; The intelligent robot dialogue platform module is used as the main interface for users to interact with the system. It uses natural language processing technology to achieve dialogue and communication with users and provide personalized service customization; The balance management module is used to process business logic related to customer balances. It supports the intelligent recharge function for branches, automatically recharges customer accounts according to demand, sets a balance warning threshold, and automatically triggers the recharge process when the customer account balance is lower than this threshold, and sends a recharge reminder to the customer via SMS; The payment module is used to be responsible for the actual payment process when the intelligent robot dialogue platform module receives the user's recharge request and it is reviewed by the balance management module; The data analysis module is used to process and analyze the data collected by the system. It uses system data to conduct statistical analysis on the average daily order volume of customers, optimize service strategies and customer management, monitor system operation status, detect abnormal situations in real time, and trigger corresponding exception handling processes to ensure the stability and reliability of the system.

2. A customer balance management system based on intelligent robot dialogue according to claim 1, characterized in that: The intelligent robot dialogue platform module specifically includes: Rasa3.6 is selected as the framework of the dialogue system; Through big data and offline interviews, we collected the balance management needs of logistics industry personnel, including balance query, recharge request and balance warning. We used JiebaTokenizer to segment the text entered by users into Chinese words, and applied LanguageModelFeaturizer combined with BERT word segmenter to generate language feature representation. Determine the conversation intent that the intelligent robot needs to recognize based on the needs of the logistics industry; Define different dialogue scenarios and user behavior paths, and write stories for each scenario, describing the interaction process between the user and the robot to construct the dialogue story; For scenarios where multiple pieces of information need to be collected, design forms and define form fields and validation rules; Use Rasa's natural language understanding component to perform intent recognition and entity extraction on the text entered by the user, and improve the accuracy of intent recognition and entity extraction through training data and machine learning algorithms; Determine the robot’s response strategy based on the identified intent and extracted entities; When you need to collect user information, start the form and guide the user to fill in the fields of the form step by step, and ensure the accuracy of the information by asking questions and confirming them.

3. A customer balance management system based on intelligent robot dialogue according to claim 2, characterized in that: The intent recognition and entity extraction specifically include: DIETClassifier is used as the intent recognition model to train and learn the intent of user input; Use RegexEntityExtractor to quickly match and extract key entities; Introduce EntitySynonymMapper to build a synonym library to solve the recognition problem caused by the diversity of user input.

4. A customer balance management system based on intelligent robot dialogue according to claim 3, characterized in that: DIETClassifier is used as the intent recognition model. The training and learning of the user input intent specifically include: Collect a large amount of user input text related to balance management in the logistics industry, manually annotate the collected text, and clarify the intent category of each text; Preprocess the data to remove impurities in the text that affect model training, including special characters and punctuation marks, and convert the text into a unified format; Set the parameters of DIETClassifier according to specific needs and input the preprocessed annotated text data into DIETClassifier; DIETClassifier first encodes the input text, converts the text into a numerical vector representation, and performs feature extraction and intent classification through a multi-layer neural network. During the training process, the weight parameters of the model are continuously adjusted to minimize the loss function. Among them, the calculation formula of the loss function is: In the formula, L represents the loss value, i represents the index of summation, and y i represents the true intent category label, p i Represents the probability distribution of intent categories predicted by the model; Use precision, recall, and F1 value metrics to evaluate model performance; The calculation formula for accuracy is: Among them, the calculation formula of recall rate Recall is: Among them, the F1 value calculation formula is: In the formula, Precision represents the accuracy, and the accuracy calculation formula is: According to the evaluation results, the problems existing in the model are analyzed and corresponding optimization is performed, which includes: adding more training data to improve the generalization ability of the model, adjusting the hyperparameters of the model, and trying different optimization algorithms.

5. A customer balance management system based on intelligent robot dialogue according to claim 1, characterized in that: The intelligent robot dialogue platform module executes corresponding business logic operations according to the identified user intentions and extracted entity information. The balance management module implements the balance warning function, sets thresholds and monitors customer balances. When the balance is lower than the threshold, the recharge process is automatically triggered, and reminders are sent through SMS services. The automatic management logic of disabling customers at outlets is introduced. When a customer is disabled, the balance in the database is automatically withdrawn to avoid waste of resources. The data analysis module provides a customer average daily order analysis function, which is calculated and displayed based on historical data to provide data support for outlet operations.

6. A customer balance management system based on intelligent robot dialogue according to claim 1, characterized in that: Intelligent robot dialogue platform modules include: Data collection unit, which is used to collect the balance management needs of logistics industry related personnel through big data and offline visits; Chinese word segmentation unit, which is used to perform Chinese word segmentation on the text input by the user using JiebaTokenizer; The language feature generation unit is used to apply LanguageModelFeaturizer in combination with BERT word segmenter to generate language feature representation and convert the text input by the user into a numerical vector form that can be processed by the model; Intent recognition unit: The intent recognition unit is used to use DIETClassifier as the intent recognition model to train and learn the intent of user input. By learning from a large amount of annotated text data, it can accurately identify the intent category of the user input text. Entity extraction unit: The entity extraction unit is used to use RegexEntityExtractor to quickly match and extract key entities, extract important information related to balance management from the text entered by the user, introduce EntitySynonymMapper, build a synonym library, solve the recognition problem caused by the diversity of user input, and improve the accuracy and robustness of entity extraction; A response strategy determination unit, the response strategy determination unit is used to determine the robot's response strategy according to the recognized intention and the extracted entity; The form processing unit is used to design forms, define form fields and validation rules for scenarios where multiple pieces of information need to be collected.

7. A customer balance management system based on intelligent robot dialogue according to claim 1, characterized in that: The balance management module includes: Recharge function unit: The recharge function unit is used to support the intelligent recharge function of the outlets for customers, automatically recharge the customer's account according to demand, and meet the customer's fund replenishment needs when the balance is insufficient; A threshold setting unit, which is used to set a balance warning threshold to provide a standard for determining whether a customer's balance is too low; Monitoring unit: The monitoring unit is used to monitor customer balances and keep an eye on customer account funds in real time; A recharge trigger unit is used to automatically trigger the recharge process and start the fund replenishment procedure when the customer account balance is lower than the warning threshold; SMS reminder unit, which is used to send recharge reminders to customers via SMS, timely inform customers of their balance status and recharge requirements, and increase customers' attention to their balances; The customer disabled management unit is used to introduce automatic management logic for disabling customers at outlets. When a customer is disabled, its balance in the library is automatically withdrawn to avoid waste of resources.

8. A customer balance management system based on intelligent robot dialogue according to claim 1, characterized in that: The balance management module includes: The audit receiving unit is used to receive user recharge requests from the intelligent robot dialogue platform and conduct audits in conjunction with the balance management module to ensure the legality and rationality of the recharge requests; The payment execution unit is responsible for the actual payment process after the recharge request is reviewed and recharges the funds into the customer account safely and accurately.

9. A customer balance management system based on intelligent robot dialogue according to claim 1, characterized in that: The balance management module includes: Data processing unit, the data processing unit is used to process the data collected by the system, including cleaning, sorting and conversion; Statistical analysis unit: The statistical analysis unit is used to use system data to conduct statistical analysis on the average daily order volume of customers and obtain the business rules and trends of customers; Status monitoring unit, the status monitoring unit is used to monitor the system operation status and detect in real time whether the system is running normally.

10. A customer balance management system based on intelligent robot dialogue according to claim 9, characterized in that: Using system data, we conduct statistical analysis on the average daily order volume of customers, and obtain the business rules and trends of customers, including: Collect data related to customer orders from the system, including order time, order amount, and customer information; Preprocessing the collected data, including: data cleaning, data sorting and data conversion; The average daily order volume is calculated by determining the time range and counting the total number of orders, where the calculation formula for the average daily order volume is: DAOQ=TO / NDR In the formula, DAOQ represents the average daily order volume, TO represents the total number of orders, and NDR represents the number of days in the time range; Analyze business patterns and trends through time series analysis, seasonality analysis, customer behavior analysis, and association analysis; The statistical analysis results are presented to branch operators in the form of charts.

Images