Communication customer service system based on artificial intelligence

Abstract

The invention discloses a communication customer service system based on artificial intelligence. The system comprises a data acquisition module, a data analysis module, an intelligent reply module and a user data analysis module. The data acquisition module is used for acquiring historical data and real-time interaction data of a user; the data analysis module is used for analyzing the key value and the final result of the data acquisition module; the intelligent reply module generates reply content according to the analysis result and sends the reply content to the user through voice or text data; and the user data analysis module is used for deeply analyzing interaction records and feedback of the customer and the intelligent customer service. The invention belongs to the technical field of communication, and achieves the technical effects that the intelligentization, individuation and high efficiency of customer service are realized through the application of big data analysis and artificial intelligence technology, and the system has high efficiency, individuation and intelligentization, and can provide more accurate and intimate service for users.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a communication customer service system based on artificial intelligence. Background Art

[0002] With the rapid development of communication technology, the demand for customer service is increasing. In terms of customer service, many industries are now beginning to use intelligent chat to answer customers' questions online. Currently, customer service personnel solve customers' problems online one-to-many. This requires more customer service personnel. Basic questions still require the next round of answers after multiple answers. The frequency of repeating answers to the same question is very high, resulting in inefficient response to questions. Therefore, traditional manual customer service methods can no longer meet the needs of efficient and accurate customer service.

[0003] The rise of big data and artificial intelligence technology has provided strong support for the intelligentization of communication customer service systems. Therefore, it is particularly important to develop a communication customer service system that can use artificial intelligence technology to achieve efficient and accurate customer service. Summary of the invention

[0004] To this end, the present invention provides a communication customer service system based on artificial intelligence to solve the above-mentioned problems in the prior art.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A communication customer service system based on artificial intelligence, comprising: a data collection module, a data analysis module, an intelligent reply module and a user data analysis module;

[0007] The data collection module collects historical data and real-time interaction data of users, defines MapReduce tasks, counts key values, and outputs the final results in text data;

[0008] The data analysis module is used to analyze the key value and final result of the data acquisition module, extract useful information, and obtain the analysis results;

[0009] The intelligent reply module generates reply content based on the analysis results and sends it to the user via voice or text data;

[0010] The user data analysis module deeply analyzes the interaction records and feedback between customers and intelligent customer service, and fine-tunes the customer model.

[0011] Furthermore, the data acquisition module includes the following steps:

[0012] S1: Collect users’ historical call records and real-time interaction data;

[0013] S2: Preprocess the collected data, including denoising, format conversion and data cleaning;

[0014] S3: Data partitioning: Divide a large data set into multiple small data sets;

[0015] S4: Build MapReduce framework configuration;

[0016] S5: Map function execution;

[0017] S6: Execute Shuffle and Sort, including:

[0018] S6.1: Data partitioning: According to the key value of the key-value pair, the intermediate results are divided into different partitions, and each partition corresponds to a Reduce task;

[0019] S6.2: Data sorting: In each partition, sort the key-value pairs so that the Reduce function can efficiently process records with the same key value;

[0020] S7: Reduce, including:

[0021] S7.1: Data reading: The Reduce task reads the intermediate results of the corresponding partition from the distributed file system;

[0022] S7.2: Reduce function execution: Iterate the key-value pairs in each partition and execute the Reduce function. The Reduce function merges or calculates the records with the same key value to generate the final output result.

[0023] S7.3: Output result storage: The final result generated by the Reduce stage is stored in the specified output path. The output result is output and stored in text format.

[0024] Furthermore, step S4 specifically includes:

[0025] S4.1: Cluster configuration: Configure the Hadoop distributed computing cluster;

[0026] S4.2: Task configuration: Define MapReduce tasks and configure task parallelism and resource allocation based on data volume and processing requirements.

[0027] Furthermore, step S5 specifically includes:

[0028] S5.1: Data reading: Read data blocks and distribute them to worker nodes in the cluster;

[0029] S5.2: Function execution: The Map function is executed on each working node to process each record in the data block. The Map function converts the input record into a series of key-value pairs, which are output as intermediate results.

[0030] S5.3: Intermediate result storage: The intermediate results generated in the Map stage are stored in the distributed file system, and the intermediate results are partitioned and sorted.

[0031] Furthermore, the intelligent reply module includes the following steps:

[0032] S1: Data preprocessing: convert text data into word vectors;

[0033] S2: Use a pre-trained large language model to extract features from text data;

[0034] S3: LSTM network construction, including:

[0035] S3.1: Add a word embedding layer. Add a word embedding layer before the LSTM network.

[0036] S3.2: Build an LSTM backbone network to process text sequence data and extract features;

[0037] S3.3: Build a fully connected layer to map the high-dimensional features extracted by the LSTM network into a low-dimensional classification space to obtain the final classification result or reply content;

[0038] S4: Model training and optimization;

[0039] S5: Reply generation. After the model training is completed, the trained model is used to generate the reply content. The text data input by the user is preprocessed and input into the model. The model generates the corresponding reply content based on the extracted features.

[0040] Further, S4 includes the following steps:

[0041] S4.1: Loss function selection;

[0042] S4.2: Optimization algorithm selection, select SGD training model;

[0043] S4.3: Model evaluation and tuning, use the validation set to evaluate the performance of the model and tune the model based on the evaluation results.

[0044] Furthermore, the user data analysis module includes the following steps:

[0045] S1: Data collection: Collect the interaction records between customers and robots, and organize customers’ feedback after using intelligent customer service;

[0046] S2: Data cleaning: Clean the collected interaction records, remove irrelevant information, and standardize the feedback data;

[0047] S3: Data analysis: Use natural language processing technology to analyze the conversation between customers and intelligent customer service, extract key information, and analyze customer feedback data;

[0048] S4: Model evaluation: Based on the interaction records and feedback data between customers and intelligent customer service, the performance of the current intelligent customer service model is evaluated;

[0049] S5: Problem location: Based on the evaluation results, locate the problems and deficiencies in the intelligent customer service model and identify the reasons why the model performs poorly or fails in specific scenarios;

[0050] S6: Model optimization: Based on the results of problem location, fine-tune the intelligent customer service model and optimize the model’s algorithm, parameters, and feature engineering;

[0051] S7: Testing and Validation: Test the fine-tuned model.

[0052] Furthermore, the user data analysis module also includes: user portrait construction, which constructs user portraits based on the interaction records and feedback data between customers and intelligent customer service;

[0053] Application of data mining algorithms: Use data mining algorithms to analyze user portraits, identify similarities between users, and discover potential patterns and trends in user preferences;

[0054] Personalized recommendation generation: Generate personalized recommendation content based on the results of data mining.

[0055] The present invention has the following advantages:

[0056] The present invention realizes intelligent, personalized and efficient customer service through the application of big data analysis and artificial intelligence technology. The system is efficient, personalized and intelligent, and can provide users with more accurate and considerate services. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A module connection diagram of an artificial intelligence-based communication customer service system provided by the present invention.

[0058] Figure 2 An operation flow chart of an intelligent reply module of an artificial intelligence-based communication customer service system provided by the present invention. DETAILED DESCRIPTION

[0059] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are 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.

[0060] See also Figure 1 to Figure 2 As shown, the present invention provides an artificial intelligence-based communication customer service system, including: a data collection module, a data analysis module, an intelligent reply module and a user data analysis module;

[0061] The data collection module collects historical data and real-time interaction data of users, defines MapReduce tasks, counts key values, and outputs the final results in text data. Specifically, it includes the following steps:

[0062] S1: The data collection module collects the user's historical call records and real-time interaction data, and captures the user's call voice, text chat records, and interaction information through other channels (such as social media, emails, etc.) in real time and accurately. The historical call records include the user's past call time, call duration, call content (processed text or voice data), and call results. The historical call records help the system analyze the user's communication habits and needs;

[0063] Real-time interaction data: mainly refers to the interaction information between the user and the system in the current session, such as the text, voice, and click behavior input by the user. Real-time interaction data can reflect the user's needs and intentions in real time;

[0064] S2: Preprocess the collected data, including denoising, format conversion and data cleaning, to ensure the accuracy and usability of the data;

[0065] S3: Data partitioning: Divide a large data set into multiple small data sets (or data blocks) for parallel processing in a distributed system;

[0066] S4: Build the MapReduce framework configuration, including:

[0067] S4.1: Cluster configuration: Configure the Hadoop distributed computing cluster to ensure that the nodes (or working nodes) in the cluster can communicate and work together normally;

[0068] S4.2: Task configuration: Define MapReduce tasks, including the path of input data, the path of output data, the implementation of Map function and Reduce function, etc. Configure the parallelism and resource allocation of tasks according to the data volume and processing requirements;

[0069] S5: Map function execution, including:

[0070] S5.1: Data reading: Read data blocks and distribute them to worker nodes in the cluster;

[0071] S5.2: Function execution: The Map function is executed on each working node to process each record in the data block. The Map function converts the input record into a series of key-value pairs, which are output as intermediate results.

[0072] S5.3: Intermediate result storage: The intermediate results generated in the Map phase are stored in a distributed file system (such as HDFS) for use in the subsequent Reduce phase. The intermediate results need to be partitioned and sorted to improve the processing efficiency of the Reduce phase.

[0073] S6: Execute Shuffle and Sort, including:

[0074] S6.1: Data partitioning: According to the key value of the key-value pair, the intermediate results are divided into different partitions, and each partition corresponds to a Reduce task;

[0075] S6.2: Data sorting: In each partition, sort the key-value pairs so that the Reduce function can efficiently process records with the same key value;

[0076] S7: Reduce, including:

[0077] S7.1: Data reading: The Reduce task reads the intermediate results of the corresponding partition from the distributed file system;

[0078] S7.2: Reduce function execution: Iterate the key-value pairs in each partition and execute the Reduce function. The Reduce function merges or calculates the records with the same key value to generate the final output result.

[0079] S7.3: Output result storage: The final result generated by the Reduce stage is stored in the specified output path. The output result is output and stored in text format.

[0080] The data analysis module is used to analyze the key values ​​and final results of the data acquisition module, extract useful information, and obtain analysis results. Specifically: use statistics (such as mean, median, mode, standard deviation, etc.) to describe the distribution and characteristics of the key values, draw charts (such as histograms, box plots, scatter plots, etc.) to show the distribution and relationship between the key values ​​and the final results, discover hidden patterns and relationships in the key values ​​through data mining techniques (such as association analysis, cluster analysis, etc.), use visualization tools (such as Tableau, Power BI, etc.) to assist the analysis process, improve the efficiency and accuracy of the analysis, and finally build a prediction model (decision tree) to predict future trends and results;

[0081] The intelligent reply module generates reply content based on the analysis results and sends it to the user via voice or text data. Specifically:

[0082] S1: Data preprocessing: convert text data into word vectors;

[0083] S2: Use pre-trained large language models (such as BERT, GPT, etc.) to extract features from text data; these models have been trained on large-scale corpora and can capture the semantic information and contextual relationships of the text. By inputting text data into the pre-trained model, vector representations of the text can be obtained. These vector representations contain the semantic features and contextual information of the text;

[0084] S3: LSTM network construction, including:

[0085] S3.1: Add a word embedding layer. Before the LSTM network, add a word embedding layer. The function of the word embedding layer is to map each word in the text into a high-dimensional vector space, so that similar words are closer in the vector space. The word embedding layer can be obtained through training;

[0086] S3.2: Build an LSTM backbone network. In the smart reply module, the LSTM network is used as the backbone network to process text sequence data and extract features. LSTM (Long Short-Term Memory) is a special recurrent neural network (RNN) that can capture long-term dependencies in sequence data. The LSTM network consists of multiple LSTM units. Each LSTM unit contains components such as input gate, forget gate, output gate, and cell state, and can process data at each time step in the input sequence.

[0087] S3.3: Construct a fully connected layer. After the LSTM backbone network, add one or more fully connected layers to further process and classify the features extracted by the LSTM network. The function of the fully connected layer is to map the high-dimensional features extracted by the LSTM network into a low-dimensional classification space, thereby obtaining the final classification result or reply content.

[0088] S4: Model training and optimization, including:

[0089] S4.1: Loss function selection. According to the specific task requirements (such as text classification, sentiment analysis, response generation, etc.), select the appropriate loss function to evaluate the performance of the model. For classification tasks, the cross entropy loss function is used; for generation tasks, the sequence generation loss function is used;

[0090] S4.2: Optimization algorithm selection, select SGD training model. The role of the optimization algorithm is to minimize the loss function by continuously adjusting the parameters of the model, thereby improving the performance of the model;

[0091] S4.3: Model evaluation and tuning: Use the validation set to evaluate the performance of the model and tune the model based on the evaluation results. The tuning process includes adjusting the model's hyperparameters (such as learning rate, batch size, etc.) and modifying the network structure.

[0092] S5: Reply generation. After the model training is completed, the trained model is used to generate the reply content. The text data input by the user is preprocessed and input into the model. The model generates the corresponding reply content based on the extracted features.

[0093] The user data analysis module deeply analyzes the interaction records and feedback between customers and intelligent customer service, fine-tunes the customer model, and uses data mining algorithms to achieve personalized recommendations based on user preferences. Specifically, it includes:

[0094] S1: Data collection: Collect the interaction records between customers and robots, including text chats, voice calls and other forms of communication, and organize customer feedback after using intelligent customer service, such as satisfaction surveys, feedback, etc.

[0095] S2: Data cleaning: Clean the collected interaction records, remove irrelevant information such as advertisements and spam, and standardize the feedback data to ensure data consistency and accuracy;

[0096] S3: Data analysis: Use natural language processing (NLP) technology to analyze the conversations between customers and intelligent customer service, extract key information, analyze customer feedback data, understand customer satisfaction and potential needs for intelligent customer service, and through data analysis, discover common problems, pain points and difficulties in the interaction between customers and intelligent customer service, and identify improvement suggestions and demand points mentioned by customers in feedback;

[0097] S4: Model evaluation: Based on the interaction records and feedback data between customers and intelligent customer service, evaluate the performance of the current intelligent customer service model and analyze the model's performance in identifying customer intentions, answering questions, and providing solutions;

[0098] S5: Problem location: Based on the evaluation results, locate the problems and deficiencies in the intelligent customer service model and identify the reasons why the model performs poorly or fails in specific scenarios;

[0099] S6: Model optimization: Based on the results of problem location, fine-tune the intelligent customer service model, optimize the model's algorithm, parameters, and feature engineering to improve the model's accuracy and efficiency.

[0100] S7: Testing and Verification: Test the fine-tuned model to ensure that the model's performance has been improved. Use new interaction records and feedback data to verify the model to ensure that the model can perform well in actual applications.

[0101] The user data analysis module also includes: user portrait construction, which is based on the interaction records and feedback data between customers and intelligent customer service to construct user portraits and extract key features such as user interests, preferences, and behavior patterns;

[0102] Application of data mining algorithms: Use data mining algorithms (such as collaborative filtering, association rule mining, etc.) to analyze user portraits, identify similarities between users, and discover potential patterns and trends in user preferences;

[0103] Personalized recommendation generation: Generate personalized recommendation content based on the results of data mining. The recommended content may include products, services, activities, etc. to meet the personalized needs of users.

Claims

1. A communication customer service system based on artificial intelligence, characterized in that: include: Data collection module, data analysis module, intelligent reply module and user data analysis module; The data collection module collects historical data and real-time interaction data of users, defines MapReduce tasks, counts key values, and outputs the final results in text data; The data analysis module is used to analyze the key value and final result of the data acquisition module, extract useful information, and obtain the analysis results; The intelligent reply module generates reply content based on the analysis results and sends it to the user via voice or text data; The user data analysis module deeply analyzes the interaction records and feedback between customers and intelligent customer service, and fine-tunes the customer model.

2. The artificial intelligence-based communication customer service system according to claim 1, characterized in that: The data acquisition module includes the following steps: S1: Collect users’ historical call records and real-time interaction data; S2: Preprocess the collected data, including denoising, format conversion and data cleaning; S3: Data partitioning: Divide a large data set into multiple small data sets; S4: Build MapReduce framework configuration; S5: Map function execution; S6: Execute Shuffle and Sort, including: S6.1: Data partitioning: According to the key value of the key-value pair, the intermediate results are divided into different partitions, and each partition corresponds to a Reduce task; S6.2: Data sorting: In each partition, sort the key-value pairs so that the Reduce function can efficiently process records with the same key value; S7: Reduce, including: S7.1: Data reading: The Reduce task reads the intermediate results of the corresponding partition from the distributed file system; S7.2: Reduce function execution: Iterate the key-value pairs in each partition and execute the Reduce function. The Reduce function merges or calculates the records with the same key value to generate the final output result. S7.3: Output result storage: The final result generated by the Reduce stage is stored in the specified output path. The output result is output and stored in text format.

3. The artificial intelligence-based communication customer service system according to claim 2, characterized in that: Step S4 specifically includes: S4.1: Cluster configuration: Configure the Hadoop distributed computing cluster; S4.2: Task configuration: Define MapReduce tasks and configure task parallelism and resource allocation based on data volume and processing requirements.

4. The artificial intelligence-based communication customer service system according to claim 2, characterized in that: Step S5 specifically includes: S5.1: Data reading: Read data blocks and distribute them to worker nodes in the cluster; S5.2: Function execution: The Map function is executed on each working node to process each record in the data block. The Map function converts the input record into a series of key-value pairs, which are output as intermediate results. S5.3: Intermediate result storage: The intermediate results generated in the Map stage are stored in the distributed file system, and the intermediate results are partitioned and sorted.

5. The artificial intelligence-based communication customer service system according to claim 1, characterized in that: The smart reply module includes the following steps: S1: Data preprocessing: convert text data into word vectors; S2: Use a pre-trained large language model to extract features from text data; S3: LSTM network construction, including: S3.1: Add a word embedding layer. Add a word embedding layer before the LSTM network. S3.2: Build an LSTM backbone network to process text sequence data and extract features; S3.3: Build a fully connected layer to map the high-dimensional features extracted by the LSTM network into a low-dimensional classification space to obtain the final classification result or reply content; S4: Model training and optimization; S5: Reply generation. After the model training is completed, the trained model is used to generate the reply content. The text data input by the user is preprocessed and input into the model. The model generates the corresponding reply content based on the extracted features.

6. The artificial intelligence-based communication customer service system according to claim 5, characterized in that: S4 includes the following steps: S4.1: Loss function selection; S4.2: Optimization algorithm selection, select SGD training model; S4.3: Model evaluation and tuning, use the validation set to evaluate the performance of the model and tune the model based on the evaluation results.

7. The artificial intelligence-based communication customer service system according to claim 1, characterized in that: The user data analysis module includes the following steps: S1: Data collection: Collect the interaction records between customers and robots, and organize customers’ feedback after using intelligent customer service; S2: Data cleaning: Clean the collected interaction records, remove irrelevant information, and standardize the feedback data; S3: Data analysis: Use natural language processing technology to analyze the conversation between customers and intelligent customer service, extract key information, and analyze customer feedback data; S4: Model evaluation: Based on the interaction records and feedback data between customers and intelligent customer service, the performance of the current intelligent customer service model is evaluated; S5: Problem location: Based on the evaluation results, locate the problems and deficiencies in the intelligent customer service model and identify the reasons why the model performs poorly or fails in specific scenarios; S6: Model optimization: Based on the results of problem location, fine-tune the intelligent customer service model and optimize the model’s algorithm, parameters, and feature engineering; S7: Testing and Validation: Test the fine-tuned model.

8. The artificial intelligence-based communication customer service system according to claim 1, characterized in that: The user data analysis module also includes: user portrait construction, which builds user portraits based on the interaction records and feedback data between customers and intelligent customer service; Application of data mining algorithms: Use data mining algorithms to analyze user portraits, identify similarities between users, and discover potential patterns and trends in user preferences; Personalized recommendation generation: Generate personalized recommendation content based on the results of data mining.

Images