Intelligent customer service session classification method and device
By extracting and analyzing the conversation content between drivers and customer service personnel in the online ride-hailing industry, and automatically classifying using natural language models and neural networks, the problem of high efficiency and low manual classification is solved, and efficient conversation classification is achieved.
Patent Information
- Application Number
- CN202411099934.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-12-06
AI Technical Summary
In the existing online ride-hailing industry, customer service conversation classification relies on labor, resulting in high costs and low efficiency.
By obtaining the target session content between the driver and the customer service staff, extracting keyword characteristics, driver intention characteristics and customer service operation characteristics, and using natural language models and neural networks for analysis and training, and automatically performing conversation classification.
It reduces the cost of session classification, improves classification efficiency, and improves the accuracy and consistency of classification results.
Smart Images

Figure CN119088962A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of natural language processing, and in particular to an intelligent customer service conversation classification method and device. Background Art
[0002] In the existing online car-hailing industry, customer service conversation classification mainly relies on manual customer service. Customer service combines past inherent experience to classify conversations after the conversation ends. Based on the final classification of the conversation, the problems that customer service is currently dealing with can be analyzed later, so as to design optimization plans to improve customer service response efficiency. However, manual classification requires a lot of manpower. For customer service who performs classification, they need to learn the classification standards and hundreds of classification categories, which is costly and inefficient. Summary of the invention
[0003] In view of this, the present invention provides an intelligent customer service conversation classification method and device to solve the problem that manual classification of customer service conversations requires a large amount of manpower, resulting in high cost and low efficiency.
[0004] In a first aspect, the present invention provides an intelligent customer service conversation classification method, which obtains the target conversation content between the driver and the customer service personnel, and extracts the target keyword features, the target driver intention features and the target customer service operation features from the target conversation content, including:
[0005] Splice the target conversation content between the driver and the customer service staff according to the standard sentences;
[0006] The natural language model is used to analyze the spliced target conversation content. After deleting the target conversation content with erroneous logic, the target keyword features, target driver intention features and target customer service operation features are extracted.
[0007] In an optional implementation, before obtaining the target conversation content between the driver and the customer service personnel and extracting keyword features, driver intention features, and customer service operation features from the target conversation content, the method further includes:
[0008] Obtaining historical conversation content for training the first target network, and extracting feature training sets corresponding to historical keyword features, historical driver intention features, and historical customer service operation features from the historical conversation content;
[0009] Training the first target network using the feature training set and obtaining an output result of the first target network;
[0010] The loss value between the output result and the true value vector corresponding to the historical session content is calculated, and the network parameters of the first target network are updated, and the first target network is trained cyclically until the preset number of training times is reached to terminate the training, so as to obtain a trained first target network.
[0011] In an optional implementation, historical conversation content for training the first target network is obtained, and feature training sets corresponding to historical keyword features, historical driver intention features, and historical customer service operation features are extracted from the historical conversation content, including:
[0012] Get historical conversation content;
[0013] Splice the historical conversation content between the driver and the customer service staff according to standard sentences;
[0014] The natural language model is used to analyze the spliced historical conversation content. After deleting the target conversation content with erroneous logic, historical keyword features, historical driver intention features, and historical customer service operation features are extracted.
[0015] In an optional implementation, before inputting the target keyword vector, the target driver intention vector, and the target customer service operation vector into the second target network, the process further includes:
[0016] Obtain vector training sets corresponding to the target keyword vector, the target driver intention vector, and the target customer service operation vector for training the second target network.
[0017] In an optional implementation, obtaining vector training sets corresponding to the target keyword vector, the target driver intention vector, and the target customer service operation vector for training the second target network includes:
[0018] Obtaining a target keyword vector and its corresponding first conversation category, a target driver intention vector and its corresponding second conversation category, and a target customer service operation vector and its corresponding third conversation category;
[0019] The target keyword vector and its corresponding first conversation category, the target driver intention vector and its corresponding second conversation category, the target customer service operation vector and its corresponding third conversation category are respectively used as vector training sets corresponding to the target conversation content, among which the first conversation category is used as the conversation label category of the target keyword vector, the second conversation category is used as the conversation label category of the target driver intention vector, and the third conversation category is used as the conversation category of the target customer service operation vector.
[0020] The technical solution of the present invention has the following advantages:
[0021] The present invention discloses an intelligent customer service conversation classification method and device, wherein the method obtains the target conversation content between the driver and the customer service personnel, and extracts the target keyword features, the target driver intention features and the target customer service operation features from the target conversation content; the target keyword features, the target driver intention features and the target customer service operation features are input into a first target network to obtain the target keyword vector, the target driver intention vector and the target customer service operation vector; the target keyword vector, the target driver intention vector and the target customer service operation vector are input into a second target network to obtain the classification result of the target conversation content. The present invention can classify a large amount of target conversation content into categories, and can not only reduce the cost but also improve the classification efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0023] Figure 1 is a flow chart of an intelligent customer service conversation classification method according to an embodiment of the present invention;
[0024] Figure 2 is a schematic diagram of feature extraction of target conversation content according to an embodiment of the present invention;
[0025] Figure 3 is a schematic diagram of vector output of target conversation content according to an embodiment of the present invention;
[0026] Figure 4 is a structural block diagram of an intelligent customer service conversation classification device according to an embodiment of the present invention;
[0027] Figure 5 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0028] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.
[0029] According to an embodiment of the present invention, an embodiment of an intelligent customer service conversation classification method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0030] In this embodiment, a smart customer service conversation classification method is provided, which can be used in mobile terminals, such as mobile phones, tablet computers, etc. Figure 1 is a flow chart of a method for classifying intelligent customer service conversations according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:
[0031] Step S101, obtaining the target conversation content between the driver and the customer service personnel, and extracting the target keyword features, the target driver intention features and the target customer service operation features from the target conversation content.
[0032] Specifically, drivers are tenants of online ride-hailing companies. When passengers place online ride-hailing orders on the online ride-hailing platform, they can choose which tenants they want to order from. Different tenants have different pricing methods and standards. Customer service staff are the staff who respond to and handle complaints, reports, and inquiries provided by drivers. The target conversation content between the driver and the customer service staff is the text record of the driver's communication with the customer service staff when making complaints, reports, and inquiries on the APP. A complete communication record is considered a conversation.
[0033] Furthermore, the target keyword feature is the keyword in each target conversation content, the target driver intention feature is the driver intention in each target conversation content, and the feature is usually not more than 20 words. The target driver intention feature is the driver intention in each target conversation content, and the target customer service operation feature is the customer service response in each target conversation content, and the feature is usually not more than 20 words. In this embodiment, the open source API of Tongyi Qianwen can be used to call the qwen1.5-110b model to process the conversation and generate keyword features, driver intention features and customer service operation features in each target conversation. The Tongyi Qianwen mentioned above is an artificial intelligence language model. With its powerful natural language processing capabilities and extensive knowledge coverage, it plays an important role in education, consulting, information retrieval and other fields. As an artificial intelligence question-and-answer system, Tongyi Qianwen's main function is to understand and generate human natural language, and can provide accurate and detailed question-answering services, whether it is knowledge query in professional fields, answering questions in daily life, or interpretation and analysis of news and current affairs. Figure 2 As shown, it is a schematic diagram of using the qwen1.5-110b model to extract target keyword features, target driver intention features, and target customer service operation features from the target conversation content.
[0034] Step S102: input the target keyword feature, the target driver intention feature and the target customer service operation feature into the first target network to obtain a target keyword vector, a target driver intention vector and a target customer service operation vector.
[0035] Specifically, in the disclosed embodiment, the first target network is a BERT network model. BERT is short for Bidirectional Encoder Representations from Transformers, which is a pre-trained natural language processing model that can be fine-tuned in various downstream tasks, such as text classification, named entity recognition, sentiment analysis, etc., and achieved the most advanced performance at the time. The core of BERT is a Transformer-based architecture that contains multiple self-attention layers, which enables the model to effectively capture long-distance dependencies. Therefore, the target keyword features, target driver intention features, and target customer service operation features are input into the first target network, and an Embedding text conversion vector is performed on each feature, so that vectors corresponding to multiple different keywords can be obtained. Figure 3 As shown, it is a schematic diagram of using the BERT network model to train target keyword features, target driver intention features and target customer service operation features to obtain target keyword vectors, target driver intention vectors and target customer service operation vectors.
[0036] Step S103: input the target keyword vector, the target driver intention vector and the target customer service operation vector into the second target network to obtain the classification result of the target conversation content.
[0037] Specifically, the second target network in the embodiment of the present disclosure can be an MLP network. MLP is the abbreviation of MultilayerPerceptron, which is a feedforward artificial neural network. MLP consists of at least three layers of nodes: an input layer, one or more hidden layers, and an output layer. The nodes of each layer are connected to each node of the next layer, each connection has a corresponding weight, and the node (except the input node) also has an activation function. MLP is trained by a back propagation algorithm, which calculates the output error and propagates it back to the network to adjust the connection weights. This process is repeated until the performance of the network on the training data reaches a satisfactory level. MLP has applications in many machine learning tasks, including classification, regression, and feature learning. Although deep learning and convolutional neural networks (CNNs) are more common in tasks such as image and video analysis, MLP is still a powerful tool widely used in many other application fields. The basic structure of MLP includes: an input layer, a hidden layer, and an output layer, wherein the input layer receives input data, each node represents a feature in the data, and the hidden layer is a node in one or more hidden layers. It can perform weighted summation on the input data and process it through an activation function to produce a nonlinear output, and the hidden layer can capture complex patterns in the data. The output layer, whose nodes represent the final prediction results. For classification problems, usually each output node represents a category, and the softmax function is used to generate probability distribution; for regression problems, there can be one output node.
[0038] Exemplarily, qwen1.5-110b is used to extract features of the target conversation content between the driver and the customer service staff, and then the extracted feature information is converted into an Embedding text vector, thereby obtaining the target keyword vector, the target driver intention vector and the target customer service operation vector.
[0039] The intelligent customer service conversation classification method in the disclosed embodiment obtains the target conversation content between the driver and the customer service personnel, and extracts the target keyword features, target driver intention features and target customer service operation features from the target conversation content; inputs the target keyword features, target driver intention features and target customer service operation features into the first target network to obtain the target keyword vector, target driver intention vector and target customer service operation vector; inputs the target keyword vector, target driver intention vector and target customer service operation vector into the second target network to obtain the classification result of the target conversation content. The present invention can classify a large amount of target conversation content into categories, and can not only reduce costs but also improve classification efficiency.
[0040] In an optional implementation, the above step S101, obtaining the target conversation content between the driver and the customer service personnel, and extracting the target keyword features, the target driver intention features and the target customer service operation features from the target conversation content, includes:
[0041] Step a1, splicing the target conversation content between the driver and the customer service staff according to the standard sentence.
[0042] Step a2, using a natural language model to analyze the spliced target conversation content, deleting the target conversation content with erroneous logic, and then extracting target keyword features, target driver intention features, and target customer service operation features.
[0043] Specifically, executing the above steps a1 and a2 is equivalent to cleaning the data. In the database, the target conversation information of the conversation between the driver and the customer service is obtained, and the complete conversation is spliced into a complete conversation based on keywords as the basic corpus. For the spliced conversation, the open source API of Tongyi Qianwen is used to call the qwen1.5-110b model to process the conversation. Combined with the corpus meaning of the conversation, check whether the classification of each conversation conforms to the basic logic, and eliminate the conversation data with erroneous logic, and further obtain the target keyword features, target driver intention features and target customer service operation features.
[0044] In the disclosed embodiment, by executing the above steps a1 to a2, it is ultimately beneficial to accurately classify the target conversation content.
[0045] In an optional implementation, in step S101, before obtaining the target conversation content between the driver and the customer service personnel and extracting keyword features, driver intention features, and customer service operation features from the target conversation content, the method further includes:
[0046] Step b1, obtaining historical conversation content for training the first target network, and extracting feature training sets corresponding to historical keyword features, historical driver intention features, and historical customer service operation features from the historical conversation content.
[0047] Step b2: train the first target network using the feature training set and obtain the output result of the first target network.
[0048] Step b3, calculate the loss value between the output result and the true value vector corresponding to the historical session content, update the network parameters of the first target network, and train the first target network in a loop until the preset training times are reached to obtain a trained first target network.
[0049] Specifically, by executing the above steps b1 to b3, it is equivalent to using the historical conversation content to train the first target network. Based on the trained first target network, the BERT model is fine-tuned using the prepared conversation corpus. The BERT model is fine-tuned on the data of the target task through pre-training to adapt to the new task. This process involves updating the parameters of some or all layers of the first target network.
[0050] In the disclosed embodiment, the first target network is pre-trained to achieve pre-training of target keyword features, target driver intention features, and target customer service operation features, thereby obtaining a target keyword vector, a target driver intention vector, and a target customer service operation vector.
[0051] Furthermore, in a specific example, the above step b1, obtaining historical conversation content for training the first target network, and extracting feature training sets corresponding to historical keyword features, historical driver intention features, and historical customer service operation features from the historical conversation content, respectively, includes:
[0052] Step b11, obtaining historical conversation content.
[0053] Step b12, splicing the historical conversation contents between the driver and the customer service staff according to standard sentences.
[0054] Step b13, using a natural language model to analyze the concatenated historical conversation content, deleting the target conversation content with erroneous logic, and then extracting historical keyword features, historical driver intention features, and historical customer service operation features.
[0055] Here, the process of processing the historical conversation content is the same as the process of processing the target conversation content, and specifically refer to the above steps a1 to a2.
[0056] In an optional implementation, before inputting the target keyword vector, the target driver intention vector, and the target customer service operation vector into the second target network, the process further includes:
[0057] Step c1, obtaining vector training sets corresponding to the target keyword vector, the target driver intention vector and the target customer service operation vector for training the second target network.
[0058] In the disclosed embodiment, the vector training sets corresponding to the target keyword vector, the target driver intention vector, and the target customer service operation vector are used to implement pre-training of the second target network.
[0059] In an optional implementation, step c1, obtaining vector training sets corresponding to the target keyword vector, the target driver intention vector, and the target customer service operation vector for training the second target network, respectively, includes:
[0060] Step c11, obtaining a target keyword vector and its corresponding first conversation category, a target driver intention vector and its corresponding second conversation category, and a target customer service operation vector and its corresponding third conversation category.
[0061] In step c12, the target keyword vector and its corresponding first conversation category, the target driver intention vector and its corresponding second conversation category, the target customer service operation vector and its corresponding third conversation category are respectively used as vector training sets corresponding to the target conversation content, wherein the first conversation category is used as the conversation label category of the target keyword vector, the second conversation category is used as the conversation label category of the target driver intention vector, and the third conversation category is used as the conversation category of the target customer service operation vector.
[0062] In the disclosed embodiment, the second target network is an MLP network, that is, a fully connected neural network, in which the classification result of the target conversation content is obtained by using the label information corresponding to the target keyword vector, the target driver intention vector and the target customer service operation vector respectively.
[0063] Therefore, the intelligent customer service conversation classification method in the embodiment of the present invention uses a model to automatically classify customer service conversations, which can simplify the operation process, improve operation efficiency, and thus reduce operation costs. In addition, the model uses a unified classification standard to reduce the understanding deviation of the same category among different customer service personnel and improve the consistency of classification results. Automatic classification of customer service conversations by the model can improve the accuracy and completeness of the data, thereby providing a more reliable basis for data analysis and decision-making. In addition, through automatic classification, it can respond to customer needs more quickly and provide better customer service, thereby improving customer satisfaction and loyalty.
[0064] In this embodiment, an intelligent customer service conversation classification device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.
[0065] This embodiment provides an intelligent customer service conversation classification device, such as Figure 4 As shown, including:
[0066] The conversation feature acquisition module 41 is used to acquire the target conversation content between the driver and the customer service personnel, and extract the target keyword feature, the target driver intention feature and the target customer service operation feature from the target conversation content;
[0067] A conversation vector acquisition module 42, used to input the target keyword feature, the target driver intention feature and the target customer service operation feature into the first target network to obtain a target keyword vector, a target driver intention vector and a target customer service operation vector;
[0068] The conversation type classification module 43 is used to input the target keyword vector, the target driver intention vector and the target customer service operation vector into the second target network to obtain the classification result of the target conversation content.
[0069] In an optional implementation, the session feature acquisition module 41 includes:
[0070] The content splicing submodule is used to splice the target conversation content between the driver and the customer service staff according to the standard sentences;
[0071] The feature extraction submodule is used to analyze the spliced target conversation content using a natural language model, and after deleting the target conversation content with erroneous logic, extract the target keyword features, target driver intention features, and target customer service operation features.
[0072] In an optional implementation, the intelligent customer service conversation classification device further includes:
[0073] A feature acquisition module, used to acquire historical conversation content for training the first target network, and extract feature training sets corresponding to historical keyword features, historical driver intention features, and historical customer service operation features from the historical conversation content;
[0074] A result acquisition module, used to train the first target network using the feature training set and obtain the output result of the first target network;
[0075] The feature training module is used to calculate the loss value between the output result and the true value vector corresponding to the historical session content, update the network parameters of the first target network, and cyclically train the first target network until the preset training times are reached to terminate the training, so as to obtain a trained first target network.
[0076] In an optional implementation, the sample acquisition module includes:
[0077] A historical conversation acquisition unit, used to acquire historical conversation content;
[0078] A conversation content splicing unit, used to splice the historical conversation content between the driver and the customer service staff according to standard sentences;
[0079] The conversation feature extraction unit is used to analyze the spliced historical conversation content using a natural language model, and after deleting the target conversation content with erroneous logic, extract historical keyword features, historical driver intention features, and historical customer service operation features.
[0080] In an optional implementation, the intelligent customer service conversation classification device further includes:
[0081] The vector acquisition module is used to obtain vector training sets corresponding to the target keyword vector, the target driver intention vector and the target customer service operation vector for training the second target network.
[0082] In an optional implementation, the vector acquisition module includes:
[0083] A conversation category acquisition unit, used to acquire a target keyword vector and its corresponding first conversation category, a target driver intention vector and its corresponding second conversation category, and a target customer service operation vector and its corresponding third conversation category;
[0084] The vector training set determination unit is used to use the target keyword vector and its corresponding first conversation category, the target driver intention vector and its corresponding second conversation category, the target customer service operation vector and its corresponding third conversation category as the vector training sets corresponding to the target conversation content, respectively, wherein the first conversation category is used as the conversation label category of the target keyword vector, the second conversation category is used as the conversation label category of the target driver intention vector, and the third conversation category is used as the conversation category of the target customer service operation vector.
[0085] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.
[0086] The intelligent customer service conversation classification device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.
[0087] The embodiment of the present invention also provides a computer device having the above Figure 5 The intelligent customer service conversation classification device shown.
[0088] See also Figure 5 , Figure 5 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 5As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.
[0089] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.
[0090] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.
[0091] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0092] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.
[0093] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.
[0094] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.
[0095] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.
[0096] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. A method for classifying intelligent customer service conversations, characterized in that: The method comprises: Obtain target conversation content between the driver and the customer service personnel, and extract target keyword features, target driver intention features, and target customer service operation features from the target conversation content; Inputting the target keyword feature, the target driver intention feature, and the target customer service operation feature into a first target network to obtain a target keyword vector, a target driver intention vector, and a target customer service operation vector; The target keyword vector, the target driver intention vector, and the target customer service operation vector are input into a second target network to obtain a classification result of the target conversation content.
2. The method according to claim 1, characterized in that The step of obtaining the target conversation content between the driver and the customer service personnel, and extracting the target keyword features, the target driver intention features, and the target customer service operation features from the target conversation content includes: splicing the target conversation content between the driver and the customer service staff according to standard sentences; The target conversation content after splicing is analyzed using a natural language model, and after deleting the target conversation content with erroneous logic, target keyword features, target driver intention features, and target customer service operation features are extracted.
3. The method according to claim 1, characterized in that Before obtaining the target conversation content between the driver and the customer service personnel and extracting keyword features, driver intention features, and customer service operation features from the target conversation content, the method further includes: Acquire historical conversation content for training the first target network, and extract feature training sets corresponding to historical keyword features, historical driver intention features, and historical customer service operation features from the historical conversation content; Training the first target network using the feature training set, and obtaining an output result of the first target network; Calculate the loss value between the output result and the true value vector corresponding to the historical session content, update the network parameters of the first target network, and train the first target network in a loop until the training is terminated after a preset number of training times, so as to obtain the trained first target network.
4. The method according to claim 3, characterized in that The acquiring of historical conversation content for training the first target network and extracting feature training sets corresponding to historical keyword features, historical driver intention features, and historical customer service operation features from the historical conversation content respectively include: Get historical conversation content; splicing the historical conversation contents between the driver and the customer service personnel according to standard sentences; The concatenated historical conversation content is analyzed using a natural language model, and after deleting the target conversation content with erroneous logic, historical keyword features, historical driver intention features, and historical customer service operation features are extracted.
5. The method according to claim 1, characterized in that Before inputting the target keyword vector, the target driver intention vector, and the target customer service operation vector into the second target network, the following step further includes: Obtain vector training sets corresponding to the target keyword vector, the target driver intention vector, and the target customer service operation vector for training the second target network.
6. The method according to claim 5, characterized in that The step of obtaining vector training sets corresponding to the target keyword vector, the target driver intention vector, and the target customer service operation vector for training the second target network includes: Obtaining the target keyword vector and its corresponding first conversation category, the target driver intention vector and its corresponding second conversation category, and the target customer service operation vector and its corresponding third conversation category; The target keyword vector and its corresponding first conversation category, the target driver intention vector and its corresponding second conversation category, the target customer service operation vector and its corresponding third conversation category are respectively used as vector training sets corresponding to the target conversation content, wherein the first conversation category is used as the conversation label category of the target keyword vector, the second conversation category is used as the conversation label category of the target driver intention vector, and the third conversation category is used as the conversation category of the target customer service operation vector.
7. An intelligent customer service conversation classification device, characterized in that: The device comprises: A conversation feature acquisition module is used to acquire the target conversation content between the driver and the customer service personnel, and extract the target keyword feature, the target driver intention feature and the target customer service operation feature from the target conversation content; A conversation vector acquisition module, used to input the target keyword feature, the target driver intention feature and the target customer service operation feature into the first target network to obtain a target keyword vector, a target driver intention vector and a target customer service operation vector; The conversation type classification module is used to input the target keyword vector, the target driver intention vector and the target customer service operation vector into the second target network to obtain a classification result of the target conversation content.
8. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the intelligent customer service conversation classification method according to any one of claims 1 to 6 by executing the computer instructions.
9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the intelligent customer service conversation classification method according to any one of claims 1 to 6.
10. A computer program product, characterized in that It includes computer instructions, and the computer instructions are used to enable a computer to execute the intelligent customer service conversation classification method described in any one of claims 1 to 6.
