Session processing method and device, computer equipment and storage medium

Abstract

The invention relates to a session processing method and device, computer equipment and a storage medium. The method comprises the steps of obtaining a to-be-processed session, wherein the to-be-processed session is associated with a corresponding session intention; obtaining a corresponding historical session set according to the session intention, wherein the historical session set comprises a plurality of historical sessions; performing session processing on each historical session to obtain a plurality of session texts; and splicing the session texts to obtain a session standard workflow corresponding to the to-be-processed session. By adopting the method, the service quality of session processing can be improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular, to a session processing method, apparatus, computer device, and storage medium. Background Art

[0002] In the current online car-hailing industry, the communication between drivers and customer service is an important part of service quality. When facing a large number of driver problems, human customer service needs to solve the problems encountered by drivers during driving efficiently and accurately to ensure driver satisfaction and passenger safety.

[0003] However, different customer service personnel have different work processes when dealing with the same or similar problems, resulting in a lack of unified and standardized operation guidelines for customer service personnel when dealing with problems, which easily leads to inconsistent processing results and affects service quality. Summary of the Invention

[0004] Based on this, in view of the above technical problems, it is necessary to provide a session processing method, apparatus, computer device, and storage medium that can improve the service quality of session processing.

[0005] A session processing method, the method includes:

[0006] Obtain a session to be processed, and the session to be processed is associated with a corresponding session intent;

[0007] Obtain a corresponding historical session set according to the session intent, and the historical session set includes multiple historical sessions;

[0008] Perform session processing on each historical session to obtain multiple session texts;

[0009] Perform splicing according to each session text to obtain a session standard work process corresponding to the session to be processed.

[0010] In one embodiment, obtaining a session to be processed, where the session to be processed is associated with a corresponding session intent, includes:

[0011] Obtain at least one session interaction response to be processed corresponding to the session to be processed;

[0012] Perform intent analysis on each session interaction response to be processed to obtain the session intent corresponding to the session to be processed.

[0013] In one embodiment, the historical session includes at least one historical session interaction response, and performing session processing on each historical session to obtain multiple session texts includes:

[0014] Perform vector conversion on each historical session interaction response to obtain a corresponding interaction response vector;

[0015] Classify each interaction response vector to obtain the corresponding category;

[0016] Classify each historical session interaction response corresponding to each historical session according to the category to obtain the session text corresponding to each category.

[0017] In one embodiment, splice according to each session text to obtain the session standard workflow corresponding to the session to be processed, including:

[0018] Perform text processing on the session text corresponding to the same category to obtain the session standard text corresponding to the category;

[0019] Splice the session standard texts corresponding to each category to obtain the session standard workflow corresponding to the session to be processed.

[0020] In one embodiment, perform vector conversion on each historical session interaction response to obtain the corresponding interaction response vector, including:

[0021] Obtain a trained natural language processing model;

[0022] Input each historical session interaction response into the natural language processing model, and perform vector conversion on each historical session interaction response through the natural language processing model to obtain the corresponding interaction response vector.

[0023] In one embodiment, classify each interaction response vector to obtain the corresponding category, including:

[0024] Obtain a trained classification model;

[0025] Input each interaction response vector into the classification model, and classify each interaction response vector through the classification model to obtain the corresponding category.

[0026] In one embodiment, obtain the corresponding historical session set according to the session intention. After the historical session set includes multiple historical sessions, it includes: calculate the similarity between each historical session to obtain the similarity between each historical session;

[0027] Screen the historical session set according to the similarity to obtain the screened historical session set.

[0028] A session processing device, the above device includes:

[0029] A first acquisition module, configured to acquire a session to be processed, and the session to be processed is associated with a corresponding session intention;

[0030] A second acquisition module, configured to acquire the corresponding historical session set according to the session intention, and the historical session set includes multiple historical sessions;

[0031] A processing module for performing session processing on each historical session to obtain multiple session texts;

[0032] A splicing module for splicing according to each session text to obtain a session standard workflow corresponding to the session to be processed.

[0033] A computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented:

[0034] Obtain a session to be processed, and the session to be processed is associated with a corresponding session intention;

[0035] Obtain a corresponding historical session set according to the session intention, and the historical session set includes multiple historical sessions;

[0036] Perform session processing on each historical session to obtain multiple session texts;

[0037] Splice according to each session text to obtain a session standard workflow corresponding to the session to be processed.

[0038] A computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0039] Obtain a session to be processed, and the session to be processed is associated with a corresponding session intention;

[0040] Obtain a corresponding historical session set according to the session intention, and the historical session set includes multiple historical sessions;

[0041] Perform session processing on each historical session to obtain multiple session texts;

[0042] Splice according to each session text to obtain a session standard workflow corresponding to the session to be processed.

[0043] For the above session processing method, device, computer device, and storage medium, obtain a session to be processed, the session to be processed is associated with a corresponding session intention, obtain a corresponding historical session set according to the session intention, the historical session set includes multiple historical sessions, perform session processing on each historical session to obtain multiple session texts, and splice according to each session text to obtain a session standard workflow corresponding to the session to be processed. Therefore, through the session standard workflow, there is a reference standard for all driver problems to solve the corresponding problems, avoiding different processes for different customer service personnel to handle the same or similar problems, and improving the service quality of session processing. Moreover, the customer service personnel can solve problems by referring to the clear session standard workflow, which can reduce the processing time of the customer service and improve the processing efficiency of the customer service. Description of the Drawings

[0044] Figure 1 It is an application environment diagram of the session processing method in an embodiment;

[0045] Figure 2 It is a schematic flowchart of the session processing method in an embodiment;

[0046] Figure 3 It is a schematic flowchart of the processing steps for the session to be processed in an embodiment;

[0047] Figure 4 It is a schematic flowchart of the historical session processing steps in an embodiment;

[0048] Figure 5 It is a schematic flowchart of the splicing steps for the standard session workflow in an embodiment;

[0049] Figure 6 It is a schematic flowchart of the conversion steps for the historical session interaction response vector in an embodiment;

[0050] Figure 7 It is a schematic flowchart of the classification steps for the interaction response vector in an embodiment;

[0051] Figure 8 It is a schematic flowchart of the screening steps for the historical session set in an embodiment;

[0052] Figure 9 It is a structural block diagram of the session processing device in an embodiment;

[0053] Figure 10 It is an internal structure diagram of a computer device in an embodiment;

[0054] Figure 11 It is an internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0055] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0056] The session processing method provided by the present application can be applied to an application environment as Figure 1 shown. Among them, the in-vehicle terminal 102 communicates with the server 104 through a network. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers and portable wearable devices connected to the vehicle, and the server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0057] Specifically, the vehicle terminal 102 obtains the session to be processed and provides it to the server 104. The server 104 obtains the session to be processed. The session to be processed is associated with a corresponding session intent. According to the session intent, the corresponding historical session set is obtained. The historical session set includes multiple historical sessions. Session processing is performed on each historical session to obtain multiple session texts. According to the session texts, they are spliced to obtain the session standard workflow corresponding to the session to be processed.

[0058] In another embodiment, the vehicle terminal 102 obtains the session to be processed. The session to be processed is associated with a corresponding session intent. According to the session intent, the corresponding historical session set is obtained. The historical session set includes multiple historical sessions. Session processing is performed on each historical session to obtain multiple session texts. According to the session texts, they are spliced to obtain the session standard workflow corresponding to the session to be processed.

[0059] In one embodiment, as Figure 2 shown, a session processing method is provided. Taking the method applied to Figure 1 the in-vehicle terminal or the server as an example for illustration, it includes the following steps:

[0060] Step 202, obtain the session to be processed. The session to be processed is associated with a corresponding session intent.

[0061] Among them, the session to be processed is the session that needs to be processed. The session content can be a multi-round conversation between the driver and the customer service, which is a complete communication between the driver and the customer service. The session intent is the purpose or idea that the session hopes to achieve. Among them, when processing the session, a corresponding session intent label will be assigned to each session. It can be manually labeled by the customer service staff, or the session intent recognition is performed on the session to assign the corresponding session intent label.

[0062] Step 204, according to the session intent, obtain the corresponding historical session set. The historical session set includes multiple historical sessions.

[0063] Among them, since each session is associated with a corresponding session intent, after obtaining the session to be processed, the historical session set with the same session intent as the session to be processed is obtained. The historical session set includes multiple historical sessions, and each historical session has the same session intent as the session to be processed.

[0064] Specifically, each stored historical session is associated with a corresponding session intent. According to the session intent corresponding to the session to be processed, historical sessions with the same session intent are found from the stored historical sessions to form the historical session set.

[0065] For example, the session intent corresponding to the stored historical session 1 Figure 1 , the session intent corresponding to the historical session 2 Figure 2 , the session intent corresponding to the historical session 3Figure 1 and the session intention of the session to be processed is the session intention Figure 1 , then the historical sessions corresponding to the same session intention are grouped into a historical session set, and the historical session set includes historical session 1 and historical session 2.

[0066] Step 206: Perform session processing on each historical session to obtain multiple session texts.

[0067] Among them, the session processing includes session vector conversion and classification of the converted session vectors. Session vector conversion is to convert each historical session in the historical session set into a corresponding session vector, and then classify the converted session vectors to obtain the corresponding categories, and classify each historical session according to the categories to obtain the session texts corresponding to each category.

[0068] Among them, the category is the structure that composes the session standard work process. For example, category 1 is to confirm the problem, category 2 is to query the problem, category 3 is to perform an operation, and category 4 is to comfort words.

[0069] Step 208: Concatenate according to each session text to obtain the session standard work process corresponding to the session to be processed.

[0070] Among them, since each category is the structure that composes the session standard work process, therefore, after obtaining the session texts corresponding to each category, concatenate the session texts corresponding to each category to obtain the session standard work process corresponding to the session to be processed. The so-called session standard work process is the session SOP (Standard Operating Procedure), that is, the standard operating procedure, which means to describe the standard operation steps and requirements of a certain event in a unified format for guiding and standardizing daily work. The essence of SOP is to quantify the details. Generally speaking, SOP is to refine and quantify the key control points in a certain procedure.

[0071] It can be understood that the session standard work process refers to decomposing a certain work process into a series of standardized and operable steps so that the staff can execute according to the specified process to improve work efficiency and quality.

[0072] Specifically, after obtaining the session texts corresponding to each category, concatenate the session texts corresponding to each category to obtain the session standard work process corresponding to the session to be processed. Therefore, in subsequent applications, when encountering sessions with the same session intention, the corresponding session standard work process can be directly called, which avoids different processing situations for the same session intention, thereby improving work efficiency and quality.

[0073] In the above session processing method, a session to be processed is obtained. The session to be processed is associated with a corresponding session intent. A corresponding historical session set is obtained according to the session intent. The historical session set includes multiple historical sessions. Session processing is performed on each historical session to obtain multiple session texts. The session texts are concatenated to obtain the session standard workflow corresponding to the session to be processed. Therefore, through the session standard workflow, there is a reference standard for all driver problems to solve the corresponding problems, avoiding different processes for different customer service personnel to handle the same or similar problems, and improving the service quality of session processing. Moreover, the customer service personnel can solve problems by referring to the clear session standard workflow, which can reduce the processing time of the customer service and improve the processing efficiency of the customer service.

[0074] In one embodiment, as Figure 3 shown, obtaining a session to be processed, where the session to be processed is associated with a corresponding session intent, includes:

[0075] Step 302, obtaining at least one interaction response of the session to be processed corresponding to the session to be processed.

[0076] Step 304, performing intent analysis on each interaction response of the session to be processed to obtain the session intent corresponding to the session to be processed.

[0077] Among them, the session to be processed includes at least one interaction response of the session to be processed. The session to be processed is composed of interaction responses of the session to be processed. An interaction response defines a conversation between a driver and a customer service as an interaction response. An interaction response can be regarded as an immediate feedback action of the customer service to the driver's problem. For example,

[0078] Driver: Hello! Is anyone there?

[0079] Driver: I want to consult about the problem complained by the previous user.

[0080] Customer service: Okay, I'm here. Wait a moment. I'll check the relevant user complaint content.

[0081] Specifically, each interaction response of the session to be processed corresponding to the session to be processed is obtained, and intent analysis is performed on each interaction response of the session to be processed to analyze the purpose and idea of each interaction response, so as to obtain the session intent corresponding to the session to be processed. Among them, the intent analysis can be to analyze the text of each interaction response and determine the session intent corresponding to the session to be processed in combination with the analysis results.

[0082] In one embodiment, as Figure 4 shown, the historical session includes at least one historical session interaction response. Session processing is performed on each historical session to obtain multiple session texts, including:

[0083] Step 402: Perform vector transformation on each historical session interaction response to obtain the corresponding interaction response vector.

[0084] Step 404: Classify each interaction response vector to obtain the corresponding category.

[0085] Step 406: Classify each historical session interaction response corresponding to each historical session according to the category to obtain the session text corresponding to each category.

[0086] Among them, at least one historical session interaction response forms the corresponding historical session. When performing vector transformation on the historical session, vector transformation is performed on each historical session interaction response to obtain the corresponding interaction response vector. By vectorizing the historical session interaction response, non-numerical historical session interaction responses can be effectively processed and analyzed by a computer, capturing features and relationships, and improving computational efficiency.

[0087] Further, after vectorizing each historical session interaction response, classify the interaction response vector to obtain the corresponding category. By classifying the interaction response vector, it is helpful to quickly understand the purpose or meaning of each historical session interaction response.

[0088] Finally, after determining the category corresponding to each interaction response vector, classify the historical session interaction response corresponding to each historical session according to the category, and gather the historical session interaction responses of the same category together, so as to obtain the session text corresponding to each category.

[0089] For example, the session texts corresponding to category 1 confirmation questions are classified together, and the session texts corresponding to category 2 query questions are classified together, etc.

[0090] In one embodiment, as Figure 5 shown, splice according to each session text to obtain the session standard workflow corresponding to the session to be processed, including:

[0091] Step 502: Perform text processing on the session text corresponding to the same category to obtain the session standard text corresponding to the category.

[0092] Step 504: Splice the session standard texts corresponding to each category to obtain the session standard workflow corresponding to the session to be processed.

[0093] Among them, after obtaining the session text corresponding to each category, since each category is the structure that composes the session standard workflow, the session text corresponding to each category is spliced to obtain the corresponding session standard workflow.

[0094] Specifically, after obtaining the conversation texts corresponding to each category, it is necessary to process the conversation texts of each category, extract the content of the conversation texts of each category, and extract the text content that can represent the corresponding category to obtain the conversation standard text corresponding to the category. For example, the category is to confirm a question, so the driver's question is extracted from the corresponding conversation text as the conversation standard text corresponding to the category.

[0095] Finally, after obtaining the conversation standard texts corresponding to each category, the conversation standard texts corresponding to each category are concatenated to obtain the conversation standard workflow corresponding to the conversation to be processed. Further, the conversation intention corresponding to the conversation to be processed is bound to the conversation standard workflow. In this way, when encountering conversations with the same conversation intention subsequently, the corresponding conversation standard workflow can be directly called for use.

[0096] In one embodiment, as Figure 6 shown, vector conversion is performed on each historical conversation interaction response to obtain the corresponding interaction response vector, including:

[0097] Step 602, obtain a trained natural language processing model.

[0098] Step 604, input each historical conversation interaction response into the natural language processing model, and perform vector conversion on each historical conversation interaction response through the natural language processing model to obtain the corresponding interaction response vector.

[0099] Among them, the natural language processing model is a deep learning model constructed through large-scale pre-training and self-supervised learning techniques, mainly to improve the computer's ability to understand and generate natural language. For example, it can be a BERT model. The BERT model (Bidirectional Encoder Representations from Transformers) is a pre-trained natural language processing model proposed by Google AI in 2018. The pre-trained model of BERT can be fine-tuned in various downstream tasks, such as text classification, named entity recognition, sentiment analysis, etc., and achieved the state-of-the-art performance at that time. The core of BERT is an architecture based on Transformer, which contains multiple self-attention layers, enabling the model to effectively capture long-distance dependencies. The proposal of BERT marks a new stage in natural language processing, that is, to build a powerful language understanding model through large-scale pre-training and fine-tuning.

[0100] Therefore, the trained natural language processing model can convert non-numerical input data into corresponding vectors, which is a method that can convert discrete data into continuous, low-dimensional real number vectors. This conversion makes the original discrete data easier to process mathematically and computationally.

[0101] Specifically, obtain a pre-trained natural language processing model, use each historical conversation interaction response as the input of the natural language processing model, and through the natural language processing model, perform vector transformation on each historical conversation interaction response to convert it into continuous and low-dimensional vectors, and output the corresponding interaction response vectors.

[0102] In one embodiment, as Figure 7 shown, classify each interaction response vector to obtain the corresponding category, including:

[0103] Step 702, obtain a trained classification model.

[0104] Step 704, input each interaction response vector into the classification model, and through the classification model, classify each interaction response vector to obtain the corresponding category.

[0105] Among them, the classification model mainly solves classification, regression, and clustering problems. Classification belongs to supervised learning algorithms, which means learning based on existing data and labels (classification categories) to predict the labels of unknown data. The goal of the classification problem is to predict the class labels of the data, and the classification problem can be divided into binary classification and multi-classification problems. Here, the multi-classification problem is used. Through the classification model, the corresponding category of each interaction response vector can be obtained, and this category is an important structure that composes the standard work process of the conversation.

[0106] Among them, the classification model here can be an MLP model. MLP (Multilayer Perceptron) 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, and each connection has a corresponding weight. The nodes (except for the input nodes) also have an activation function. MLP is trained through the backpropagation 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.

[0107] Among them, MLP is applied 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 is as follows:

[0108] Input layer: Receive input data, and each node represents a feature in the data.

[0109] Hidden layer: The nodes in one or more hidden layers perform a weighted sum of the input data and process it through an activation function to produce a nonlinear output. Hidden layers can capture complex patterns in the data.

[0110] Output layer: The nodes in the output layer represent the final prediction results. For classification problems, each output node usually represents a category, and the softmax function is used to generate a probability distribution; for regression problems, there may be only one output node.

[0111] Specifically, a pre-trained classification model is obtained, each interaction response vector after vectorization is used as an input of the classification model, and each interaction response vector is classified by the classification model to obtain a category corresponding to each interaction response vector.

[0112] In one embodiment, Figure 8 As shown, the corresponding historical conversation set is obtained according to the conversation intent. The historical conversation set includes multiple historical conversations, including:

[0113] Step 802: Calculate the similarity of each historical conversation to obtain the similarity between each historical conversation.

[0114] Step 804: filter the historical conversation set according to the similarity to obtain a filtered historical conversation set.

[0115] After obtaining the corresponding historical conversation set according to the conversation intent, the historical conversation set needs to be cleaned, and abnormal historical conversations or invalid historical conversations need to be screened to obtain a cleaned historical conversation set. The screening of historical conversations can be screening the historical conversation interaction responses that constitute the historical conversations, and the screening can be kmeans clustering processing or similarity processing.

[0116] Among them, the K-means algorithm is an iterative clustering method based on distance, which aims to divide a given data set into K non-overlapping clusters so that the sum of the squared distances between each data point and the center point of its cluster is minimized.

[0117] The execution process of the algorithm is as follows: First, randomly select K data points as the initial cluster centers, or select better initial centers through some heuristic methods; then, enter the iterative process. For each data point in the dataset, calculate the distance between it and each cluster center, and assign it to the cluster represented by the nearest cluster center; after all data points are assigned, recalculate the center point of each cluster, that is, average the coordinates of all data points belonging to the cluster; then, repeat the above steps of assignment and center update until the stopping condition is met, such as the change in the cluster center is less than a preset threshold, or a predetermined number of iterations is reached. Due to its high computational efficiency and ease of understanding, the K-means algorithm is still widely used in data mining, image processing, and many other fields.

[0118] The similarity calculation is similar to the K-means algorithm. By calculating the similarity between multiple entities, it is determined whether they are similar. If the similarity deviates from the preset similarity, the corresponding historical session can be determined as an abnormal session.

[0119] Specifically, calculate the similarity of each historical session to obtain the similarity between each historical session, obtain the preset similarity, compare the calculated similarity with the preset similarity, and determine the historical session with a similarity less than the preset similarity as an abnormal session and filter it out.

[0120] Among them, a historical session consists of at least one historical session interaction response. Therefore, the center point can be determined according to each historical session interaction response, and then calculate the distance between each historical session interaction response and the center point. When the distance is greater than the threshold, it can be regarded as an abnormal interaction response and excluded. Finally, the filtered historical session interaction responses in this dimension are obtained.

[0121] It should be understood that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0122] In one embodiment, as Figure 9 shown, a session processing device 900 is provided, including: a first acquisition module 902, a second acquisition module 904, a processing module 906, and a splicing module 908, where:

[0123] The first acquisition module 902 is configured to acquire a session to be processed, and the session to be processed is associated with a corresponding session intention.

[0124] The second acquisition module 904 is configured to acquire a corresponding historical session set according to the session intention, and the historical session set includes multiple historical sessions.

[0125] The processing module 906 is configured to perform session processing on each historical session to obtain multiple session texts.

[0126] The splicing module 908 is configured to splice according to each session text to obtain a session standard workflow corresponding to the session to be processed.

[0127] In one embodiment, the first acquisition module 902 acquires at least one session interaction response to be processed corresponding to the session to be processed, performs intention analysis on each session interaction response to be processed, and obtains the session intention corresponding to the session to be processed.

[0128] In one embodiment, the historical session includes at least one historical session interaction response. The processing module 906 performs vector conversion on each historical session interaction response to obtain a corresponding interaction response vector, classifies each interaction response vector to obtain a corresponding category, and classifies each historical session interaction response corresponding to each historical session according to the category to obtain session texts corresponding to each category.

[0129] In one embodiment, the splicing module 908 performs text processing on the session texts corresponding to the same category to obtain a session standard text corresponding to the category, and splices the session standard texts corresponding to each category to obtain a session standard workflow corresponding to the session to be processed.

[0130] In one embodiment, the processing module 906 acquires a trained natural language processing model, inputs each historical session interaction response into the natural language processing model, and performs vector conversion on each historical session interaction response through the natural language processing model to obtain a corresponding interaction response vector.

[0131] In one embodiment, the processing module 906 acquires a trained classification model, inputs each interaction response vector into the classification model, and classifies each interaction response vector through the classification model to obtain a corresponding category.

[0132] In one embodiment, the session processing device 900 calculates the similarity between each historical session to obtain the similarity between each historical session, and filters the historical session set according to the similarity to obtain a filtered historical session set.

[0133] For the specific limitations of the session processing device, reference may be made to the limitations on the session processing method in the foregoing text, which will not be elaborated here. Each module in the foregoing session processing device may be implemented in whole or in part by software, hardware, or a combination thereof. The foregoing modules may be embedded in the processor of the computer device in hardware form or independent thereof, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the foregoing modules.

[0134] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as Figure 10 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the session standard workflow. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a session processing method.

[0135] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as Figure 11 shown. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a session processing method. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device may be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the computer device housing, or an external keyboard, touchpad, or mouse, etc.

[0136] Those skilled in the art can understand that Figure 10 or Figure 11 the structures shown in are merely block diagrams of some structures related to the solution of this application, and do not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0137] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: obtaining a session to be processed, where the session to be processed is associated with a corresponding session intent, obtaining a corresponding set of historical sessions according to the session intent, the set of historical sessions including multiple historical sessions, performing session processing on each historical session to obtain multiple session texts, and splicing the session texts according to each session text to obtain a session standard workflow corresponding to the session to be processed.

[0138] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0139] obtaining at least one session interaction response to be processed corresponding to the session to be processed;

[0140] performing intent analysis on each session interaction response to be processed to obtain a session intent corresponding to the session to be processed.

[0141] In one embodiment, the historical session includes at least one historical session interaction response. When the processor executes the computer program, the following steps are further implemented:

[0142] performing vector transformation on each historical session interaction response to obtain a corresponding interaction response vector;

[0143] classifying each interaction response vector to obtain a corresponding category;

[0144] classifying each historical session interaction response corresponding to each historical session according to the category to obtain session texts corresponding to each category.

[0145] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0146] performing text processing on the session texts corresponding to the same category to obtain a session standard text corresponding to the category;

[0147] splicing the session standard texts corresponding to each category to obtain a session standard workflow corresponding to the session to be processed.

[0148] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0149] obtaining a trained natural language processing model;

[0150] inputting each historical session interaction response into the natural language processing model, and performing vector transformation on each historical session interaction response through the natural language processing model to obtain a corresponding interaction response vector.

[0151] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0152] Obtain the trained classification model;

[0153] Input each interaction response vector into the classification model, and classify each interaction response vector through the classification model to obtain the corresponding category.

[0154] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0155] Calculate the similarity of each historical session to obtain the similarity between each historical session;

[0156] Filter the historical session set according to the similarity to obtain the filtered historical session set.

[0157] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: obtain the session to be processed, the session to be processed is associated with a corresponding session intention, obtain the corresponding historical session set according to the session intention, the historical session set includes multiple historical sessions, perform session processing on each historical session to obtain multiple session texts, and splice according to each session text to obtain the session standard workflow corresponding to the session to be processed.

[0158] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0159] Obtain at least one interaction response of the session to be processed corresponding to the session to be processed;

[0160] Perform intention analysis on each interaction response of the session to be processed to obtain the session intention corresponding to the session to be processed.

[0161] In one embodiment, the historical session includes at least one historical session interaction response. When the processor executes the computer program, the following steps are further implemented:

[0162] Perform vector conversion on each historical session interaction response to obtain the corresponding interaction response vector;

[0163] Classify each interaction response vector to obtain the corresponding category;

[0164] Classify each historical session interaction response corresponding to each historical session according to the category to obtain the session text corresponding to each category.

[0165] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0166] Perform text processing on the session text corresponding to the same category to obtain the session standard text corresponding to the category;

[0167] Concatenate the session standard texts corresponding to each category to obtain the session standard workflow corresponding to the session to be processed.

[0168] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0169] Obtain a trained natural language processing model;

[0170] Input each historical session interaction response into the natural language processing model, and perform vector conversion on each historical session interaction response through the natural language processing model to obtain the corresponding interaction response vector.

[0171] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0172] Obtain a trained classification model;

[0173] Input each interaction response vector into the classification model, and classify each interaction response vector through the classification model to obtain the corresponding category.

[0174] In one embodiment, when the processor executes the computer program, the following steps are further implemented:

[0175] Calculate the similarity between each historical session to obtain the similarity between each historical session.

[0176] Filter the historical session set according to the similarity to obtain the filtered historical session set.

[0177] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0178] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0179] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A session processing method, the method comprising: Acquire a session to be processed, wherein the session to be processed is associated with a corresponding session intent; Acquire a corresponding historical conversation set according to the conversation intention, where the historical conversation set includes a plurality of historical conversations; Performing session processing on each of the historical sessions to obtain a plurality of session texts; The conversation texts are spliced ​​together to obtain a conversation standard workflow corresponding to the conversation to be processed.

2. The method according to claim 1, characterized in that The acquiring of the to-be-processed session, wherein the to-be-processed session is associated with a corresponding session intention, includes: Obtaining at least one pending session interaction response corresponding to the pending session; Performing intent analysis on each of the to-be-processed session interaction responses to obtain a session intent corresponding to the to-be-processed session.

3. The method according to claim 1, characterized in that The historical conversation includes at least one historical conversation interaction response, and the conversation processing is performed on each of the historical conversations to obtain a plurality of conversation texts, including: Performing vector conversion on each of the historical conversation interaction responses to obtain a corresponding interaction response vector; Classifying each of the interactive response vectors to obtain a corresponding category; The historical conversation interaction responses corresponding to the historical conversations are classified according to the categories to obtain conversation texts corresponding to the categories.

4. The method according to claim 3, characterized in that The step of splicing the conversation texts to obtain a conversation standard workflow corresponding to the conversation to be processed includes: Performing text processing on the conversation text corresponding to the same category to obtain a conversation standard text corresponding to the category; The conversation standard texts corresponding to the categories are spliced ​​together to obtain the conversation standard workflow corresponding to the conversation to be processed.

5. The method according to claim 3, characterized in that: The performing vector conversion on each of the historical session interaction responses to obtain a corresponding interaction response vector includes: Get the trained natural language processing model; Each of the historical conversation interaction responses is input into the natural language processing model, and each of the historical conversation interaction responses is vector-converted by the natural language processing model to obtain a corresponding interaction response vector.

6. The method according to claim 3, characterized in that: The classifying each of the interaction response vectors to obtain a corresponding category includes: Get the trained classification model; Each of the interaction response vectors is input into the classification model, and each of the interaction response vectors is classified by the classification model to obtain a corresponding category.

7. The method according to claim 1, characterized in that The step of acquiring a corresponding historical conversation set according to the conversation intention, wherein the historical conversation set includes a plurality of historical conversations, includes: Calculating the similarity of each of the historical conversations to obtain the similarity between the historical conversations; The historical conversation set is filtered according to the similarity to obtain a filtered historical conversation set.

8. A conversation processing device, characterized in that: The device comprises: A first acquisition module is used to acquire a session to be processed, wherein the session to be processed is associated with a corresponding session intention; A second acquisition module, configured to acquire a corresponding historical conversation set according to the conversation intention, wherein the historical conversation set includes a plurality of historical conversations; A processing module, used for performing session processing on each of the historical sessions to obtain a plurality of session texts; The splicing module is used to perform splicing according to each of the conversation texts to obtain a conversation standard workflow corresponding to the conversation to be processed.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images