Intelligent question and answer method and device, computer equipment and storage medium

Abstract

The invention belongs to the field of artificial intelligence, and relates to an intelligent question and answer method, which comprises the steps of receiving a consultation question, and performing intention recognition on the consultation question by using an intention recognition model to obtain an intention label; selecting a consultation parameter template corresponding to the intention label from a consultation parameter template library, and pushing the consultation parameter template to the client terminal; acquiring necessary consultation parameters on the consultation parameter template, processing the necessary consultation parameters by utilizing a retrieval module of the question and answer generation model, and retrieving from the technical document set to obtain reply fragments; and combining the consultation question and the reply fragment to obtain a context fragment, processing the context fragment by using a generation module of a question and answer generation model, generating a reply text corresponding to the error-reported consultation question, and feeding back the reply text to the customer. The invention further provides an intelligent question answering device, computer equipment and a storage medium. According to the method, the reply generation efficiency and accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to an intelligent question-answering method. Background Art

[0002] With the rapid development of information technology and the increasing variety of business types, there has been a surge in various consulting needs. When customers need to understand or consult a certain business, they usually ask questions directly to the customer service department, and the customer service staff of the customer service department will answer the questions raised by the customers; for example, taking non-auto insurance (purchasing insurance products unrelated to auto insurance) as an example, the customer service department receives a large number of error consultations about this type of business every day; however, as the number of consultations continues to rise, customer service staff gradually find it difficult to cope, resulting in a significant decrease in the efficiency of answering customer consultation questions;

[0003] At present, in order to improve the efficiency of answering questions raised by customers, standardized answer documents have been developed for various types of customer questions. When users encounter questions that need to be consulted, they can refer to the standardized answer documents to obtain answers, thereby reducing dependence on customer service personnel and improving the efficiency of answering questions.

[0004] However, in the above solution, customers still need to spend a lot of time to familiarize themselves with and search for standardized answer documents, and ultimately many customers choose to ask customer service staff questions, and there are still technical problems with low response efficiency. Summary of the invention

[0005] The purpose of the embodiments of the present application is to propose an intelligent question-answering method, apparatus, computer device and storage medium, the main purpose of which is to improve the efficiency of answering.

[0006] In order to solve the above technical problems, the present application embodiment provides an intelligent question-answering method, which adopts the following technical solutions:

[0007] Receive consulting questions raised by customers, use a pre-trained intent recognition model to perform intent recognition on the consulting questions, and obtain intent labels corresponding to the consulting questions;

[0008] Selecting a consulting parameter template corresponding to the intent tag from a preset consulting parameter template library, and pushing the consulting parameter template to the client terminal;

[0009] Obtaining the necessary consulting parameters filled in by the customer on the consulting parameter template, and searching from a preset technical document set using a retrieval module of a pre-trained question-answer generation model according to the necessary consulting parameters to obtain a reply fragment;

[0010] The consulting question and the answer segment are merged to obtain a context segment, the context segment is processed using the generation module of the pre-trained question-answer generation model, a reply text corresponding to the error consulting question is generated, and the reply text is fed back to the customer.

[0011] Furthermore, after receiving the consulting questions raised by the customer, the method also includes:

[0012] Performing form recognition on the consulting question to obtain the question form of the consulting question;

[0013] According to the question form, a conversion technology corresponding to the question form is used to convert the consulting question into a text form to obtain the consulting question in text form.

[0014] Furthermore, the use of a pre-trained intent recognition model to perform intent recognition on the consulting question to obtain an intent label corresponding to the consulting question includes:

[0015] The consulting question is converted into a consulting question vector by using a preset vector conversion technology through the input layer of the pre-trained intention recognition model;

[0016] Extracting key feature vectors of the consultation question vector through the feature extraction layer of the pre-trained intent recognition model;

[0017] Performing intent recognition on the key feature vector through the intent recognition layer of the pre-trained intent recognition model to obtain an intent vector;

[0018] The intent vector is matched with a predefined intent label through the output layer of the pre-trained intent recognition model, and the intent label of the consulting question is output.

[0019] Furthermore, before selecting the consulting parameter template corresponding to the intention tag from the preset consulting parameter template library, the method further includes:

[0020] Acquire a historical consultation question data set, perform a clustering operation on the historical consultation question data set, and obtain consultation question data subsets of multiple categories;

[0021] According to a predefined label set, assign a label to the subset of consulting question data of each category;

[0022] Create a consultation parameter template corresponding to each tag according to the tag;

[0023] All of the consulting parameter templates are stored in the preset consulting parameter template library.

[0024] Furthermore, selecting a consulting parameter template corresponding to the intention tag from a preset consulting parameter template library and pushing the consulting parameter template to the client terminal includes:

[0025] Traversing the preset consultation parameter template library, and querying from the preset consultation parameter template library whether there is a tag that is the same as the intended tag;

[0026] If it exists, the consulting parameter template corresponding to the tag is pushed to the client terminal;

[0027] If it does not exist, a consulting parameter template corresponding to the intention tag is created according to the intention tag and a preset information document in a predefined template format, and the consulting parameter template is pushed to the client terminal.

[0028] Furthermore, before obtaining the answer fragment by searching from a preset technical document set using a retrieval module of a pre-trained question-answer generation model according to the necessary consultation parameters, the method further includes:

[0029] Obtain a preset data set, and perform vectorization processing on each preset data in the preset data set to obtain a preset data vector set;

[0030] The preset technical document set is constructed according to the preset data vector set.

[0031] Further, according to the necessary consultation parameters, the retrieval module of the pre-trained question-answer generation model is used to search from a preset technical document set to obtain a reply fragment, including:

[0032] Vectorizing the necessary consulting parameters to obtain a necessary consulting parameter vector;

[0033] Calculating a similarity score between each preset data vector in the preset technical document set and the necessary consulting parameter vector;

[0034] The preset data corresponding to the highest similarity score is selected, and the preset data is used as the reply segment.

[0035] In order to solve the above technical problems, the embodiment of the present application further provides an intelligent question-answering device, which adopts the following technical solution:

[0036] A receiving module is used to receive consulting questions raised by customers, and use a pre-trained intent recognition model to perform intent recognition on the consulting questions to obtain intent labels corresponding to the consulting questions;

[0037] A push module, used to select a consulting parameter template corresponding to the intention tag from a preset consulting parameter template library, and push the consulting parameter template to the client terminal;

[0038] A retrieval module, used to obtain the necessary consulting parameters filled in by the customer on the consulting parameter template, and to retrieve from a preset technical document set according to the necessary consulting parameters using the retrieval module of the pre-trained question-answer generation model to obtain a reply fragment;

[0039] A generation module is used to merge the consulting question and the answer fragment to obtain a context fragment, use the generation module of the pre-trained question-answer generation model to process the context fragment, generate a reply text corresponding to the error consulting question, and feed back the reply text to the customer.

[0040] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0041] Receive consulting questions raised by customers, use a pre-trained intent recognition model to perform intent recognition on the consulting questions, and obtain intent labels corresponding to the consulting questions;

[0042] Selecting a consulting parameter template corresponding to the intent tag from a preset consulting parameter template library, and pushing the consulting parameter template to the client terminal;

[0043] Obtaining the necessary consulting parameters filled in by the customer on the consulting parameter template, and searching from a preset technical document set using a retrieval module of a pre-trained question-answer generation model according to the necessary consulting parameters to obtain a reply fragment;

[0044] The consulting question and the answer segment are merged to obtain a context segment, the context segment is processed using the generation module of the pre-trained question-answer generation model, a reply text corresponding to the error consulting question is generated, and the reply text is fed back to the customer.

[0045] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0046] Receive consulting questions raised by customers, use a pre-trained intent recognition model to perform intent recognition on the consulting questions, and obtain intent labels corresponding to the consulting questions;

[0047] Selecting a consulting parameter template corresponding to the intent tag from a preset consulting parameter template library, and pushing the consulting parameter template to the client terminal;

[0048] Obtaining the necessary consulting parameters filled in by the customer on the consulting parameter template, and searching from a preset technical document set using a retrieval module of a pre-trained question-answer generation model according to the necessary consulting parameters to obtain a reply fragment;

[0049] The consulting question and the answer segment are merged to obtain a context segment, the context segment is processed using the generation module of the pre-trained question-answer generation model, a reply text corresponding to the error consulting question is generated, and the reply text is fed back to the customer.

[0050] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0051] In this embodiment, a pre-trained intent recognition model is used to perform intent recognition on the consulting question to obtain the intent label corresponding to the consulting question, which can quickly locate the intent label of the user's consulting question, thereby improving the processing efficiency and accuracy of the subsequent generation of answers to the consulting question; and by automatically identifying the intent label corresponding to the consulting question, the intervention of manual customer service can be reduced, thereby reducing operating costs.

[0052] Selecting a consulting parameter template corresponding to the intent tag from a preset consulting parameter template library and pushing the consulting parameter template to the customer terminal can automatically respond to customer needs and reduce the workload of customer service personnel; and the feedback consulting parameter template can standardize the necessary consulting parameters filled in by the customer, improve the accuracy of the customer input, thereby improving the accuracy and efficiency of subsequent reply generation.

[0053] By obtaining the necessary consulting parameters filled in by the customer on the consulting parameter template, accurate retrieval can be performed based on the necessary parameters, thereby improving the accuracy of reply generation; based on the necessary consulting parameters, a retrieval module of a pre-trained question-answer generation model is used to search from a preset technical document set to obtain reply fragments, and reply fragments are generated by using a pre-trained model, which can improve the efficiency of subsequent reply generation and reduce labor costs.

[0054] By merging the consulting questions and answer fragments and analyzing them as the basis to generate a reply text, the model can more comprehensively understand the customer's questions and background information, thereby generating more accurate and relevant replies; through the pre-trained question and answer generation model, the efficiency of answer generation can be accelerated, manual intervention and waiting time can be reduced, and customer experience can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0057] Figure 2 A flowchart of an embodiment of an intelligent question-answering method according to the present application;

[0058] Figure 3 is a structural schematic diagram of an embodiment of an intelligent question-answering device according to the present application;

[0059] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0061] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0062] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0063] like Figure 1As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables.

[0064] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0065] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to a laptop computer 1011, a tablet computer 1012 or a mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0066] The server 103 may be a server that provides various services, such as a background server that provides support for a web page displayed on the terminal device 101 .

[0067] It should be noted that the intelligent question-answering method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the intelligent question-answering device is generally arranged in the server / terminal device.

[0068] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0069] Continue to refer Figure 2 , showing a flow chart of an embodiment of an intelligent question-answering method according to the present application.

[0070] In this embodiment, a consulting question raised by a customer is received, and the intent of the consulting question raised by the customer is identified through a pre-trained intent recognition model to obtain an intent label corresponding to the consulting question; based on the intent label, a consulting parameter template corresponding to the intent label is selected from a pre-classified consulting parameter template library, and the consulting parameter template is fed back to the customer terminal; after the customer receives the consulting parameter template at the customer terminal, the customer fills in the necessary consulting parameters in the consulting parameter template according to the prompt of the consulting parameter template; the necessary consulting parameters filled in by the customer are received, and the RAG technology is used to retrieve and generate an answer segment to the consulting question from the preset technical document; the consulting question and the answer segment are input into the pre-trained question-answering model together to obtain an answer text corresponding to the consulting question, and the answer text is fed back to the customer terminal.

[0071] The intelligent question-answering method comprises the following steps:

[0072] S201. Receive consulting questions raised by customers, use a pre-trained intent recognition model to perform intent recognition on the consulting questions, and obtain intent labels corresponding to the consulting questions.

[0073] In this embodiment, the consulting question refers to a problem that a customer encounters in business that cannot be solved or does not understand, such as an error problem, a query problem, etc.; the pre-trained intent recognition model refers to a neural network model, which includes at least a feature extraction layer and an intent recognition layer; the feature extraction layer is used to extract the key features of the consulting question raised by the user; the intent recognition layer is used to identify the key features to determine the intent label corresponding to the key features.

[0074] In this embodiment, consulting questions raised by customers through terminals are received, and the terminals include but are not limited to mobile terminals (mobile phones, iPads, etc.) and fixed terminals (desktop computers, etc.). Preprocessing operations are performed on the consulting questions, and the consulting questions after the preprocessing operation are input into a pre-trained intent recognition model. The pre-trained intent recognition model is used to perform intent recognition on the consulting questions to obtain intent labels corresponding to the consulting questions.

[0075] Specifically, the preprocessing operations on the consulting questions include but are not limited to text cleaning operations, removing irrelevant characters, punctuation marks, special symbols, etc., to reduce the noise of the consulting text and improve the text quality; character standardization operations, converting the format of the consulting text according to a pre-set format; word segmentation operations, segmenting the consulting text to obtain a word segmentation set.

[0076] In one embodiment, after receiving the consultation question raised by the customer, the method further includes:

[0077] Performing form recognition on the consulting question to obtain the question form of the consulting question;

[0078] According to the question form, a conversion technology corresponding to the question form is used to convert the consulting question into a text form to obtain the consulting question in text form.

[0079] In this embodiment, the consulting questions may exist in various forms, such as consulting questions in image form and consulting questions in voice form. In order to facilitate the subsequent identification of the intent label of the consulting questions, the form of the consulting questions is identified, and according to the form of the consulting questions, a conversion technology corresponding to the form is used to uniformly convert the consulting questions in different forms into text form to obtain consulting questions in text form; for example, when it is identified that the consulting question is in image form, image recognition technology is used, such as OCR technology (Optical Character Recognition) to detect and extract text information in the image to obtain consulting questions in text form; when it is identified that the consulting question is in voice form, ASR technology (Automatic Speech Recognition) is used to convert the consulting questions in voice form into consulting questions in text form.

[0080] In this embodiment, by converting consulting questions in different forms into consulting questions in text form, the efficiency of subsequent consulting question processing can be improved, and the accuracy of consulting question processing can be improved.

[0081] In one embodiment, the use of a pre-trained intent recognition model to perform intent recognition on the consulting question to obtain an intent label corresponding to the consulting question includes:

[0082] The consulting question is converted into a consulting question vector by using a preset vector conversion technology through the input layer of the pre-trained intention recognition model;

[0083] Extracting key feature vectors of the consultation question vector through the feature extraction layer of the pre-trained intent recognition model;

[0084] Performing intent recognition on the key feature vector through the intent recognition layer of the pre-trained intent recognition model to obtain an intent vector;

[0085] The intent vector is matched with a predefined intent label through the output layer of the pre-trained intent recognition model, and the intent label of the consulting question is output.

[0086] In this embodiment, it is described above that the pre-trained intention recognition model at least includes a feature extraction layer and an intention recognition layer. Taking the above description as an example, the pre-trained intention recognition model is composed of an input layer, a feature extraction layer, an intention recognition layer, and an output layer, wherein the feature extraction layer includes multiple convolutional layers, and the intention recognition layer includes a classifier; through the output layer of the intention recognition model, a preset word embedding technology (such as embedding technology, BERT technology, etc.) is used to convert each word of the consulting question into a vector, and then the vector of each word is combined to obtain a consulting question vector, wherein the form of combination includes but is not limited to splicing, etc.; the consulting question vector is convoluted through the feature extraction layer of the intention recognition model to extract the key feature vector of the consulting question vector; the key feature vector is processed through the intention recognition layer of the intention recognition model to extract the intention vector of the key feature vector; the output layer of the intention recognition model is matched with each label of the preset label library, and the similarity value between the intention vector and each label is calculated, and the label with the highest similarity value is used as the intention label of the consulting question, wherein the similarity value calculation method includes but is not limited to cosine similarity and Euclidean distance method.

[0087] In this embodiment, the intent of consulting questions is recognized through a pre-trained intent model, which can accurately identify the intent of the consulting questions and generate corresponding labels based on the intent, thereby avoiding misidentification of intent that leads to errors in subsequent replies, improving the accuracy of intent recognition, and further improving the accuracy of subsequent generated replies; and by identifying consulting questions through the model, the automation and efficiency of the processing flow are improved.

[0088] In an implementable example A, a consultation question is received from a customer named Xiao Ming in the chat box of the APP through his mobile phone: "My car was flooded. Can I apply for compensation?" The consultation question is preprocessed to obtain a clean consultation text "My car was flooded. Can I apply for compensation?" The preprocessed consultation question is input into a pre-trained intent recognition model, and each word is converted into a vector form through the embedding technology, and the vector splicing operation is performed to obtain a consultation question vector; the key feature vectors "car was flooded" and "apply for compensation" are extracted through the feature extraction layer of the intent recognition model; the intent recognition layer of the intent recognition model identifies the intent vectors of "car was flooded" and "apply for compensation", and the intent vectors are matched with each label in the preset label library to determine that the label of the consultation question is "auto insurance claims consultation (flooded car)".

[0089] In one embodiment, before performing intent recognition on the consulting question using the pre-trained intent recognition model, the method further includes:

[0090] Constructing an initial intent recognition model, and obtaining a consulting question dataset with label annotations, and dividing the consulting question dataset into a training set and a test set according to a preset ratio;

[0091] Using the training set to train the initial intent recognition model to obtain an intermediate intent recognition model;

[0092] Using the test set to test the intermediate intent recognition model, and obtaining the accuracy of the intermediate intent recognition model;

[0093] When the accuracy is greater than a preset accuracy threshold, a trained intent recognition model is obtained.

[0094] In this embodiment, a large number of labeled consulting questions are obtained, and a large number of labeled consulting questions are aggregated into a consulting question data set, and the consulting question data set is split into a training set and a test set according to a preset ratio; the initial intent recognition model is trained with the training set to obtain an intermediate intent recognition model; the intent recognition model is tested with the test set, and the obtained result is compared with the original label to obtain the accuracy of the intermediate intent recognition model, and it is determined whether the accuracy is greater than a preset accuracy threshold. When the accuracy is greater than the preset accuracy threshold, the training is completed to obtain a trained intent recognition model. When the accuracy is less than or equal to the preset accuracy threshold, the initial intent recognition model is continued to be trained with the training set until the accuracy of the intermediate intent recognition model is greater than the preset accuracy threshold, and the training is completed to obtain a trained intent recognition model.

[0095] In this embodiment, a pre-trained intent recognition model is used to perform intent recognition on the consulting question to obtain the intent label corresponding to the consulting question, which can quickly locate the intent label of the user's consulting question, thereby improving the processing efficiency and accuracy of the subsequent generation of answers to the consulting question; and by automatically identifying the intent label corresponding to the consulting question, the intervention of manual customer service can be reduced, thereby reducing operating costs.

[0096] S202: Select a consulting parameter template corresponding to the intention tag from a preset consulting parameter template library, and push the consulting parameter template to the client terminal.

[0097] In this embodiment, the preset consulting parameter template library refers to a storage space storing a plurality of consulting parameter templates; the consulting parameter template is a template used to prompt the user to input necessary consulting parameters.

[0098] In this embodiment, after the intent tag of the consulting question is obtained, a match is performed from a preset consulting parameter template library based on the intent tag to obtain a consulting parameter template corresponding to the intent tag, and the consulting parameter template is pushed to the customer terminal so that the customer can fill in the corresponding necessary consulting parameters according to the prompts of the consulting parameter template.

[0099] In one embodiment, before selecting the consulting parameter template corresponding to the intention tag from the preset consulting parameter template library, the method further includes:

[0100] Acquire a historical consultation question data set, perform a clustering operation on the historical consultation question data set, and obtain consultation question data subsets of multiple categories;

[0101] According to a predefined label set, assign a label to the subset of consulting question data of each category;

[0102] Create a consultation parameter template corresponding to each tag according to the tag;

[0103] All of the consulting parameter templates are stored in the preset consulting parameter template library.

[0104] In this embodiment, the historical consulting question data set refers to the consulting questions that users have asked to customer service personnel in the past; the predefined tag set is consistent with the intention tags corresponding to the consulting questions mentioned in step S201; the consulting parameter template refers to a predefined template, which includes a series of parameter filling boxes or placeholders, used to guide users to enter the correct necessary consulting parameters.

[0105] In this embodiment, before selecting the consulting parameter template corresponding to the intention label from the preset consulting parameter template library, a large number of historical consulting questions are obtained from various data channels (for example, obtaining customer service records of historical questions consulted by users to customer service personnel, user feedback, etc.), and all historical consulting questions are aggregated into a historical consulting question data set. The historical consulting question data set is clustered by a preset clustering algorithm to obtain consulting question subsets of multiple categories, wherein the preset consulting algorithm includes but is not limited to a K-means algorithm, a hierarchical clustering algorithm, etc., by text vectorizing each historical consulting question, calculating the similarity of any two historical consulting questions by a cosine similarity calculation method, a Euclidean distance calculation method, etc., traversing all historical consulting questions, and dividing the historical consulting question data set into historical consulting question subsets of multiple categories; assigning a label to each category according to a predefined label set, and the method of assigning labels also adopts a similarity calculation method; for each label, a corresponding consulting parameter template is created and all consulting parameter templates are stored in the preset consulting parameter template library.

[0106] In this embodiment, similar historical consulting questions are classified into the same category through clustering operations, so that the consulting parameter templates created for each category are more accurate and targeted, thereby improving the accuracy of subsequent generated replies; when new consulting questions arise, the corresponding consulting parameter templates are quickly found according to their intent labels, thereby improving the efficiency of subsequent generated replies.

[0107] Continuing with the above-mentioned feasible example A, when the intention label of the consultation question input by customer Xiao Ming is identified as "auto insurance claim consultation (flooded car)", the label "auto insurance claim consultation (flooded car)" is matched in the preset consultation parameter template library to obtain the consultation parameter template corresponding to the label "auto insurance claim consultation (flooded car)", such as: Do you buy motor vehicle loss insurance? Was the vehicle flooded while driving or while parked? Was the vehicle re-ignited after being flooded? And so on; the consultation parameter template is fed back to the terminal of customer Xiao Ming for Xiao Ming to fill in.

[0108] In one embodiment, selecting a consulting parameter template corresponding to the intention tag from a preset consulting parameter template library and pushing the consulting parameter template to the client terminal includes:

[0109] Traversing the preset consultation parameter template library, and querying from the preset consultation parameter template library whether there is a tag that is the same as the intended tag;

[0110] If it exists, the consulting parameter template corresponding to the tag is pushed to the client terminal;

[0111] If it does not exist, a consulting parameter template corresponding to the intention tag is created according to the intention tag and a preset information document in a predefined template format, and the consulting parameter template is pushed to the client terminal.

[0112] In this embodiment, the preset information document includes but is not limited to the car insurance claims system, etc.; when the consulting parameter template corresponding to the intent tag is selected from the preset consulting parameter template library through the intention tag, the entire preset consulting parameter template library is traversed according to the intent tag, and it is queried whether there is a tag identical to the intent tag in the preset consulting parameter template library. If so, the consulting parameter template corresponding to the tag is pushed to the user. If not, a consulting parameter template corresponding to the intent tag is created according to the intent tag and the preset information document in accordance with a predefined template format. For example, when there is no tag identical to the tag "car insurance claims consultation (flooded car)" in the preset consulting parameter template library, the parameters corresponding to "car insurance claims consultation (flooded car)" are obtained from the car insurance claims system, and a consulting parameter template with the tag "car insurance claims consultation (flooded car)" is created according to the predefined template format and the parameters corresponding to "car insurance claims consultation (flooded car)", and the consulting parameter template is pushed to the client terminal.

[0113] In this embodiment, by traversing the preset consultation parameter template library, the template that matches the intent tag can be quickly found, thereby improving the efficiency of user consultation; by pushing the consultation parameter template to the customer terminal, the customer can make customized replies based on the template content, avoiding entering irrelevant or less useful content, thereby improving the accuracy of subsequent replies; by automatically pushing templates corresponding to intent tags, the workload of customer service personnel is reduced and the efficiency of answering consultation questions is improved.

[0114] In this embodiment, a consulting parameter template corresponding to the intent tag is selected from a preset consulting parameter template library, and the consulting parameter template is pushed to the customer terminal, which can automatically respond to customer needs and reduce the workload of customer service personnel; and the consulting parameter template fed back can standardize the necessary consulting parameters filled in by the customer, improve the accuracy of the customer input, thereby improving the accuracy and efficiency of subsequent reply generation.

[0115] S203, obtaining the necessary consulting parameters filled in by the customer on the consulting parameter template, and searching from a preset technical document set based on the necessary consulting parameters using a retrieval module of a pre-trained question-answer generation model to obtain a reply fragment.

[0116] In this embodiment, the necessary consulting parameters refer to key parameters of the consulting question. For example, in the claim settlement process for a flooded vehicle, the necessary consulting parameters are the flooding situation of the vehicle and whether to restart the vehicle after the flooding.

[0117] After the customer fills in the necessary consulting parameters on the consulting parameter template, the consulting parameters filled in by the customer are obtained, and the necessary consulting parameters are processed using a pre-trained answer generation model. By means of similarity calculation, the preset technical document set is searched to obtain the answer fragment corresponding to the necessary consulting parameters.

[0118] Among them, the pre-trained response generation model adopts RAG technology. The RAG technology (Retrieval-Augmented Generation) is an artificial intelligence technology that combines information retrieval technology and language generation model. This technology retrieves relevant information from an external knowledge base and inputs the retrieved relevant information as prompts to large language models (LLMs), thereby assisting large language models to generate more accurate and richer text content.

[0119] The pre-trained answer generation model includes a retrieval module (Retriever) and a generation module (Generator). The retrieval module adopts a dual-tower model (Dual-Encoder), which consists of two independent encoders, one for encoding queries and the other for encoding documents. The two encoders map queries and documents into the same vector space for similarity calculation; the generation module refers to a large language model (LLMs, Large Language Models), which refers to a neural network model, a model for understanding and generating natural language text, including an input layer, an encoder, an attention mechanism, a decoder and an output layer.

[0120] In one embodiment, before searching from a preset technical document set using a retrieval module of a pre-trained question-answer generation model according to the necessary consultation parameters to obtain a reply fragment, the method further includes:

[0121] Obtain a preset data set, and perform vectorization processing on each preset data in the preset data set to obtain a preset data vector set;

[0122] The preset technical document set is constructed according to the preset data vector set.

[0123] In this embodiment, the preset data set refers to data related to consulting questions or a certain business collected from various sources such as documents, web pages, databases, etc., and can also be various data related to a preset question-and-answer system, for example, collecting all data related to auto insurance; performing text preprocessing operations on the collected preset data set, and the text preprocessing operations include data cleaning, noise removal, duplicate removal, and irrelevant information removal to obtain a preprocessed preset data set, and for each preset data in the preset data set, using embedded vector technology or a text embedding model (GLM) to convert each preset data into a vector, and aggregating all preset data vectors to obtain a preset data vector set; constructing a preset technical document based on the preset data vector set.

[0124] In this embodiment, by obtaining preset data from various sources, the accuracy and richness of subsequent response generation can be improved; by vectorizing the obtained data set, an efficient vector index can be constructed, thereby utilizing similarity calculations in the vector space to improve retrieval efficiency.

[0125] In another embodiment, according to the necessary consultation parameters, a retrieval module of a pre-trained question-answer generation model is used to search from a preset technical document set to obtain a reply fragment, including:

[0126] Vectorizing the necessary consulting parameters to obtain a necessary consulting parameter vector;

[0127] Calculating a similarity score between each preset data vector in the preset technical document set and the necessary consulting parameter vector;

[0128] The preset data corresponding to the highest similarity score is selected, and the preset data is used as the reply segment.

[0129] In this embodiment, the necessary consulting parameters are also vectorized by using embedded vector technology or text embedding model (GLM) to obtain the necessary consulting parameter vector, and the similarity scores between the necessary consulting parameter vector and each preset data vector in the preset technical document are calculated respectively by a similarity score calculation method, and all similarity scores are sorted in a high to low order, and the preset data corresponding to the preset data vector with the highest similarity score is selected, and the preset data is used as the reply fragment. The similarity score calculation method includes but is not limited to cosine similarity and Euclidean distance method.

[0130] In this embodiment, by calculating the similarity between the necessary consulting parameters and each preset data vector in the preset technical document set, the answer fragment most relevant to the necessary consulting parameters can be retrieved from the preset technical document set to improve the accuracy of subsequent answer generation.

[0131] Continuing with the above feasible example A, the necessary consulting parameters filled in by customer Xiao Ming on the consulting parameter template are obtained: the motor vehicle loss insurance has been purchased; the vehicle was flooded while parked; the vehicle was not re-ignited after being flooded; the above necessary consulting parameters are vectorized using the text embedding model, and the similarity score between the necessary consulting parameter vector and each preset data vector in the preset technical document set is calculated to obtain the response fragments "Motor vehicle loss insurance claim process", "Precautions for flooded vehicles", and "Compensation regulations for damage during parking".

[0132] In this embodiment, the necessary consulting parameters filled in by the customer on the consulting parameter template are obtained, and accurate retrieval can be performed based on the necessary parameters, thereby improving the accuracy of reply generation; based on the necessary consulting parameters, a retrieval module of a pre-trained question and answer generation model is used to search from a preset technical document set to obtain reply fragments, and reply fragments are generated by using a pre-trained model, which can improve the efficiency of subsequent reply generation and reduce labor costs.

[0133] S204: Merge the consulting question and the answer segment to obtain a context segment, process the context segment using the generation module of the pre-trained question-answer generation model, generate an answer text corresponding to the error consulting question, and feed back the answer text to the customer.

[0134] In this embodiment, the consulting question and the answer fragment are merged into a coherent text sequence to obtain a context fragment, and the context fragment is preprocessed, and the preprocessing includes word segmentation, stop word removal, etc. The generation module of the pre-trained question-answer generation model mentioned above includes an encoder and a decoder, and the encoder is used to process the preprocessed context fragment, and the context fragment is converted into a vector form. The context information and semantic information of the context fragment are captured through the self-attention mechanism to obtain a context vector, and the decoder is used to generate words or characters in the reply text one by one according to the context vector until the generation is completed to obtain the reply text, and the reply text is fed back to the client terminal.

[0135] Continuing with the above feasible example A, after obtaining the reply fragments "Claim process of motor vehicle loss insurance", "Precautions for flooded vehicles", and "Compensation regulations for damage during parking", Xiao Ming's original consultation question "My car was flooded. Can I apply for compensation?" is merged to obtain the context fragment, and the context fragment is processed by the large language model to generate the reply text "If your vehicle has purchased motor vehicle loss insurance, and the insurance terms include compensation for vehicle losses caused by natural disasters (such as flooding), then you are eligible to apply for compensation. The claim process for motor vehicle loss insurance usually includes the following steps: Reporting: After discovering that the vehicle is damaged, you should immediately contact the insurance company to report the case and provide vehicle information, accident details, and your contact information. On-site inspection: The insurance company will arrange for professionals to go to the accident site for inspection and assess the vehicle loss. Damage assessment: Based on the inspection results, the insurance company will determine the extent of the vehicle's loss and the amount of compensation. Claim application :You need to submit relevant claim application materials, such as insurance policy, ID card, driver's license, vehicle registration, accident certificate, etc. Compensation: Once the claim application is approved, the insurance company will pay you the compensation in accordance with the agreed compensation method and time. In the process of applying for claims, you need to pay special attention to the following points about the precautions for flooded cars: Make sure the vehicle is not restarted in the water, because the second start may cause engine damage, and some insurance companies may not compensate for such losses. Keep all evidence related to the accident, such as photos, videos, etc., for use in the claims process. Communicate with the insurance company in a timely manner to understand the progress of the claims and the materials that need to be supplemented. In addition, regarding the compensation provisions for damage during parking, if your vehicle suffers losses due to natural disasters (such as flooding) during parking, and your insurance terms include such compensation, then the insurance company will usually compensate according to the agreed compensation standards. However, please note that the specific compensation amount and conditions may vary depending on the insurance company and insurance terms."

[0136] In this embodiment, the consultation questions and answer fragments are merged and analyzed as a basis to generate a reply text, so that the model can more comprehensively understand the customer's questions and background information, thereby generating more accurate and relevant replies; through the pre-trained question and answer generation model, the efficiency of answer generation can be accelerated, manual intervention and waiting time can be reduced, and customer experience can be improved.

[0137] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0138] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0139] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0140] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0141] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of an intelligent question-answering device, which is similar to Figure 2 Corresponding to the method embodiment shown, the apparatus can be specifically applied to various computer devices.

[0142] like Figure 3 As shown, the intelligent question-answering device 300 described in this embodiment includes: a receiving module 301, a push module 302, a retrieval module 303, and a generation module 304. Among them:

[0143] The receiving module 301 is used to receive the consulting questions raised by the customers, and use the pre-trained intention recognition model to perform intention recognition on the consulting questions to obtain the intention labels corresponding to the consulting questions;

[0144] A push module 302, configured to select a consulting parameter template corresponding to the intention tag from a preset consulting parameter template library, and push the consulting parameter template to the client terminal;

[0145] A retrieval module 303 is used to obtain the necessary consulting parameters filled in by the customer on the consulting parameter template, and to retrieve from a preset technical document set according to the necessary consulting parameters using a retrieval module of a pre-trained question-answer generation model to obtain a reply fragment;

[0146] The generation module 304 is used to merge the consulting question and the answer fragment to obtain a context fragment, use the generation module of the pre-trained question-answer generation model to process the context fragment, generate a reply text corresponding to the error consulting question, and feed back the reply text to the customer.

[0147] In this embodiment, a pre-trained intent recognition model is used to perform intent recognition on the consulting question to obtain the intent label corresponding to the consulting question, which can quickly locate the intent label of the user's consulting question, thereby improving the processing efficiency and accuracy of the subsequent generation of answers to the consulting question; and by automatically identifying the intent label corresponding to the consulting question, the intervention of manual customer service can be reduced, thereby reducing operating costs.

[0148] Selecting a consulting parameter template corresponding to the intent tag from a preset consulting parameter template library and pushing the consulting parameter template to the customer terminal can automatically respond to customer needs and reduce the workload of customer service personnel; and the feedback consulting parameter template can standardize the necessary consulting parameters filled in by the customer, improve the accuracy of the customer input, thereby improving the accuracy and efficiency of subsequent reply generation.

[0149] By obtaining the necessary consulting parameters filled in by the customer on the consulting parameter template, accurate retrieval can be performed based on the necessary parameters, thereby improving the accuracy of reply generation; based on the necessary consulting parameters, a retrieval module of a pre-trained question-answer generation model is used to search from a preset technical document set to obtain reply fragments, and reply fragments are generated by using a pre-trained model, which can improve the efficiency of subsequent reply generation and reduce labor costs.

[0150] By merging the consulting questions and answer fragments and analyzing them as the basis to generate a reply text, the model can more comprehensively understand the customer's questions and background information, thereby generating more accurate and relevant replies; through the pre-trained question and answer generation model, the efficiency of answer generation can be accelerated, manual intervention and waiting time can be reduced, and customer experience can be improved.

[0151] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0152] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected and communicated through a system bus. It should be noted that the figure only shows a computer device 4 with a memory 41, a processor 42, and a network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0153] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0154] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the intelligent question-answering method, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0155] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the intelligent question-answering method.

[0156] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0157] In the implementation process of the electronic device of the present application, the pre-trained intention recognition model is used to perform intention recognition on the consulting question, and the intention label corresponding to the consulting question is obtained, which can quickly locate the intention label of the user's consulting question, improve the processing efficiency and the accuracy of the reply for the subsequent generation of the reply to the consulting question; and by automatically identifying the intention label corresponding to the consulting question, the intervention of manual customer service can be reduced, and the operating cost can be reduced. The consulting parameter template corresponding to the intention label is selected from the preset consulting parameter template library, and the consulting parameter template is pushed to the customer terminal, which can automatically respond to customer needs and reduce the workload of customer service personnel; and the feedback consulting parameter template can standardize the necessary consulting parameters filled in by the customer, improve the accuracy of the customer input, and thus improve the accuracy and efficiency of the subsequent reply generation. Obtain the necessary consulting parameters filled in by the customer on the consulting parameter template, and can accurately search according to the necessary parameters, thereby improving the accuracy of the reply generation; according to the necessary consulting parameters, the retrieval module of the pre-trained question and answer generation model is used to search from the preset technical document set to obtain the reply fragment, and the reply fragment is generated by the pre-trained model, which can improve the efficiency of the subsequent reply generation and reduce labor costs. By merging the consulting questions and answer fragments and analyzing them as the basis to generate a reply text, the model can more comprehensively understand the customer's questions and background information, thereby generating more accurate and relevant replies; through the pre-trained question and answer generation model, the efficiency of answer generation can be accelerated, manual intervention and waiting time can be reduced, and customer experience can be improved.

[0158] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the intelligent question-answering method as described above.

[0159] In the implementation process of the computer-readable storage medium of the present application, the pre-trained intention recognition model is used to perform intention recognition on the consulting question, and the intention label corresponding to the consulting question is obtained, which can quickly locate the intention label of the user's consulting question, improve the processing efficiency and the accuracy of the reply for the subsequent generation of the consultation question; and by automatically identifying the intention label corresponding to the consultation question, the intervention of manual customer service can be reduced, and the operating cost can be reduced. The consulting parameter template corresponding to the intention label is selected from the preset consulting parameter template library, and the consulting parameter template is pushed to the customer terminal, which can automatically respond to customer needs and reduce the workload of customer service personnel; and the feedback consulting parameter template can standardize the necessary consulting parameters filled in by the customer, improve the accuracy of customer input, and thus improve the accuracy and efficiency of subsequent reply generation. Obtain the necessary consulting parameters filled in by the customer on the consulting parameter template, and can accurately search according to the necessary parameters, thereby improving the accuracy of reply generation; according to the necessary consulting parameters, the retrieval module of the pre-trained question-answer generation model is used to search from the preset technical document set to obtain the reply fragment, and the reply fragment is generated by the pre-trained model, which can improve the efficiency of subsequent reply generation and reduce labor costs. By merging the consulting questions and answer fragments and analyzing them as the basis to generate a reply text, the model can more comprehensively understand the customer's questions and background information, thereby generating more accurate and relevant replies; through the pre-trained question and answer generation model, the efficiency of answer generation can be accelerated, manual intervention and waiting time can be reduced, and customer experience can be improved.

[0160] The non-Company software tools or components appearing in the embodiments of this application are merely examples and do not represent actual use.

[0161] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0162] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. An intelligent question-answering method, characterized in that: The steps include: Receive consulting questions raised by customers, use a pre-trained intent recognition model to perform intent recognition on the consulting questions, and obtain intent labels corresponding to the consulting questions; Selecting a consulting parameter template corresponding to the intent tag from a preset consulting parameter template library, and pushing the consulting parameter template to the client terminal; Obtaining the necessary consulting parameters filled in by the customer on the consulting parameter template, and searching from a preset technical document set using a retrieval module of a pre-trained question-answer generation model according to the necessary consulting parameters to obtain a reply fragment; The consulting question and the answer segment are merged to obtain a context segment, the context segment is processed using the generation module of the pre-trained question-answer generation model to generate an answer text corresponding to the consulting question, and the answer text is fed back to the customer.

2. The intelligent question-answering method according to claim 1, characterized in that: After receiving the inquiry question from the customer, the method further includes: Performing form recognition on the consulting question to obtain the question form of the consulting question; According to the question form, a conversion technology corresponding to the question form is used to convert the consulting question into a text form to obtain the consulting question in text form.

3. The intelligent question-answering method according to claim 1, characterized in that: The using of the pre-trained intent recognition model to perform intent recognition on the consulting question to obtain the intent label corresponding to the consulting question includes: The consulting question is converted into a consulting question vector by using a preset vector conversion technology through the input layer of the pre-trained intention recognition model; Extracting key feature vectors of the consulting question vector through the feature extraction layer of the pre-trained intent recognition model; Performing intent recognition on the key feature vector through the intent recognition layer of the pre-trained intent recognition model to obtain an intent vector; The intent vector is matched with a predefined intent label through the output layer of the pre-trained intent recognition model, and the intent label of the consulting question is output.

4. The intelligent question-answering method according to claim 1, characterized in that: Before selecting the consulting parameter template corresponding to the intention tag from the preset consulting parameter template library, the method further includes: Acquire a historical consultation question data set, perform a clustering operation on the historical consultation question data set, and obtain consultation question data subsets of multiple categories; According to a predefined label set, assign a label to the subset of consulting question data of each category; Create a consultation parameter template corresponding to each tag according to the tag; All of the consulting parameter templates are stored in the preset consulting parameter template library.

5. The intelligent question-answering method according to claim 1, characterized in that: The selecting a consulting parameter template corresponding to the intention tag from a preset consulting parameter template library and pushing the consulting parameter template to the client terminal includes: Traversing the preset consultation parameter template library, and querying from the preset consultation parameter template library whether there is a tag that is the same as the intended tag; If it exists, the consulting parameter template corresponding to the tag is pushed to the client terminal; If it does not exist, a consulting parameter template corresponding to the intention tag is created according to the intention tag and a preset information document in a predefined template format, and the consulting parameter template is pushed to the client terminal.

6. The intelligent question-answering method according to claim 1, characterized in that: Before searching from a preset technical document set using a retrieval module of a pre-trained question-answer generation model according to the necessary consultation parameters to obtain a reply fragment, the method further includes: Obtain a preset data set, and perform vectorization processing on each preset data in the preset data set to obtain a preset data vector set; The preset technical document set is constructed according to the preset data vector set.

7. The intelligent question-answering method according to claim 6, characterized in that: According to the necessary consultation parameters, the retrieval module of the pre-trained question-answer generation model is used to search from a preset technical document set to obtain answer fragments, including: Vectorizing the necessary consulting parameters to obtain a necessary consulting parameter vector; Calculating a similarity score between each preset data vector in the preset technical document set and the necessary consulting parameter vector; The preset data corresponding to the highest similarity score is selected, and the preset data is used as the reply segment.

8. An intelligent question-answering device, characterized in that: The device comprises: A receiving module is used to receive consulting questions raised by customers, and use a pre-trained intent recognition model to perform intent recognition on the consulting questions to obtain intent labels corresponding to the consulting questions; A push module, used to select a consulting parameter template corresponding to the intention tag from a preset consulting parameter template library, and push the consulting parameter template to the client terminal; A retrieval module, used to obtain the necessary consulting parameters filled in by the customer on the consulting parameter template, and to retrieve from a preset technical document set according to the necessary consulting parameters using the retrieval module of the pre-trained question-answer generation model to obtain a reply fragment; A generation module is used to merge the consulting question and the answer segment to obtain a context segment, use the generation module of the pre-trained question-answer generation model to process the context segment, generate a reply text corresponding to the consulting question, and feed back the reply text to the customer.

9. A computer device, characterized in that: The computer device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the intelligent question-answering method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the intelligent question-answering method according to any one of claims 1 to 7 is implemented.

Images