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

Abstract

According to the intelligent question and answer method and device, the storage medium and the computer equipment provided by the invention, during intelligent question and answer, the question input by the user and the context information can be firstly obtained, and the target text is obtained after the question is preprocessed; and then, keyword extraction and intention recognition can be performed on the target text, retrieval is performed in the knowledge base according to the extracted keywords, the recognized user intention and the context information, after retrieval results are obtained, the retrieval results can be summarized, the summarized text is optimized, and then a question and answer result is output. According to the process, the intention of the user can be determined through intention recognition, accurate retrieval can be carried out through keywords in the target text, context information and the intention of the user, and the accuracy of the question and answer result can be further improved by summarizing and optimizing the retrieval result; and high requirements of the fields of power dispatching and the like on professionality, accuracy and continuity of question and answer results are met.

Description

Technical Field

[0001] This application relates to the technical fields of artificial intelligence and natural language processing, and particularly to an intelligent question-answering method, device, storage medium, and computer device. Background Art

[0002] With the rapid development of artificial intelligence technology, intelligent question-answering systems have been widely used in various fields, especially in professional fields such as power dispatching. However, although existing general large models perform excellently in language understanding and generation, when faced with problems in professional fields, they often expose inaccurate answers, lack of basis, or even generate false answers (i.e., hallucination problems).

[0003] For example, when existing general large models process relevant problems in highly professional fields such as power dispatching, they often cannot deeply understand the user's questions, resulting in inaccurate replies to the questions. Moreover, existing intelligent question-answering systems cannot maintain coherence in continuous complex question-answering scenarios, leading to poor question-answering effects. Summary of the Invention

[0004] The purpose of this application aims to solve at least one of the above technical defects, especially the technical defect that when using general large models to process relevant problems in highly professional fields such as power dispatching, the output question-answering results are inaccurate and the question-answering effects are poor.

[0005] This application provides an intelligent question-answering method, and the method includes:

[0006] Obtain the question input by the user and the context information, and after preprocessing the question, obtain the target text;

[0007] Extract keywords and identify the intent from the target text, and retrieve in the knowledge base according to the extracted keywords, the identified user intent, and the context information to obtain the retrieval result;

[0008] Summarize the retrieval result, and after optimizing the summarized text, output the question-answering result.

[0009] Optionally, the obtaining of the context information includes:

[0010] According to the question input by the user, use a long short-term memory network to capture the context information in the current conversation.

[0011] Optionally, the obtaining of the target text after preprocessing the question includes:

[0012] After removing the stop words in the question, perform word segmentation on the question after removal, and perform lemmatization on the segmented question to obtain the target text.

[0013] Optionally, the keyword extraction from the target text includes:

[0014] Performing keyword extraction on the target text using natural language processing techniques.

[0015] Optionally, the intention recognition of the target text includes:

[0016] Performing intention recognition on the target text using a pre-configured target intention recognition model, and classifying the recognized user intention using a pre-trained classifier;

[0017] Among them, the target intention recognition model uses historical Q&A information in the field of power dispatching as training data, and fine-tunes a pre-trained large model of the Transformer architecture using the training data.

[0018] Optionally, the retrieving in the knowledge base according to the extracted keywords, the recognized user intention, and the context information to obtain a retrieval result includes:

[0019] Using a preset retrieval algorithm, and performing retrieval in the knowledge base according to the extracted keywords, the recognized user intention, and the context information to obtain a preliminary retrieval result;

[0020] Using a learning-to-rank model to rank the preliminary retrieval results, and filtering the ranked preliminary retrieval results using a filter to obtain the retrieval result.

[0021] Optionally, the summarizing the retrieval result, optimizing the summarized text, and outputting a Q&A result includes:

[0022] Using a pre-configured summary generation model to summarize the retrieval result, and performing semantic optimization on the summarized text to output the Q&A result.

[0023] This application also provides an intelligent Q&A device, including:

[0024] A text processing module, configured to obtain a question input by a user and context information, and perform preprocessing on the question to obtain a target text.

[0025] A knowledge base retrieval module, configured to perform keyword extraction and intention recognition on the target text, and perform retrieval in the knowledge base according to the extracted keywords, the recognized user intention, and the context information to obtain a retrieval result.

[0026] A result optimization module, configured to summarize the retrieval result, optimize the summarized text, and output the Q&A result.

[0027] The present application also provides a computer-readable storage medium, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the steps of the intelligent question-answering method according to any one of the above embodiments.

[0028] The present application also provides a computer device, including: one or more processors, and a memory;

[0029] The memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the intelligent question-answering method according to any one of the above embodiments are executed.

[0030] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages:

[0031] For the intelligent question-answering method, device, storage medium and computer device provided by the present application, when performing intelligent question-answering, the question and context information input by the user can be obtained first, and after preprocessing the question, a target text can be obtained; then, the present application can extract keywords and identify the intention of the target text, and retrieve in the knowledge base according to the extracted keywords, the identified user intention and the context information. After obtaining the retrieval result, the present application can also summarize the retrieval result and optimize the summarized text to output the question-answering result. This process can not only determine the user intention through intention recognition, but also perform accurate retrieval through the keywords, context information and user intention in the target text, and can further improve the accuracy of the question-answering result through the summary and optimization of the retrieval result, so as to meet the high requirements for the professionalism, accuracy and coherence of the question-answering result in fields such as power dispatching. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0033] Figure 1 It is a schematic flowchart of an intelligent question-answering method provided by an embodiment of the present application;

[0034] Figure 2 It is a schematic flowchart when performing retrieval in the knowledge base provided by an embodiment of the present application;

[0035] Figure 3 It is a schematic structural diagram of an intelligent question-answering device provided by an embodiment of the present application;

[0036] Figure 4 The figure is a schematic internal structure diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0037] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0038] In one embodiment, as Figure 1 shown, Figure 1 is a schematic flowchart of an intelligent question-answering method provided by an embodiment of the present application; the present application provides an intelligent question-answering method, and the method may include:

[0039] S110: Obtain the question and context information input by the user, and after preprocessing the question, obtain the target text.

[0040] In this step, when the intelligent question-answering system conducts intelligent question-answering, it can first obtain the question and context information input by the user, so that the obtained question can be preprocessed and then retrieved together with the context information, and the context memory function is used to ensure that the answers are more coherent when the user asks consecutive questions.

[0041] It can be understood that the intelligent question-answering system of the present application can interact with the user through the user interface, and the core of its design lies in convenience and diversity. The user can interact with the system in various ways, mainly including two forms: text input and voice input. Among them, the text input interface is that the user inputs the question through the text box, and the system further analyzes and processes the received text after receiving it. The text input interface of the present application has the characteristics of instant response and simple usability, and at the same time supports multi-round conversations in complex and professional fields. The voice input interface provides a more convenient interaction method for the user, especially more efficient in on-site environments or when quick problem-solving is required.

[0042] Furthermore, since there are certain differences in the usage habits, speaking styles, etc. of different users, in order to improve the accuracy of question retrieval in the present application, after receiving the question input by the user, the question can be preprocessed, and the preprocessing process includes but is not limited to optimizing some meaningless words, adjusting the word order of the text, etc., which can be set according to the actual situation and will not be limited here.

[0043] S120: Extract keywords and identify the intent from the target text, and retrieve in the knowledge base based on the extracted keywords, the identified user intent, and the context information to obtain a retrieval result.

[0044] In this step, after obtaining the user input question and context information through S110, this application can extract keywords and identify the intent from the preprocessed target text, so that it can retrieve in the knowledge base based on the extracted keywords, the identified user intent, and the context information, and then obtain a retrieval result.

[0045] Among them, the knowledge base of this application is a professional knowledge base for power dispatching, which covers data such as power grid operation, equipment information, and accident plans. This application integrates the large model with the professional knowledge base for power dispatching to build an accurate and efficient intelligent question-answering system, and through knowledge extraction, association analysis, and knowledge graph construction technologies, it models the structured and unstructured data in the field of power dispatching into knowledge, so as to provide professional knowledge support for the large model. At the same time, the system analyzes the user intent and context information through the large model, combines with the knowledge base for retrieval and summarization, and generates accurate and natural answers.

[0046] S130: Summarize the retrieval result, optimize the summarized text, and then output the question-answering result.

[0047] In this step, after retrieving in the knowledge base through S120 based on the extracted keywords, the identified user intent, and the context information to obtain a retrieval result, this application can also summarize the retrieval result, optimize the summarized text, and then output the question-answering result, which can effectively improve the accuracy of the question-answering result and enhance the user experience.

[0048] For example, in the application of the intelligent question-answering system in the field of power dispatching, users can input dispatching-related questions through text or voice, such as "How to handle power grid failures?". The system first uses the large model to analyze the user intent, combines with the professional knowledge base for power dispatching for retrieval. This knowledge base covers data such as power grid operation, equipment information, and accident plans. The system extracts keywords and classifies the intent to retrieve relevant document and plan information. Subsequently, the large model is used to summarize the retrieval result and generate a professional answer in natural language.

[0049] Furthermore, after outputting the question-answering result, this application can also dynamically optimize the question-answering effect through manual feedback and reinforcement learning, so as to continuously improve the answer quality of the system and user satisfaction.

[0050] Among them, this application can use DQN (Deep Q-Network) in reinforcement learning to optimize the question-answering strategy, and the specific formula is as follows:

[0051]

[0052] Among them, and are the current and next states, is the action, is the reward, is the discount factor.

[0053] In the above embodiment, when performing intelligent question answering, the question input by the user and the context information can be obtained first, and after preprocessing the question, the target text can be obtained; then, the present application can extract keywords and recognize the intention from the target text, and retrieve in the knowledge base according to the extracted keywords, the recognized user intention and the context information. After obtaining the retrieval result, the present application can also summarize the retrieval result and optimize the summarized text to output the question answering result. This process can not only determine the user intention through intention recognition, but also perform accurate retrieval through the keywords, context information and user intention in the target text, and can further improve the accuracy of the question answering result through the summary and optimization of the retrieval result, thereby meeting the high requirements for the professionalism, accuracy and coherence of the question answering result in fields such as power dispatching.

[0054] In one embodiment, obtaining the context information in S110 may include:

[0055] According to the question input by the user, use a long short-term memory network to capture the context information in the current conversation.

[0056] In this embodiment, when the intelligent question answering system obtains the context information, it can extract relevant information from the context record according to the question input by the user and use the relevant information as the context information. In addition, when performing knowledge base retrieval, the Lucene retrieval algorithm can also be used to accelerate the search.

[0057] Specifically, when the present application uses a long short-term memory network (LSTM) to capture the context information in the current conversation, it can model multiple rounds of conversations through the long short-term memory network (LSTM) to capture the context information. The specific modeling process is as follows:

[0058]

[0059] Among them, , , are the activation functions of the forget gate, input gate, and output gate respectively, is the Sigmoid activation function, is the weight matrix of each activation function, is the bias term, Denote the sequence of dense vectors of sentences in the t-th round of conversation, where t represents the conversation round. Denote the encoding result of sentences in the (t - 1)-th round of conversation.

[0060] Through the above modeling process, this application enables the system to support multi-round conversations and ensures more coherent answers when users ask consecutive questions through the context memory function.

[0061] In one embodiment, after preprocessing the question in S110, the target text can be obtained, which may include:

[0062] After removing the stop words in the question, segment the question after removing the stop words, and then lemmatize the segmented question to obtain the target text.

[0063] In this embodiment, since there are certain differences in the usage habits, speaking styles, etc. of different users, therefore, in order to improve the accuracy of question retrieval, this application can preprocess the question after receiving the question input by the user.

[0064] Specifically, when preprocessing the question input by the user, this application can first remove the stop words in the question, then segment the question after removing the stop words, and then lemmatize the segmented question to obtain the target text.

[0065] In a specific implementation manner, this application can perform the following preprocessing operations after text input:

[0066] a. Remove stop words: Stop words such as "de", "shi", "zai", etc., which have little meaning, are removed. Among them, the common stop word list of this application can be obtained through stopwords in the NLTK library;

[0067] b. Segment the text: Use the jieba library to divide the text into meaningful words;

[0068] c. Lemmatize: Restore the word to its basic form, such as running -> run, which is processed through the WordNetLemmatizer in NLTK.

[0069] Through the above preprocessing operations, the question input by the user in this application can be integrated into the target text, so that the intelligent question-answering system can retrieve the question-answering result more accurately.

[0070] In one embodiment, keyword extraction from the target text in S120 can include: using natural language processing technology to extract keywords from the target text.

[0071] In this embodiment, when extracting keywords from the target text, natural language processing technology can be used for keyword extraction to extract words or phrases that best represent its theme or content from the target text, thereby further improving the accuracy of retrieval results.

[0072] For example, this application can use the natural language processing technology TF-IDF for keyword extraction, and the specific formula is as follows:

[0073]

[0074] Among them, is the term, is the document, represents the document containing the specific term, is the total number of documents, is the document set, is used to calculate the number of documents containing the term of.

[0075] In one embodiment, the intention recognition of the target text in S120 may include:

[0076] Using a pre-configured target intention recognition model to perform intention recognition on the target text, and using a pre-trained classifier to classify the recognized user intention.

[0077] Among them, the target intention recognition model is obtained by using the historical Q&A information in the power dispatching field as training data and fine-tuning the pre-trained large model of the Transformer architecture with the training data.

[0078] In this embodiment, when performing intention recognition on the target text, a pre-trained large model of the Transformer architecture can be first selected as the initial intention recognition model, then the historical Q&A information in the power dispatching field is used as training data, and the target intention recognition model is obtained by fine-tuning it with the training data. Then, this application uses the target intention recognition model to perform intention recognition on the target text, and finally uses a pre-trained classifier to classify the recognized user intention, so as to determine the user intention corresponding to the target text.

[0079] Among them, when this application fine-tunes the initial intention recognition model, the pre-trained large model of the Transformer architecture can be used to parse the user's question, and its multi-head self-attention mechanism is as follows:

[0080]

[0081] Among them, respectively represent the query, key, and value matrices, For the dimension of the key.

[0082] Next, this application identifies the user's intent by fine-tuning the pre-trained model. The loss function for fine-tuning the model can adopt the following cross-entropy loss function:

[0083]

[0084] where, is the true label, is the probability distribution predicted by the model, represents the index of the current sample, represents the total number of samples.

[0085] Finally, this application can use the trained classifier for intent classification. For example, this application can use the support vector machine (SVM) to classify the extracted intent, and the SVM formula is as follows:

[0086]

[0087] where, is the weight vector, is the bias, is the label, is the input sample, represents the index of the current sample.

[0088] In one embodiment, as Figure 2 shown, Figure 2 is a schematic flowchart of the process when retrieving in the knowledge base provided by the embodiment of this application; in S120, retrieving in the knowledge base according to the extracted keywords, the identified user intent, and the context information to obtain a retrieval result may include:

[0089] S121: Use a preset retrieval algorithm and retrieve in the knowledge base according to the extracted keywords, the identified user intent, and the context information to obtain a preliminary retrieval result.

[0090] S122: Use a learning-to-rank model to rank the preliminary retrieval results, and use a filter to filter the ranked preliminary retrieval results to obtain a retrieval result.

[0091] In this embodiment, when retrieving in the knowledge base according to the extracted keywords, the identified user intent, and the context information, the system can use a preset retrieval algorithm for retrieval. During retrieval, it can perform targeted retrieval based on the extracted keywords, the identified user intent, and the context information. After obtaining the preliminary retrieval results, use a learning-to-rank model to rank the preliminary retrieval results, and use a filter to filter the ranked preliminary retrieval results, then the final retrieval result can be obtained.

[0092] In a specific implementation manner, the present application can use the Okapi BM25 text retrieval algorithm for retrieval. The specific algorithm is as follows:

[0093]

[0094] Wherein, is the word in the query, is the frequency of the word in the document, is the document length, is the average length of the document set, and are hyperparameters, represents the inverse document frequency, and is used to measure the importance of a certain term.

[0095]

[0096] Wherein, is the total number of documents, is the number of documents containing the term, The higher the value of, the less frequently the term appears in the document set, and thus the stronger the ability to distinguish documents.

[0097] Next, the present application can use the learning-to-rank model LambdaMART to rank the retrieval results:

[0098]

[0099] Wherein, represents the "increment" or "gradient" between sample and sample and is used to adjust the ranking model, and are the scores of two different results, is the Sigmoid activation function.

[0100] Finally, the present application can use rules based on domain knowledge to filter out irrelevant entries, such as the credibility of data sources, release time, etc., which can be specifically set according to the actual situation and are not limited herein.

[0101] In one embodiment, after summarizing the retrieval results in S130 and optimizing the summarized text, the Q&A results are output, which may include:

[0102] Summarize the retrieval results using a pre-configured summary generation model, and after semantic optimization of the summarized text, output the Q&A results.

[0103] In this embodiment, after obtaining the retrieval results, the present application can also summarize and optimize the retrieval results to output more accurate Q&A results.

[0104] Specifically, the present application can use a pre-configured summary generation model to summarize the retrieval results to form a natural language answer, and after semantic optimization of the summarized text, output the Q&A result.

[0105] For example, the present application can use a Seq2Seq model to generate a summary of the retrieval results. The Seq2Seq architecture uses an encoder-decoder structure, and the formula is as follows:

[0106]

[0107] Among them, is the hidden state, and are the encoding and decoding functions of the model, represents the input vector at the th time step in the input sequence, represents the hidden state at time step , represents the hidden state at time step .

[0108] Finally, the present application can use the BERT (Bidirectional Encoder Representations from Transformers) model for semantic optimization of sentences. The BERT model is a pre-trained language model based on the Transformer architecture, and this model can capture context information through the self-attention mechanism to optimize the generated answer.

[0109] Next, the intelligent Q&A device provided by the embodiments of the present application will be described. The intelligent Q&A device described below can be correspondingly referred to the intelligent Q&A method described above.

[0110] In one embodiment, as Figure 3 shown, Figure 3 is a schematic structural diagram of an intelligent Q&A device provided by an embodiment of the present application; the present application also provides an intelligent Q&A device, which can include a text processing module 210, a knowledge base retrieval module 220, and a result optimization module 230, specifically including the following:

[0111] The text processing module 210 is used to obtain the question input by the user and the context information, and after preprocessing the question, obtain the target text.

[0112] The knowledge base retrieval module 220 is configured to extract keywords and identify the intent from the target text, and perform a retrieval in the knowledge base based on the extracted keywords, the identified user intent, and the context information to obtain a retrieval result.

[0113] The result optimization module 230 is configured to summarize the retrieval result, and after optimizing the summarized text, output a question-and-answer result.

[0114] In the above embodiments, when performing intelligent question answering, the user input question and context information can be obtained first, and after preprocessing the question, a target text is obtained; then, the present application can extract keywords and identify the intent from the target text, and perform a retrieval in the knowledge base based on the extracted keywords, the identified user intent, and the context information. After obtaining the retrieval result, the present application can also summarize the retrieval result, and after optimizing the summarized text, output a question-and-answer result. This process can not only determine the user intent through intent recognition, but also perform accurate retrieval through the keywords, context information, and user intent in the target text, and can further improve the accuracy of the question-and-answer result by summarizing and optimizing the retrieval result, thereby meeting the high requirements for professionalism, accuracy, and coherence of the question-and-answer result in fields such as power dispatching.

[0115] In one embodiment, the present application further provides a computer-readable storage medium, characterized in that: computer-readable instructions are stored in the computer-readable storage medium, and when the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the steps of the intelligent question answering method according to any one of claims 1 to 7.

[0116] In one embodiment, the present application further provides a computer device, characterized in that it includes: one or more processors, and a memory.

[0117] Computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the one or more processors, the steps of the intelligent question answering method according to any one of claims 1 to 7 are executed.

[0118] Schematically, as Figure 4 shown, Figure 4 is an internal structure schematic diagram of a computer device provided by an embodiment of the present application. The computer device 300 can be provided as a server. Refer to Figure 4, the computer device 300 includes a processing component 302, which further includes one or more processors, and memory resources represented by a memory 301 for storing instructions executable by the processing component 302, such as application programs. The application programs stored in the memory 301 may include one or more modules each corresponding to a set of instructions. In addition, the processing component 302 is configured to execute instructions to perform the intelligent question-answering method of any of the above embodiments.

[0119] The computer device 300 may further include a power component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate based on an operating system stored in the memory 301, such as Windows Server TM, Mac OS XTM, Unix TM, Linux TM, Free BSDTM, or the like.

[0120] Those skilled in the art can understand that Figure 4 the structure shown in

[0121] is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0122] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The embodiments can be combined according to needs, and the same or similar parts can be referred to each other.

[0123] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent question-answering method, characterized in that, The method includes: Obtaining the question input by the user and the context information, and after preprocessing the question, obtaining a target text; Performing keyword extraction and intent recognition on the target text, and retrieving in a knowledge base according to the extracted keywords, the recognized user intent, and the context information to obtain a retrieval result; Summarizing the retrieval result, and after optimizing the summarized text, outputting a question-and-answer result.

2. The intelligent question-answering method according to claim 1, wherein The obtaining of the context information includes: According to the question input by the user, using a long short-term memory network to capture the context information in the current conversation.

3. The intelligent question and answer method according to claim 1, characterized in that The obtaining of the target text after preprocessing the question includes: Removing the stop words in the question, then performing word segmentation on the question after removal, and performing lemmatization on the segmented question to obtain the target text.

4. The intelligent question-answering method according to claim 1, wherein The performing of keyword extraction on the target text includes: Using natural language processing technology to perform keyword extraction on the target text.

5. The intelligent question-answering method according to claim 1, characterized in that The performing of intent recognition on the target text includes: Adopting a pre-configured target intent recognition model to perform intent recognition on the target text, and using a pre-trained classifier to classify the recognized user intent; Wherein, the target intent recognition model uses the historical question-and-answer information in the power dispatching field as training data, and fine-tunes a pre-trained large model of the Transformer architecture using the training data.

6. The intelligent question-answering method according to claim 1, characterized in that, The retrieving in the knowledge base according to the extracted keywords, the recognized user intent, and the context information to obtain a retrieval result includes: Using a preset retrieval algorithm, and retrieving in the knowledge base according to the extracted keywords, the recognized user intent, and the context information to obtain a preliminary retrieval result; Using a learning-to-rank model to rank the preliminary retrieval result, and using a filter to filter the ranked preliminary retrieval result to obtain the retrieval result.

7. The intelligent question-answering method according to claim 1, characterized in that The summarizing of the retrieval result, and after optimizing the summarized text, outputting a question-and-answer result includes: Using a pre-configured summary generation model to summarize the retrieval result, and after performing semantic optimization on the summarized text, outputting the question-and-answer result.

8. An intelligent question-answering device, characterized in that, It includes: A text processing module, configured to obtain the question input by the user and the context information, and after preprocessing the question, obtain a target text; A knowledge base retrieval module, configured to perform keyword extraction and intent recognition on the target text, and retrieve in the knowledge base according to the extracted keywords, the recognized user intent, and the context information to obtain a retrieval result; A result optimization module, configured to summarize the retrieval result, and after optimizing the summarized text, output a question-and-answer result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors are caused to execute the steps of the intelligent question-and-answer method according to any one of claims 1 to 7.

10. A computer device, characterized in that, It includes: One or more processors, and a memory; Computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the one or more processors, the steps of the intelligent question-answering method according to any one of claims 1 to 7 are executed.

Images