Intelligent question-answering system intention recognition method and device and storage medium

Abstract

The invention relates to an intelligent question answering system intention recognition method and device and a storage medium, and belongs to the technical field of knowledge base question answering. The method comprises the steps of receiving a query question input by a user; key words are extracted from the query question, and semantic features of the query question are obtained; converting the semantic features of the query problem into a vector form, and generating a query problem vector; inputting the query question vector into a question-answer pair database, and retrieving a question-answer pair matched with the query question based on the similarity between the query question vector and the vector of each question in the question-answer pair database; and taking the intention type of the retrieved question and answer pair as the intention type of the query question so as to generate an answer of the query question in combination with the intention type. The time delay in the intention recognition process is reduced, meanwhile, the uncertainty probability of intention recognition is reduced, and the performance of the whole system is improved.

Description

Technical Field

[0001] The present application belongs to the field of knowledge base question answering technology, and specifically relates to an intention recognition method, device and storage medium for an intelligent question answering system. Background Art

[0002] When the intelligent question-answering system identifies the intent of a user's query, the general method is to pre-construct a prompt word for intent identification, and then input the user's query question and the constructed prompt word into the large language model. Through the large language model's super logical understanding ability, the corresponding intent classification is given.

[0003] The advantage of this is that it effectively utilizes the capabilities of the large language model and has a relatively high accuracy rate, but the problems are also quite obvious: (1) The reasoning of the large language model takes a long time. An intent recognition may take several seconds or even longer. At the same time, due to the instability of the output of the large language model, the results given may not be what is preset. In this case, it is necessary to input the intent into the large language model multiple times, resulting in a poor user experience. (2) If each query question is identified through the large language model, the demand for computing power is relatively high, resulting in a significant reduction in the overall processing power of the system.

[0004] Therefore, it is very necessary to find an effective intent recognition method to effectively reduce the overall time consumption and uncertainty of the system and save the system's computing power. Summary of the invention

[0005] In view of the above analysis, the embodiments of the present invention aim to provide a method, device and storage medium for intent recognition in an intelligent question and answer system, so as to solve the problems of long overall system time consumption, high computing power requirements and poor user experience in the existing intent recognition technology.

[0006] In a first aspect of the present application, a method for identifying intentions in an intelligent question-answering system is provided, comprising:

[0007] Receive query questions input by users;

[0008] Extract keywords from the query question to obtain semantic features of the query question;

[0009] Converting the semantic features of the query question into a vector form to generate a query question vector;

[0010] Inputting the query question vector into a question-answer pair database, and retrieving a question-answer pair matching the query question based on the similarity between the query question vector and the vectors of each question in the question-answer pair database;

[0011] The intent type of the retrieved question-answer pair is used as the intent type of the query question, so as to generate an answer to the query question in combination with the intent type.

[0012] Optionally, the process of constructing the question-answer pair database includes:

[0013] Collecting a plurality of question-answer pair data, wherein the question-answer pair data includes a user's query question and a corresponding answer;

[0014] The intent type is predefined, and a classification model is used to mark a corresponding intent type for each question-answer pair; the classification model is trained using a training data set, and the training data set includes multiple question-answer pairs and intent type marking data corresponding to the question-answer pairs.

[0015] Optionally, the question-answer pair includes a question-answer pair generated by a knowledge document and a question-answer pair of a user's historical answer; the generation process of the question-answer pair generated by the knowledge document includes:

[0016] Divide the knowledge document into multiple slice data;

[0017] Inputting the slice data into a large language model, extracting and understanding the semantics of the slice data, generating questions and answers corresponding to the questions, and obtaining question-answer pairs generated by knowledge documents;

[0018] The generation process of the question-answer pair of the user's historical answers includes:

[0019] An instruction of selecting a user's historical answer is received, and a question-answer pair of the user's historical answer is generated from the selected user's historical answer.

[0020] Optionally, the question-answer pair database is a vector database, and the information stored in the vector database is vector information obtained by converting the questions in the question-answer pair from text form to vector form.

[0021] Optionally, retrieving a question-answer pair matching the query question based on the similarity between the query question vector and the vectors of each question in the question-answer pair database comprises:

[0022] Calculate the similarity score between the query question vector and the vectors of each question in the vector database, and retrieve the top k question-answer pairs with higher similarity scores;

[0023] Compare the similarity scores of the first k question-answer pairs with a preset threshold;

[0024] When at least one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, the intent type of the retrieved question-answer pair is used as the intent type of the query question;

[0025] When the similarity scores of the first k question-answer pairs are not greater than the preset threshold, the user query is input into the large language model for intent recognition.

[0026] Optionally, when at least one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, taking the intent type of the retrieved question-answer pair as the intent type of the query question includes:

[0027] When only one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, the intent type of the question-answer pair is used as the intent type of the query question;

[0028] When there are multiple similarity scores of the first k question-answer pairs that are greater than the preset threshold, the intent types of the first k question-answer pairs are deduplicated to obtain deduplicated intent types, and the first n intent types are taken as the intent types of the query question according to the similarity scores, where n≤k.

[0029] Optionally, before extracting keywords from the query question and obtaining the semantic features of the query question, the method further includes:

[0030] Modify the set vocabulary in the query question based on a preset vocabulary library, wherein the set vocabulary includes professional terms, abbreviations, and typos;

[0031] By analyzing the query language pattern, keyword completeness and context consistency, it is determined whether the query question is complete. If the query question is incomplete, a dialogue is used to request information to clarify the question from the user;

[0032] If the query question is complete, the step of extracting keywords from the query question and obtaining semantic features of the query question is performed.

[0033] Optionally, calculating the similarity score between the query question vector and the vectors of each question in the vector database includes:

[0034] The cosine similarity algorithm is used to calculate the similarity score between the query question vector and the vectors of each question in the vector database.

[0035] In a second aspect of the present application, an intelligent question and answer system intention recognition device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any of the above-mentioned intelligent question and answer system intention recognition methods is implemented.

[0036] According to a third aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, any of the above-mentioned methods for identifying intentions in an intelligent question-and-answer system is implemented.

[0037] The method for identifying intentions in an intelligent question-and-answer system provided in the present application receives a query question input by a user; extracts keywords from the query question to obtain semantic features of the query question; converts the semantic features of the query question into a vector form to generate a query question vector; inputs the query question vector into a question-and-answer pair database, and retrieves a question-and-answer pair that matches the query question based on the similarity between the query question vector and the vectors of each question in the question-and-answer pair database; uses the intent type of the retrieved question-and-answer pair as the intent type of the query question to generate an answer to the query question in combination with the intent type. While ensuring accuracy, the present application greatly reduces the delay in the intent identification process, and also reduces the probability of uncertainty in intent identification, thereby improving the performance of the entire system and ensuring user experience. In addition, the present application also provides an intelligent question-and-answer system intent identification device and storage medium having the above-mentioned technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this specification. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0039] Figure 1 A flowchart of a specific implementation method of the intelligent question-answering system intention recognition method provided in this application;

[0040] Figure 2 A schematic diagram of the process of retrieving question-answer pairs matching the query question for this application;

[0041] Figure 3 A flowchart of another specific implementation of the method for identifying intentions in an intelligent question-answering system provided in this application;

[0042] Figure 4 This is a structural block diagram of the intention recognition device of the intelligent question and answer system provided in this application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. It should be noted that, in the absence of conflict, the embodiments in the present disclosure and the features in the embodiments can be combined, separated, interchanged and / or rearranged with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0044] The terms used here are for the purpose of describing specific embodiments, and are not intended to be restrictive. As used here, unless the context clearly indicates otherwise, the singular forms "one (kind, person)" and "said (the)" are also intended to include plural forms. In addition, when the terms "comprise" and / or "include" and their variations are used in this specification, it is explained that there are stated features, integral bodies, steps, operations, parts, assemblies and / or their groups, but it is not excluded that there are or add one or more other features, integral bodies, steps, operations, parts, assemblies and / or their groups. It should also be noted that, as used here, the terms "substantially", "approximately" and other similar terms are used as approximate terms and not as degree terms, so that they are used to explain the inherent deviations of the measured values, calculated values ​​and / or the values ​​provided that will be recognized by those of ordinary skill in the art.

[0045] A flowchart of a specific implementation method of the intelligent question-answering system intention recognition method provided in this application is as follows Figure 1 As shown, the implementation specifically includes:

[0046] S101: receiving a query question input by a user.

[0047] Specifically, a query question input by a user may be received through a user interface. After receiving the input of the query question, basic text cleaning steps may be further performed, such as removing redundant spaces, punctuation marks, and the like.

[0048] S102: extracting keywords from the query question to obtain semantic features of the query question.

[0049] Extract keywords from the query questions entered by the user. This step usually uses natural language processing techniques, such as word segmentation, part-of-speech tagging, and named entity recognition, to obtain the semantic features of the query questions.

[0050] S103: Convert the semantic features of the query question into a vector form to generate a query question vector.

[0051] This step is used to convert text data into mathematical vector representations. These vectors capture the semantic information of the text, allowing machines to understand and process it. This conversion is very important for implementing subsequent similarity calculations and retrieval matching tasks.

[0052] It can be understood that text vectorization is the process of converting words, phrases, or entire documents into real vectors. These vectors are in a high-dimensional space, and each dimension represents a certain language feature or semantic feature.

[0053] S104: Input the query question vector into a question-answer pair database, and retrieve a question-answer pair that matches the query question based on the similarity between the query question vector and the vectors of each question in the question-answer pair database.

[0054] The query vector is input into the question-answer pair database, and the similarity between the query vector and each question vector in the database is compared using a vector similarity measurement method. Based on these similarity scores, the question-answer pair that best matches the user query can be retrieved. The implementation of this process requires the construction of a question-answer pair database in advance. It can be understood that the questions in each question-answer pair are converted into vector form and stored to facilitate the retrieval of questions matching the query in the question-answer pair database, thereby further determining the intent type of the query based on the intent type of the matching question.

[0055] The process of constructing a question-answer pair database includes: collecting a large amount of question-answer pair data, wherein the question-answer pair data includes the user's query questions and corresponding answers; pre-defining the intent type, and using a classification model to mark each question-answer pair with a corresponding intent type tag; the classification model is trained using a training data set, and the training data set includes multiple question-answer pairs and the corresponding intent type tag data of the question-answer pairs.

[0056] Intent type refers to the purpose of the user asking a question or the type of information they need to satisfy. Intent type helps the system understand and classify questions more accurately, thereby providing more precise answers.

[0057] The intent type labeling of question-answer pairs can be done by automatic labeling or manual labeling. Automatic labeling can be: training a classification model to automatically identify and label intent types. The model can be trained using a training data set that has already been labeled with intent types. Manual labeling can be done by having experts manually review and label the intent type of each question-answer pair before the establishment of an automated process or to improve labeling quality.

[0058] The generated question-answer pairs and their intent types are marked and saved in a question-answer pair database. The question-answer pair database can be specifically a vector database, which involves converting the questions in the question-answer pair from text to vector representation, which can be done through models such as TF-IDF, Word2Vec or BERT, and then storing these vectors for subsequent retrieval. This process ensures fast and efficient information retrieval, especially when facing large-scale data.

[0059] S105: Using the intent type of the retrieved question-answer pair as the intent type of the query question, so as to generate an answer to the query question in combination with the intent type.

[0060] The intent type in the retrieved question-answer pair is used as the intent type of the query question. Based on the intent type, a suitable answer can be further generated or selected. If the retrieved question-answer pair directly provides an answer, the answer can also be directly returned. Of course, the answer can also be further processed or personalized to better meet user needs.

[0061] This application obtains the intent type by calculating the similarity between the query question vector and the vectors of each question in the question-answer database, which greatly reduces the time for intent recognition and improves the response time of the entire system. The method of using vector similarity calculation requires less computing power than the method of using a large model for intent recognition, reduces the system's reasoning cost, and improves the processing power of the entire system. In addition, the use of a large model for intent recognition has the disadvantage of uncertainty. This application can reduce the probability of this uncertainty through the method of vector similarity calculation, improve the accuracy of the answer, and improve the overall user experience.

[0062] As a specific implementation, the question-answer pair database is a vector database, and the information stored in the vector database is the vector information corresponding to converting the questions in the question-answer pair from text form to vector form.

[0063] Reference Figure 2 A schematic diagram of a process of retrieving a question-answer pair that matches the query question, wherein based on the similarity between the query question vector and the vectors of each question in the question-answer pair database, retrieving a question-answer pair that matches the query question includes:

[0064] S201: Calculate the similarity scores between the query question vector and the vectors of each question in the vector database, and retrieve the top k question-answer pairs with higher similarity scores.

[0065] Specifically, a cosine similarity algorithm may be used to calculate the similarity score between the query question vector and the vectors of each question in the vector database.

[0066] S202: Compare the similarity scores of the first k question-answer pairs with a preset threshold.

[0067] The similarity scores of the retrieved question-answer pairs are compared with a preset threshold value. It is understandable that the preset threshold value can be set by performing multiple experiments based on the data characteristics of the project.

[0068] S203: When at least one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, the intent type of the retrieved question-answer pair is used as the intent type of the query question.

[0069] Among them, when only one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, the intent type of the question-answer pair is used as the intent type of the query question; when there are multiple similarity scores of the first k question-answer pairs that are greater than the preset threshold, the intent types of the first k question-answer pairs are deduplicated to obtain deduplicated intent types, and the first n intent types are taken as the intent types of the query question according to the similarity scores.

[0070] S204: When the similarity scores of the first k question-answer pairs are not greater than the preset threshold, the user query is input into the large language model for intent recognition.

[0071] If at least one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, the intent type of the retrieved question-answer pair is used as the intent type of the query question. Otherwise, the large model intent recognition process is followed to execute subsequent business processes.

[0072] In the case where multiple question-answer pairs exceed the threshold, the intent types of these question-answer pairs can be deduplicated first to ensure that each intent type is considered only once. After deduplication, the top n (n≤k) intent types with the highest scores are selected as the intent types of the user's query based on the similarity score. This step helps avoid bias caused by duplication of intent labels in the database.

[0073] One or more intent types will be output, which will guide the subsequent answer generation or information retrieval process, ensuring that users get responses that match their query intent.

[0074] This embodiment combines multiple potential matches to improve the accuracy and robustness of intent recognition, especially when facing ambiguous or ambiguous queries. At the same time, by setting thresholds and considering multiple candidates, various uncertainties can be handled more flexibly, improving overall user satisfaction.

[0075] It is understandable that the question-answer pair database can be a vector database. Specifically, the text form of the question in the question-answer pair can be converted into a vector form for storage. In this way, after obtaining the query question input by the current user and obtaining the query question vector, the query question vector can be directly calculated for similarity with each vector in the vector database to find a matching question-answer pair. A flowchart of another specific implementation method of the intelligent question-answer system intention recognition method provided in this application is shown in FIG. Figure 3 As shown, the implementation specifically includes:

[0076] S300: constructing a vector database offline.

[0077] The question-answer pair database stores multiple question-answer pairs, including question-answer pairs generated by knowledge documents and question-answer pairs of user historical answers.

[0078] The generation process of the question-answer pair generated by the knowledge document includes: dividing the knowledge document into multiple slice data; inputting the slice data into a large language model, extracting and understanding the semantics of the slice data, generating questions and answers corresponding to the questions, and obtaining the question-answer pair generated by the knowledge document.

[0079] Split the knowledge document into multiple slices for easier processing. These slices should be small enough to ensure that the model can understand its context, but large enough to contain complete information.

[0080] The process of generating the question-answer pair of the user's historical answers includes: receiving an instruction from the user to select the user's historical answers, and generating the question-answer pair of the user's historical answers from the selected user's historical answers. In this way, the question-answer pair with better user's historical answers can be manually screened out.

[0081] Each of the two types of question-answer pairs has an intent type tag, and the question-answer pairs and the corresponding intent tags are saved in a question-answer pair database.

[0082] S301: receiving a query question input by a user.

[0083] S302: Pre-processing the query question. The pre-processing may include but is not limited to: rewriting the query question, clarifying the question, and other operations.

[0084] The pre-operation of question rewriting specifically includes: automatically modifying the original query of the user to improve the expression of the query or increase its clarity, including but not limited to changing the word order, replacing synonyms, etc. For example, the set word list in the query question can be modified based on a preset word list library, and the set word list includes professional terms, abbreviations, and typos.

[0085] The pre-operation for question clarification specifically includes: determining whether the query is complete by analyzing the query language pattern, keyword integrity and context consistency, and if the query is incomplete, requesting information to clarify the query from the user in a dialogue form. If the query is complete, executing the step of extracting keywords from the query to obtain the semantic features of the query.

[0086] S303: extracting keywords from the query question to obtain semantic features of the query question.

[0087] S304: Convert the semantic features of the query question into a vector form to generate a query question vector.

[0088] Specifically, a vectorization module is called to perform vectorization processing on the query question.

[0089] S305: Input the query question vector into the question-answer pair database, and retrieve the first k question-answer pairs that match the query question based on the similarity between the query question vector and the vectors of each question in the question-answer pair database.

[0090] Use the generated query question vector to retrieve the top k question-answer pairs with the highest similarity from the constructed vector database. This step involves calculating the similarity between vectors and selecting the question-answer pair that is closest to the semantics of the user's query question.

[0091] S306: Using the intent type of the retrieved question-answer pair as the intent type of the query question, so as to generate an answer to the query question in combination with the intent type.

[0092] In addition, the present application also provides an intelligent question-answering system intention recognition device, such as Figure 4 As shown in the structural block diagram of the intelligent question and answer system intention recognition device provided in the present application, it includes a memory 41 and a processor 42, and the memory 41 stores a computer program. When the computer program is executed by the processor 42, any one of the above-mentioned intelligent question and answer system intention recognition methods is implemented.

[0093] In addition, the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements any of the above-mentioned methods for identifying intentions in the intelligent question and answer system.

[0094] Computer readable storage media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0095] The professionals should also be further aware that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0096] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0097] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only the specific implementation method of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for identifying intentions in an intelligent question-answering system, characterized in that: include: Receive query questions input by users; Extract keywords from the query question to obtain semantic features of the query question; Converting the semantic features of the query question into a vector form to generate a query question vector; Inputting the query question vector into a question-answer pair database, and retrieving a question-answer pair matching the query question based on the similarity between the query question vector and the vectors of each question in the question-answer pair database; The intent type of the retrieved question-answer pair is used as the intent type of the query question, so as to generate an answer to the query question in combination with the intent type.

2. The method for identifying intentions in an intelligent question-answering system according to claim 1, characterized in that: The process of constructing the question-answer pair database includes: Collecting a plurality of question-answer pair data, wherein the question-answer pair data includes a user's query question and a corresponding answer; The intent type is predefined, and a classification model is used to mark a corresponding intent type for each question-answer pair; the classification model is trained using a training data set, and the training data set includes multiple question-answer pairs and intent type marking data corresponding to the question-answer pairs.

3. The method for identifying intentions in an intelligent question-answering system according to claim 2, characterized in that: The question-answer pairs include question-answer pairs generated by knowledge documents and question-answer pairs of user historical answers; The generation process of the question-answer pair generated by the knowledge document includes: Divide the knowledge document into multiple slice data; Inputting the slice data into a large language model, extracting and understanding the semantics of the slice data, generating questions and answers corresponding to the questions, and obtaining question-answer pairs generated by knowledge documents; The generation process of the question-answer pair of the user's historical answers includes: An instruction of selecting a user's historical answer is received, and a question-answer pair of the user's historical answer is generated from the selected user's historical answer.

4. The method for identifying intentions in an intelligent question-answering system according to any one of claims 1 to 3, characterized in that: The question-answer pair database is a vector database, and the information stored in the vector database is vector information obtained by converting the questions in the question-answer pair from text form to vector form.

5. The method for identifying intentions in an intelligent question-answering system according to claim 4, characterized in that: The retrieving a question-answer pair matching the query question based on the similarity between the query question vector and the vectors of each question in the question-answer pair database comprises: Calculate the similarity score between the query question vector and the vectors of each question in the vector database, and retrieve the top k question-answer pairs with higher similarity scores; Compare the similarity scores of the first k question-answer pairs with a preset threshold; When at least one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, the intent type of the retrieved question-answer pair is used as the intent type of the query question; When the similarity scores of the first k question-answer pairs are not greater than the preset threshold, the user query is input into the large language model for intent recognition.

6. The method for identifying intentions in an intelligent question-answering system according to claim 5, characterized in that: When at least one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, taking the intent type of the retrieved question-answer pair as the intent type of the query question includes: When only one of the similarity scores of the first k question-answer pairs is greater than the preset threshold, the intent type of the question-answer pair is used as the intent type of the query question; When there are multiple similarity scores of the first k question-answer pairs that are greater than the preset threshold, the intent types of the first k question-answer pairs are deduplicated to obtain deduplicated intent types, and the first n intent types are taken as the intent types of the query question according to the similarity scores, where n≤k.

7. The method for identifying intentions in an intelligent question-answering system according to claim 6, characterized in that: Before extracting keywords from the query question and obtaining the semantic features of the query question, the method further includes: Modify the set vocabulary in the query question based on a preset vocabulary library, wherein the set vocabulary includes professional terms, abbreviations, and typos; By analyzing the query language pattern, keyword completeness and context consistency, it is determined whether the query question is complete. If the query question is incomplete, a dialogue is used to request information to clarify the question from the user; If the query question is complete, the step of extracting keywords from the query question and obtaining semantic features of the query question is performed.

8. The method for identifying intentions in an intelligent question-answering system according to claim 5, characterized in that: The calculating of the similarity score between the query question vector and the vectors of each question in the vector database comprises: The cosine similarity algorithm is used to calculate the similarity score between the query question vector and the vectors of each question in the vector database.

9. An intention recognition device for an intelligent question-answering system, characterized in that: It includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the intelligent question and answer system intention recognition method according to any one of claims 1-8 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the method for identifying intentions in an intelligent question and answer system according to any one of claims 1 to 8 is implemented.

Images