Intelligent question answering system and intelligent question answering method based on large model

Abstract

The invention provides an intelligent question-answering system and an intelligent question-answering method based on a large model. In the system, an intention recognition module is used for receiving a natural language question input by a user, and semantic recognition is performed on the natural language question by a large language model to obtain an intention recognition result; the SQL conversion module is used for obtaining a structured SQL query statement corresponding to the intention recognition result; the data retrieval module is used for performing retrieval in a database by utilizing the SQL query statement obtained by the SQL conversion module to obtain a retrieval result; the content analysis module is used for analyzing the retrieval result obtained by the data retrieval module by using a deep learning technology so as to obtain an analysis conclusion, and the analysis conclusion comprises key information and a mode corresponding to the key information; and the result display module is used for displaying the retrieval result and the analysis conclusion to the user in a visual form. According to the method, the user can retrieve the information in the database by using the intelligent question and answer system without needing to have SQL (Structured Query Language) knowledge.

Description

Technical Field

[0001] One or more embodiments of the present invention relate to network communication technologies, and in particular, to an intelligent question-answering system and an intelligent question-answering method based on large models. Background Art

[0002] Currently, intelligent question-answering systems are becoming more and more popular. Through intelligent question-answering systems, users can query the information they want to know. For example, if a user wants to obtain the sales record of a certain product in the previous month, then the user can input corresponding query information to the intelligent question-answering system, and the intelligent question-answering system retrieves and queries and provides an answer to the user.

[0003] Currently, traditional data query methods rely on users having SQL knowledge. For example, SQL statements need to be used to query in the database, which limits the access and understanding of data by non-professionals. For example, if a user wants to perform the above query, they need to input a query statement in SQL format to the intelligent question-answering system to query in the database. However, most users do not have SQL knowledge, which greatly limits the use of intelligent question-answering systems. Summary of the Invention

[0004] One or more embodiments of the present invention describe an intelligent question-answering system and an intelligent question-answering method based on large models, which can solve at least one problem in the prior art.

[0005] According to a first aspect, there is provided an intelligent question-answering system based on a large model, the system comprising: an intent recognition module, an SQL conversion module, a data retrieval module, a database, a content analysis module, and a result display module; wherein,

[0006] The intent recognition module is configured to receive a natural language question input by a user, preprocess the question to obtain a preprocessed text-form question; input the preprocessed text-form question to a large language model, and the large language model performs semantic recognition on the received natural language question, understands the user's natural language question and determines the query intent, so as to obtain an intent recognition result;

[0007] The SQL conversion module is configured to, for the intent recognition result output by the large language model, use NL2SQL technology and based on the storage structure, data type, and required data operation type of the data, obtain a structured SQL query statement corresponding to the intent recognition result, and the SQL query statement defines the data that needs to be extracted from the database;

[0008] The data retrieval module is configured to use the SQL query statement obtained by the SQL conversion module to perform retrieval in the database to obtain a retrieval result;

[0009] A content analysis module, which is used to analyze the retrieval results obtained by the data retrieval module by using deep learning technology, so as to obtain an analysis conclusion, and the analysis conclusion includes key information and the pattern corresponding to the key information;

[0010] A result display module, which is used to present the retrieval results and the analysis conclusion to the user in a visual form.

[0011] The intention recognition module includes:

[0012] A text input information processing sub-module, which is used to receive the text-format question input by the user, and perform at least one of corpus cleaning, word segmentation, semantic error correction, and part-of-speech tagging on the text-format question, so as to obtain the preprocessed text-format question;

[0013] And / or,

[0014] A voice input information processing sub-module, which is used to receive the voice-format question input by the user, convert the voice-format question into a text-format question, and then perform at least one of corpus cleaning, word segmentation, semantic error correction, and part-of-speech tagging on the text-format question, so as to obtain the preprocessed text-format question.

[0015] The intention recognition module learns the intention information contained in different words, phrases, sentences, and contexts through a pre-training process and using semantic analysis and sentiment calculation.

[0016] The data retrieval module is also used to adjust the structure of the SQL query statement and perform retrieval in the database using an index and a data access algorithm.

[0017] The SQL conversion module uses NL2SQL technology to perform pre-training with data in a predetermined domain, so as to obtain an SQL conversion module that meets the conversion accuracy rate for the predetermined domain.

[0018] The result display module is used to display the data change trend to the user by using charts and / or graphs; and / or, it is used to convert the retrieval results into charts and / or graphs and display them to the user.

[0019] The SQL conversion module converts the natural language intention recognition result into a structured SQL query statement through deep learning and a sequence-to-sequence model.

[0020] The content analysis module further includes: a trend analysis unit, which is used to identify the trend of the retrieval results changing over time as the analysis conclusion;

[0021] And / or,

[0022] The content analysis module is also used to assist the user in understanding the meaning of the data and its analysis results.

[0023] The database includes: a database in the financial field, a database in the medical field, and / or a database in the retail field.

[0024] According to a second aspect, there is provided an intelligent question-answering method for an intelligent question-answering system based on a large model, the method comprising:

[0025] Receiving a question in natural language input by the user, preprocessing the question to obtain a preprocessed question in text form;

[0026] Inputting the preprocessed question in text form into a large language model;

[0027] The large language model performs semantic recognition on the received question in natural language, understands the user's natural language question and determines the query intention, thereby obtaining an intention recognition result;

[0028] For the intention recognition result output by the large language model, using the NL2SQL technology and based on the storage structure, data type, and required data operation type of the data, obtaining a structured SQL query statement corresponding to the intention recognition result, and this SQL query statement defines the data that needs to be extracted from the database;

[0029] Using the SQL query statement to retrieve in the database to obtain a retrieval result;

[0030] Analyzing the retrieval result to obtain an analysis conclusion, and this analysis conclusion includes key information and the pattern corresponding to this key information;

[0031] Presenting the retrieval result and the analysis conclusion to the user in a visual form.

[0032] The intelligent question-answering system and intelligent question-answering method based on a large model provided by various embodiments of the present invention have at least the following beneficial effects:

[0033] 1. In the present invention, the user does not need to have SQL knowledge. When the user inputs a question to the intelligent question-answering system, the user only needs to input a question in natural language (such as a question in voice form, or a question in text form, etc.), and the intention recognition module can recognize the user's intention, that is, recognize what information the user needs to know. After that, the SQL conversion module can automatically generate a structured SQL query statement corresponding to the intention recognition result (that is, the information the user needs to know), without the user inputting an SQL query statement and without the user having SQL knowledge. Subsequently, after the data retrieval module, the database, the content analysis module, and the result display module cooperate, the retrieval result can be output to the user.

[0034] 2. The present invention uses a large language model for natural language understanding and generation, and combines NL2SQL technology to convert natural language questions into SQL query statements for database queries. At the same time, a deep learning model can be used to analyze the query results to provide more comprehensive explanations and insights. The intelligent question-answering system of the present invention can not only understand the user's natural language questions and convert them into SQL query statements, but also perform in-depth content analysis on the query results, and finally intuitively display them to the user in a visual form for easy understanding and decision-making.

[0035] 3. In the system of the present invention, complex query results are converted into easy-to-understand charts and graphs through data processing. These visualization elements are not only colorful but also exquisitely designed, and can intuitively show the relationship and trend between data. Users can quickly capture key information without digging deep into the meaning behind the data. The intelligent function of the system of the present invention is not limited to the display of data. It can also automatically identify and extract key information from the data and make effective induction and summary. This process not only saves users' time, but also improves the accuracy of information. In addition, the system will provide professional explanations and practical suggestions based on the analysis results, which are designed to help users understand the meaning behind the data more deeply and make more informed decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 It is a structural diagram of an intelligent question-answering system based on a large model in one embodiment of the present invention.

[0038] Figure 2 It is a schematic diagram of an intelligent question-answering method based on a large model in one embodiment of the present invention.

[0039] Figure 3 It is a flow chart of an intelligent question-answering method based on a large model in another embodiment of the present invention. DETAILED DESCRIPTION

[0040] First of all, it should be noted that the terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "the" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates otherwise.

[0041] It should be understood that the term "and / or" used herein is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. Additionally, the character " / " in this text generally indicates that the associated objects before and after are in an "or" relationship.

[0042] Figure 1 is a schematic structural diagram of an intelligent question-answering system based on a large model in an embodiment of the present invention. Refer to Figure 1 This system includes: an intent recognition module 101, an SQL conversion module 102, a data retrieval module 103, at least one database 104, a content analysis module 105, and a result display module 106; wherein,

[0043] The intent recognition module 101 is configured to receive a question in natural language input by a user, preprocess the question to obtain a preprocessed text-form question; input the preprocessed text-form question into a large language model, and the large language model performs semantic recognition on the received natural language question, understands the user's natural language question, and determines the query intent, thereby obtaining an intent recognition result;

[0044] The SQL conversion module 102 is configured to, for the intent recognition result output by the large language model, use NL2SQL technology and based on the storage structure, data type, and required data operation type of the data, obtain a structured SQL query statement corresponding to the intent recognition result, and this SQL query statement defines the data that needs to be extracted from the database;

[0045] The data retrieval module 103 is configured to use the SQL query statement obtained by the SQL conversion module to retrieve in the database 104 to obtain a retrieval result;

[0046] The content analysis module 105 is configured to use deep learning technology to analyze the retrieval result obtained by the data retrieval module, thereby obtaining an analysis conclusion, and this analysis conclusion includes key information and the pattern corresponding to the key information;

[0047] The result display module 106 is configured to present the retrieval result and the analysis conclusion to the user in a visual form.

[0048] According to the above Figure 1As can be seen from the system shown, in the present invention, the user does not need to have SQL knowledge. When the user inputs a question to the intelligent question-answering system, the user only needs to input a question in natural language (such as a question in voice form, or a question in text form, etc.), and the intention recognition module can recognize the user's intention, that is, recognize what information the user needs to know. After that, the SQL conversion module can automatically generate a structured SQL query statement corresponding to the intention recognition result (that is, the information the user needs to know), without the user inputting an SQL query statement and without the user having SQL knowledge. Subsequently, after the data retrieval module, the database, the content analysis module, and the result display module cooperate, the retrieval result can be output to the user, and the analysis conclusion of the system can be further provided.

[0049] It can be seen that the present invention uses a large language model for natural language understanding and generation, combines NL2SQL technology to convert natural language questions into SQL query statements for database queries, and can also use a deep learning model to analyze the query results to provide more comprehensive explanations and insights. The intelligent question-answering system of the present invention can not only understand the user's natural language questions and convert them into SQL query statements, but also perform in-depth content analysis on the query results, and finally visually display them to the user in a visual form for easy understanding and decision-making.

[0050] In the system of the present invention, for the intention recognition module 101:

[0051] Participate Figure 1 , in an embodiment of the system of the present invention, the intention recognition module 101 includes:

[0052] A text input information processing sub-module, configured to receive a question in text format input by the user, and perform at least one of corpus cleaning processing, word segmentation processing, semantic error correction processing, and part-of-speech tagging processing on the question in text format, so as to obtain a preprocessed question in text form;

[0053] And / or,

[0054] A voice input information processing sub-module, configured to receive a question in voice format input by the user, convert the question in voice format into a question in text format, and then perform at least one of corpus cleaning processing, word segmentation processing, semantic error correction processing, and part-of-speech tagging processing on the question in text format, so as to obtain a preprocessed question in text form.

[0055] It can be seen that based on the intention recognition module 101, the user can ask questions in any form of natural language, such as the above-mentioned text format or voice format, thus providing great convenience for the user's use.

[0056] In an embodiment of the system of the present invention, the intent recognition module 101 learns the intent information contained in different words, phrases, sentences, and contexts through a pre-training process and by using semantic analysis and sentiment computing.

[0057] Intent recognition is a key technology in the field of artificial intelligence. It involves in-depth analysis of the information input by users to determine their true needs and purposes. This process not only requires the machine to have a high level of language understanding ability but also demands the ability to handle various complex contexts and implicit intents. Through precise intent recognition, the system can provide more accurate and personalized responses, greatly enhancing the user experience.

[0058] In the process of implementing intent recognition, machine learning and natural language processing technologies play a core role. By training a large number of data models, the machine can learn the intent information contained in different words, phrases, and even sentence structures. In addition, context understanding is an indispensable part of intent recognition, which helps the machine better grasp the flow of the conversation and the true intent of the user.

[0059] To improve the accuracy of intent recognition, modern AI systems also combine multiple technologies, such as semantic analysis and sentiment computing, to further refine and optimize the recognition results. For example, through sentiment computing, the system can identify the emotional state of the user when asking for information, so as to be more considerate and context-appropriate in the answer.

[0060] With the continuous progress of technology, the application scope of intent recognition is also constantly expanding. From the initial simple question-and-answer systems to the current intelligent assistants, customer service robots, etc., the application scenarios of intent recognition technology are becoming more and more extensive, and its accuracy and response speed are also constantly improving, greatly promoting the development and application of artificial intelligence technology.

[0061] When the user asks a question in natural language, the system uses a large language model (LLM) for in-depth semantic analysis to accurately capture and understand the essential needs of the question. This process involves a detailed interpretation of the user's query to ensure that the true intent of the user can be grasped, providing a solid foundation for subsequent processing.

[0062] In the system of the present invention, for the SQL conversion module 102:

[0063] The SQL conversion module 102 uses NL2SQL technology to perform pre-training with data in a predetermined domain, so as to obtain an SQL conversion module that meets the conversion accuracy rate for that predetermined domain. NL2SQL usually performs pre-training based on specific domain data, requiring it to have solid SQL knowledge and rich practical experience to ensure that it can efficiently and accurately generate and retrieve the required data.

[0064] The SQL conversion module 102 can convert the result of natural language intent recognition into a structured SQL query statement through deep learning and sequence-to-sequence models.

[0065] NL2SQL generates specific SQL statements based on the previously recognized intents, and these statements can precisely define which data needs to be extracted from the database.

[0066] When generating SQL statements, it is necessary to consider the storage structure of the data, data types, and the types of data operations required. The correct SQL statements can ensure the accuracy and efficiency of data operations, avoiding errors and performance bottlenecks.

[0067] In the system of the present invention, for the data retrieval module 103:

[0068] The data retrieval module 103 is also used to adjust the structure of the SQL query statement and perform retrieval in the database using indexes and data access algorithms.

[0069] The retrieval process is to execute the written SQL statement to query or perform other operations on the database through a database management system (DBMS). This process usually involves filtering, sorting, and presenting data to meet specific information needs. The accuracy and speed of the retrieval results directly affect the performance of the application and the user experience.

[0070] In addition, SQL generation and retrieval can also include the process of optimizing the query to ensure the speed and efficiency of data processing. This may involve adjusting the structure of the SQL statement, using indexes, or adopting more efficient data access algorithms.

[0071] It can be seen that in the system of the present invention, the NL2SQL technology is used to convert the natural language query input by the user into a structured SQL query statement. This process involves deeply understanding the meaning of natural language and accurately mapping it to the database query language. In this way, even users without professional SQL knowledge can easily obtain the required information from the database. After completing the semantic-to-SQL conversion, the system further executes these SQL queries. This usually involves connecting to a database in a specific field, such as a database in the financial, medical, or retail industries. The system retrieves relevant data in these databases according to the generated SQL statements. This step ensures that users can obtain accurate and relevant data, thus supporting advanced tasks such as decision-making and data analysis.

[0072] In the system of the present invention, for the database 104:

[0073] The database can include: a database in the financial field, a database in the medical field, and / or a database in the retail industry.

[0074] In the system of the present invention, for the content analysis module 105:

[0075] The content analysis module 105 further includes: a trend analysis unit for identifying the trend of the search results changing over time as the analysis conclusion;

[0076] and / or,

[0077] The content analysis module 105 is also used to assist users in understanding the meaning of data and its analysis results.

[0078] In the content analysis stage, we dig deep and evaluate the core value of information. This process involves a careful review of the data to identify key trends, patterns and insights. Content analysis also includes the classification and labeling of information to facilitate subsequent retrieval and application. In the content analysis stage, we also focus on the timeliness and relevance of the content to ensure that the analysis results reflect the latest market dynamics and social trends. On this basis, we further explore the potential impact and value of the content to provide strong data support for decision-making.

[0079] After the data retrieval is completed, the content analysis module 105 uses advanced deep learning technology to deeply process the acquired data. In this process, key information is accurately extracted and the hidden patterns and trends in the data are identified. The deep learning algorithm can reveal the complex relationship between the data and provide strong support for trend analysis. This comprehensive analysis not only improves the efficiency of data processing, but also ensures the accuracy and reliability of the results.

[0080] In the system of the present invention, for the result display module 106:

[0081] The result display module 106 is used to display data change trends to the user using charts and / or graphs; and / or, to convert the search results into charts and / or graphs and display them to the user.

[0082] The result display module 106 displays the search results to all relevant parties in a variety of ways, and can use detailed data reports, intuitive chart analysis, and actual product demonstrations to present the search results.

[0083] Use charts and graphs to depict data trends and help users quickly grasp the key points of information. These visual tools not only enhance the appeal of reports, but also make it easier to interpret complex data.

[0084] In the system of the present invention, complex query results are converted into easy-to-understand charts and graphs through data processing. These visualization elements are not only colorful but also exquisitely designed, and can intuitively show the relationship and trend between data. Users can quickly capture key information without having to dig deep into the meaning behind the data.

[0085] The intelligent functions of the system of the present invention are not limited to data display. It can also automatically identify and extract key information from the data and perform effective summarization. This process not only saves the user's time but also improves the accuracy of information. In addition, the system will provide professional explanations and practical suggestions based on the analysis results, which are designed to help users understand the meaning behind the data more deeply and thus make more informed decisions.

[0086] The system of the present invention can significantly improve the efficiency of data query and analysis, reduce the learning cost of users, and enhance the user experience. Through in-depth learning content analysis, the system can reveal the implicit information behind the data and promote more informed decision-making.

[0087] An embodiment of the present invention proposes an intelligent question-answering method based on a large model. See Figure 2 、 Figure 3 , and the method includes:

[0088] Step 301: Receive a question in natural language input by the user, preprocess the question, and obtain a preprocessed text-form question.

[0089] Step 303: Input the preprocessed text-form question into a large language model.

[0090] Step 305: The large language model performs semantic recognition on the received question in natural language, understands the user's natural language question, and determines the query intention, thereby obtaining an intention recognition result.

[0091] Step 307: For the intention recognition result output by the large language model, use the NL2SQL technology and based on the storage structure, data type, and required data operation type of the data, obtain a structured SQL query statement corresponding to the intention recognition result, and this SQL query statement defines the data that needs to be extracted from the database.

[0092] Step 309: Use the SQL query statement to retrieve in the database and obtain a retrieval result.

[0093] Step 311: Analyze the retrieval result to obtain an analysis conclusion, and this analysis conclusion includes key information and the pattern corresponding to the key information.

[0094] Step 313: Present the retrieval result and the analysis conclusion to the user in a visual form.

[0095] It should be noted that the above-mentioned modules are generally implemented on the server side. They can be respectively set on independent servers, or some or all of the devices can be combined and set on the same server. The server can be a single server or a server cluster composed of multiple servers. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system. The above-mentioned devices can also be implemented on a computer terminal with strong computing power.

[0096] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed on a computer, the computer is made to execute the method in any one of the embodiments in the specification.

[0097] An embodiment of the present invention provides a computing device, including a memory and a processor. An executable code is stored in the memory. When the processor executes the executable code, the method in any one of the embodiments in the specification is implemented.

[0098] It can be understood that the structure schematically shown in the embodiments of the present invention does not constitute a specific limitation on the devices in the embodiments of the present invention. In other embodiments of the specification, the above-mentioned devices may include more or fewer components than shown in the figures, or combine certain components, or split certain components, or have different component arrangements. The components shown in the figures can be implemented in hardware, software, or a combination of software and hardware.

[0099] The various embodiments of the present invention are described in a progressive manner. The same or similar parts among the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0100] Those skilled in the art should be able to realize that in the above one or more examples, the functions described in the present invention can be implemented by hardware, software, add-ons, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.

[0101] The above-mentioned specific implementation manners have further elaborated on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above is only the specific implementation manner of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solution of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent question-answering system based on a large model, characterized in that The system includes: an intent recognition module, an SQL conversion module, a data retrieval module, a database, a content analysis module, and a result display module; wherein, The intent recognition module is configured to receive a question in natural language input by a user, preprocess the question to obtain a preprocessed question in text form; input the preprocessed question in text form to a large language model, and the large language model performs semantic recognition on the received natural language question, understands the user's natural language question and determines the query intent, so as to obtain an intent recognition result; The SQL conversion module is configured to, for the intent recognition result output by the large language model, use the NL2SQL technology and based on the storage structure, data type, and required data operation type of the data, obtain a structured SQL query statement corresponding to the intent recognition result, and this SQL query statement defines the data that needs to be extracted from the database; The data retrieval module is configured to use the SQL query statement obtained by the SQL conversion module to perform a retrieval in the database to obtain a retrieval result; The content analysis module is configured to use deep learning technology to analyze the retrieval result obtained by the data retrieval module, so as to obtain an analysis conclusion, and this analysis conclusion includes key information and the pattern corresponding to the key information; The result display module is configured to present the retrieval result and the analysis conclusion to the user in a visual form.

2. The system according to claim 1, wherein The intent recognition module includes: A text input information processing sub-module, configured to receive a question in text format input by a user, and perform at least one of corpus cleaning processing, word segmentation processing, semantic error correction processing, and part-of-speech tagging processing on the question in text format, so as to obtain a preprocessed question in text form; And / or, A voice input information processing sub-module, configured to receive a question in voice format input by a user, convert the question in voice format into a question in text format, and then perform at least one of corpus cleaning processing, word segmentation processing, semantic error correction processing, and part-of-speech tagging processing on the question in text format, so as to obtain a preprocessed question in text form.

3. The system according to claim 1, wherein The intent recognition module learns the intent information contained in different words, phrases, sentences, and contexts through a pre-training process and using semantic analysis and sentiment calculation.

4. The system according to claim 1, characterized in that, The data retrieval module is further configured to adjust the structure of the SQL query statement and perform a retrieval in the database using an index and a data access algorithm.

5. The system according to claim 1, wherein The SQL conversion module uses data in a predetermined domain for pre-training using the NL2SQL technology, so as to obtain an SQL conversion module that meets the conversion accuracy rate for the predetermined domain.

6. The system according to claim 1, characterized in that The result display module is configured to use charts and / or graphs to display the data change trend to the user; and / or, configured to convert the retrieval result into charts and / or graphs and display them to the user.

7. The system according to claim 1, characterized in that, The SQL conversion module converts the natural language intent recognition result into a structured SQL query statement through deep learning and a sequence-to-sequence model.

8. The system according to claim 1, wherein The content analysis module further includes: a trend analysis unit, configured to identify the trend of the retrieval result changing over time as the analysis conclusion; And / or, The content analysis module is also used to assist users in understanding the meaning of data and its analysis results.

9. The system according to any one of claims 1 to 8, characterized in that, The database includes: a database in the financial field, a database in the medical field, and / or a database in the retail industry field.

10. The intelligent question-answering method of the large model-based intelligent question-answering system according to any one of claims 1 to 9, characterized in that, The method includes: Receiving a question in natural language input by the user, preprocessing the question, and obtaining the preprocessed question in text form; Inputting the preprocessed question in text form into a large language model; The large language model performs semantic recognition on the received question in natural language, understands the user's natural language question and determines the query intention, so as to obtain an intention recognition result; For the intention recognition result output by the large language model, using the NL2SQL technology and based on the storage structure of the data, the data type, and the required data operation type, obtaining a structured SQL query statement corresponding to the intention recognition result, and the SQL query statement defines the data that needs to be extracted from the database; Using the SQL query statement to retrieve in the database and obtaining a retrieval result; Analyzing the retrieval result to obtain an analysis conclusion, and the analysis conclusion includes key information and the pattern corresponding to the key information; Presenting the retrieval result and the analysis conclusion to the user in a visual form.

Images