Financial concept question and answer method and system for personal wealth management standard concept library based on dynamic amplification

Abstract

The invention relates to a financial concept question-answering method and system based on a dynamically-amplified personal wealth management standard concept library, and discloses a financial concept question-answering method based on the dynamically-amplified personal wealth management standard concept library, which comprises the following steps of: 1, preprocessing a user question; comprising problem segmentation and concept extraction; 2, inquiring whether each concept after preprocessing is in a database or not; 3, performing synonym judgment on each concept in the step 2 and a concept in a database; 4, performing networking retrieval on each concept in the step 3 to obtain explanations corresponding to the concepts; and 5, finally embedding the prompt words to generate a final prompt word, and submitting the final prompt word to the Deepseek-v3 model to generate a final answer. According to the method, the dynamically updated financial concept knowledge base is constructed and realized, so that the continuously expanded and rapidly evolved knowledge requirements in the financial field can be efficiently met.

Description

Technical Field

[0001] This invention relates to a financial concept question-and-answer method and system, and more particularly to a financial concept question-and-answer method and system based on a dynamically expanding personal wealth management standard concept library, which belongs to the application of artificial intelligence technology in the financial field. Background Technology

[0002] With the rapid development of financial markets, the demand for personal wealth management is growing, making accurate and timely updated financial concept knowledge crucial for investors. Currently, financial concepts exhibit dynamic expansion and diversified evolution: technology-driven innovations, such as blockchain finance and robo-advisors, and policy-driven expansion into areas like green finance and inclusive finance. These concepts encompass both traditional financial terms like "credit" and "stocks," as well as innovative financial expressions such as "shadow banking" and "circuit breaker mechanisms," significantly enhancing their professionalism and technological integration. Ensuring accurate and standardized answers to emerging financial concepts in question-and-answer systems has become a key requirement in the financial Q&A field.

[0003] Currently, most question-answering systems rely on large language models for responses, and the knowledge within these models is fixed after training. Large language models primarily use retrieval-enhanced knowledge generation, with knowledge sources mainly including vector databases, relational databases, and the internet. For vector databases and internet retrieval methods, it's difficult to ensure consistency in the definitions of new concepts introduced into the large language model, thus compromising the accuracy and standardization of responses and making the model susceptible to misleading information from irrelevant search results. Furthermore, existing personal wealth management standard concept libraries are mostly static or semi-static, resulting in slow updates to meet users' demands for the latest financial knowledge in the face of constantly emerging new financial concepts, products, and investment strategies. Therefore, in financial concept question-answering scenarios, ensuring consistency in the definitions of the same concept retrieved from the knowledge base and the real-time updates of the knowledge base are key directions for improving question-answering quality. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a financial concept question-and-answer method and system based on a dynamically expanded personal wealth management standard concept library, aiming to ensure the real-time nature, accuracy, and reliability of definitions and answers.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] The first aspect of this invention discloses a financial concept question-and-answer method based on a dynamically expanded personal wealth management standard concept library, comprising the following steps:

[0007] Step 1: User question preprocessing, including question segmentation and concept extraction;

[0008] Step 2: For each preprocessed concept To check if it already exists in the database;

[0009] Step 3: For those still in Step 2 Each concept in It performs synonym comparison with concepts in the database;

[0010] Step 4: For those still in step 3 Each concept in To retrieve the corresponding explanations of the concepts through online searches;

[0011] Step 5: Finally, The final prompt is generated by embedding prompts into the question and then handed over to the deepseek-v3 model to generate the final answer.

[0012] Step 1 includes:

[0013] Step 1.1: Receive user questions. To prevent excessively long user input, the user input is split into segments, generating a list of text fragments: contents;

[0014] Step 1.2: Traverse the text fragment list `contents`, embed the `content` into the prompt words, and call the deepseek-v3 model to extract concepts, obtaining a concept list. ;

[0015] Step 1.3: List the concepts The suggestion words are embedded and the deepseek-v3 model is called to determine which concepts belong to the financial field, resulting in a filtered list of concepts. .

[0016] Step 2 includes:

[0017] Step 2.1: If concept c belongs to a concept stored in the database or is a synonym of a concept stored in the database, then remove c from the database. Remove and add to dictionary ,in Explanation of concepts retrieved from the database;

[0018] Step 2.2: If If there are still concepts that have not been removed, proceed to step 3; if If the value is empty, proceed to step 5.

[0019] Step 3 includes:

[0020] Step 3.1: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require Split into multiple lists The number of partitions n is determined by The total number of concepts and the maximum number of concepts k1 after the hyperparameter segmentation are determined by the total number of concepts in the dataset. The specific formula is as follows: ;

[0021] Step 3.2: Divide all concepts in the database into multiple lists. The number of lists *m* to be split is determined by the number of concepts *N* in the database and the maximum number of concepts *k2* after splitting, as set by the hyperparameter. The specific formula is as follows: ;

[0022] Step 3.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require and The permutations and combinations of embedded prompts form n * m reasoning tasks, which are then fed into the deepseek-v3 model to generate a thesaurus. ;

[0023] Step 3.4: Remove all concepts c that have synonyms from... Remove and add to dictionary And append the synonym c to the corresponding synonym set in the database of c';

[0024] Step 3.5: If If there are still concepts that have not been removed, proceed to step 4; if If empty, proceed to step 5.

[0025] Step 4 includes:

[0026] Step 4.1: For each filtered concept The system uses an online search tool to retrieve the content W of the webpage that ranks first in relevance among the relevant webpages. It then recursively divides the content W into a list of texts of size W_max. To prevent the size of the prompt word from exceeding the window size of the large language model;

[0027] Step 4.2: Traversal The deepseek-v3 model determines whether it contains [the specified content]. The definition and output of the result in JSON format includes whether it was generated by a large language model and the definition of the concept; if it was not generated by a large language model, the iteration ends, because this means the model is in W. i Found in Definition;

[0028] Step 4.3: Put all Add the corresponding explanations obtained in step 4.2 to the dictionary. And update the database.

[0029] In a second aspect, this invention discloses a financial concept question-and-answer system based on a dynamically expanded personal wealth management standard concept library, including a financial concept extraction module, a financial concept retrieval module, a financial concept addition module, and a question-and-answer module;

[0030] The financial concept extraction module is responsible for identifying potential financial concepts from input documents or user-submitted questions.

[0031] The financial concept retrieval module receives concept information extracted by the financial concept extraction module and performs a retrieval of concept definitions.

[0032] The aforementioned financial concept addition module is used to add concepts from the financial concept retrieval module that are not found in the relational database to the database.

[0033] The question-and-answer module receives the definitions of concepts retrieved by the above module and generates answers to the questions.

[0034] The financial concept extraction module also includes:

[0035] The financial concept recognition unit uses prompts and a language model to identify financial concepts in user input.

[0036] The financial concept screening unit receives the recognition results from the financial concept identification unit and filters out concepts that truly belong to the financial field, ensuring that the identified concepts belong to the financial field.

[0037] The financial concept retrieval module includes:

[0038] The structured data retrieval unit is specifically designed for precise retrieval in structured databases to identify known definitions related to the input concept;

[0039] The concept semantic extension unit is mainly responsible for determining whether there are synonyms for the input concept in the structured database;

[0040] The unstructured data retrieval unit is responsible for retrieving the definitions corresponding to concepts from the documents input by the user;

[0041] The network retrieval unit is responsible for accessing the Internet to retrieve more conceptual information when the above retrieval units cannot provide a conceptual definition. If the Internet resources still cannot find a clear definition, the system will generate a reasonable conceptual definition based on the existing information.

[0042] In a third aspect, the present invention discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0043] In a fourth aspect, the present invention discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0044] The beneficial effects of adopting the above technical solution are as follows:

[0045] 1. This invention constructs and implements a dynamically updated financial concept knowledge base, which can efficiently address the ever-expanding and rapidly evolving knowledge needs in the financial field. In this process, the system ensures the idempotency of concept retrieval operations, meaning that regardless of the number of searches, the retrieval results for the same concept remain consistent, and it ensures the real-time nature, accuracy, and reliability of definitions and answers.

[0046] 2. This invention, through synonym determination, not only reduces the number of network queries and significantly reduces the number of database updates, but also improves the understanding of financial synonyms in user questions by large models, thereby reducing the illusion of large models and improving the quality of answers.

[0047] 3. This invention organizes and summarizes new financial concepts that can be retrieved from the Internet through prompt word engineering, simplifying the process of users searching for and learning new concepts, and helping users to correctly understand new financial concepts and grasp their core points in a short time. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the financial concept question-and-answer method of the present invention.

[0049] Figure 2 This is a block diagram of the financial concept question-and-answer system of the present invention. Detailed Implementation

[0050] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0051] Example 1

[0052] like Figure 1 As shown, the first aspect of this invention discloses a financial concept question-and-answer method based on a dynamically expanded personal wealth management standard concept library, comprising the following steps:

[0053] Step 1: User question preprocessing, including question segmentation and concept extraction;

[0054] Step 1.1: Receive user question. If the question length is greater than the set maximum text fragment length l, segment the user input according to semantics and generate a list of text fragments with a length less than l: contents; otherwise, contents is a list containing only the question.

[0055] Step 1.2: Traverse the text fragments (contents), embed the content into the prompt words, and call the deepseek-v3 model to extract concepts, obtaining a concept list. ;

[0056] Step 1.3: List the concepts The prompts are embedded and the DeepSeek-v3 model is called to determine which concepts belong to the financial field. Concepts that are not in the financial field are then filtered out to obtain a list of concepts. This list of concepts contains at least one concept. ;

[0057] Step 2: For each concept after preprocessing and filtering in Step 1.3 Perform a query, query concept Is it already in the database?

[0058] Including step 2.1: If concept c belongs to a concept stored in the database or a synonym of a concept stored in the database, then remove c from... Remove from the dictionary and add the concept of removal, c, to the dictionary. ,in This is an explanation of a concept retrieved from the database. For example, if the synonym for "savings" is "deposit" in the database, then "savings" will be explained from... Delete the relevant entries and add the corresponding definitions of savings and deposits to the dictionary.

[0059] Step 2.2: If If there are still concepts that have not been removed, proceed to step 3; if If empty, proceed to step 5;

[0060] Step 3: For the steps in step 2.2 that are still in progress... Each concept in This involves identifying synonyms with concepts in the database; specifically, common synonyms in the financial field include: inflation and inflation, gross domestic product and GDP, bull market and bullish market.

[0061] Including step 3.1: to Split into multiple lists The number of partitions n is determined by The total number of concepts and the maximum number of concepts k1 after the hyperparameter segmentation are determined by the total number of concepts in the dataset. The specific formula is as follows: For the n lists after splitting, the first n-1 lists have k1 concepts, and the last list has the remaining concepts;

[0062] Step 3.2: Divide all concepts in the database into multiple lists. The number of lists *m* to be split is determined by the number of concepts *N* in the database and the maximum number of concepts *k2* after splitting, as set by the hyperparameter. The specific formula is as follows: For the resulting m lists, the first m-1 lists contain k2 concepts, and the last list contains the remaining concepts.

[0063] Step 3.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require and The permutations and combinations of embedded prompts form n * m reasoning tasks, which are then fed into the deepseek-v3 model to generate a thesaurus. ;

[0064] Step 3.4: Remove all concepts c that have synonyms from... Remove, and add the concept of removal c to the dictionary. And append the synonym c to the corresponding synonym set in the database of c';

[0065] Step 3.5: If If there are still concepts that have not been removed, proceed to step 4; if If empty, proceed to step 5;

[0066] Step 4: For the steps still in step 3.5 Each concept in To retrieve the corresponding explanations of the concepts through online searches;

[0067] Including step 4.1: For steps 3.5 still in progress... For each concept in the process, an online search tool is invoked to obtain the content W of the webpage with the highest relevance among the relevant webpages. The content W of the webpage is then recursively divided into a list of texts of size W_max. To prevent the size of the prompt word from exceeding the window size of the large language model;

[0068] Step 4.2: Traversal The deepseek-v3 model determines whether it contains [the specified content]. The definition and output of the result in JSON format includes whether it was generated by a large language model and the definition of the concept; if it was not generated by a large language model, the iteration ends, because this means the model is in W. i Found in Definition;

[0069] Step 4.3: Put all Add the corresponding explanations obtained in step 4.2 to the dictionary. And update the database;

[0070] Step 5: Finally, The final prompt is generated by embedding prompts into the question and then handed over to the deepseek-v3 model to generate the final answer.

[0071] The following is a detailed explanation of the specific situations that may arise during use.

[0072] Example 1: All concepts in the problem are in the concept bank.

[0073] User question: "What's the difference between stocks and the stock market?"

[0074] Step 1: User Issue Preprocessing

[0075] Step 1.1: Receive user questions such as "What is the difference between stocks and the stock market?";

[0076] The question is divided into text segments; if the question is too short, it is divided into one segment.

[0077] contents = [“What’s the difference between stocks and the stock market?”].

[0078] Step 1.2: Use the deepseek-v3 model to extract concepts and obtain a concept list. ={Stocks, Stock Market, Difference}.

[0079] Step 1.3: Use the deepseek-v3 model to determine which concepts belong to the financial domain, and obtain a filtered list of concepts. ={stocks, stock market}.

[0080] Step 2: Query the concept in the database

[0081] Step 2.1: Check if "stocks" and "stock market" are in the database; both "stocks" and "stock market" are in the database, and the database already exists.

[0082] Step 2.2: Remove "stocks" and "stock market". Retrieve relevant definitions from the database and build a dictionary. ={Stocks: Stocks refer to shares issued by a company, which are a portion of the company's equity; Stock Market: The stock market is the market where stocks are traded, also known as a stock exchange.}

[0083] Step 2.3 Due to If empty, proceed to step 5.

[0084] Step 5: Generate the final answer

[0085] Dictionary The prompt is embedded in the user's question "What is the difference between stocks and the stock market?" and then fed into the deepseek-v3 model to generate the final answer.

[0086] The final answer is: "Stocks refer to shares issued by a company, representing a portion of the company's equity; the stock market is a market for trading stocks, also called a stock exchange. The difference is that stocks are financial assets, while the stock market is a place where stocks are traded."

[0087] Example 2: The word does not exist in the concept library, but a matching synonym exists.

[0088] Step 1: User Issue Preprocessing

[0089] Step 1.1: Receive the user's question "What are bonds?";

[0090] The question is broken down into text fragments: contents = [“What is a bond?”].

[0091] Step 1.2: Use the deepseek-v3 model to extract concepts and obtain a concept list. ={bonds}.

[0092] Step 1.3: Use the deepseek-v3 model to determine which concepts belong to the financial domain, and obtain a filtered list of concepts. ={bonds}.

[0093] Step 2: Query the concept in the database

[0094] Step 2.1: Check if "bonds" are in the database; "bonds" are not in the database.

[0095] Step 2.2: No change Empty.

[0096] Step 2.3 Due to If not empty, proceed to step 3.

[0097] Step 3: Synonym Judgment

[0098] Step 3.1: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require ={bonds} is divided into multiple lists. Since there is only one concept, the list is {bonds}.

[0099] Step 3.2: Divide the concepts in the database into multiple lists for comparison. Assume the database has synonyms such as "securities" and "notes".

[0100] Step 3.3: Perform a synonym check between "bonds" and "securities" and "notes" in the database;

[0101] The deepseek-v3 model determines that "bonds" and "notes" are synonymous.

[0102] Step 3.4: Transfer the "bonds" from Remove and add "bonds" to the dictionary. And add the synonym "bond" to the "notes" synonym set in the database;

[0103] ={Bonds: A type of fixed-income financial instrument that provides investors with periodic interest payments after purchase.}

[0104] Step 3.5: Due to The page is empty. Skip step 4 and proceed to step 5.

[0105] Step 5: Generate the final answer

[0106] Dictionary The prompt word is embedded in the user's question "What is a bond?" and then fed into the deepseek-v3 model to generate the final answer.

[0107] The final answer: "Bonds are a common investment tool, essentially a 'promissory note.' When you buy bonds, you are essentially lending money to the issuer (such as a government or corporation), who promises to pay interest as agreed in the future and repay the principal at maturity. Bonds are generally considered a low-risk, stable-return investment suitable for investors seeking long-term returns."

[0108] Example 3: There are no related concepts or synonyms in the concept base.

[0109] Step 1: User Issue Preprocessing

[0110] Step 1.1: Receive user questions "What is quantitative trading?";

[0111] The question is split into text fragments: contents = [“What is quantitative trading?”].

[0112] Step 1.2: Use the deepseek-v3 model to extract concepts and obtain a concept list. ={Quantitative Trading}.

[0113] Step 1.3: Use the deepseek-v3 model to determine which concepts belong to the financial domain, and obtain a filtered list of concepts. ={Quantitative Trading}.

[0114] Step 2: Query the concept in the database

[0115] Step 2.1: Check if "Quantitative Trading" is in the database; "Quantitative Trading" is not in the database.

[0116] Step 2.2: No change Empty.

[0117] Step 2.3 Due to If not empty, proceed to step 3.

[0118] Step 3: Synonym Judgment

[0119] Step 3.1: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require ={Quantitative Trading} is divided into multiple lists. Since there is only one concept, the list is {Quantitative Trading}.

[0120] Step 3.2: Divide the concepts in the database into multiple lists for comparison.

[0121] Step 3.3: Perform a synonym comparison between "bonds" and the concepts in the database; the deepseek-v3 model determines that there are no synonyms.

[0122] Step 3.4: No change Empty.

[0123] Step 3.5: Due to If not empty, proceed to step 4.

[0124] Step 4: Online search

[0125] Step 4.1: Use an online search tool to retrieve relevant web page content; use the online tool to retrieve relevant web pages and segment the text.

[0126] Step 4.2: Traverse the webpage content and use the deepseek-v3 model to determine the definition of "quantitative trading";

[0127] The relevant definition is: "Quantitative trading is an investment method that uses computer technology to automatically execute trading decisions by establishing mathematical models."

[0128] Step 4.3: Add the definition to the dictionary And update the database;

[0129] = {Quantitative Trading: Quantitative trading is an investment method that uses computer technology to automatically execute trading decisions by establishing mathematical models.}

[0130] Step 5: Generate the final answer

[0131] Dictionary The prompt word is embedded in the user's question "What is quantitative trading?" and then fed into the deepseek-v3 model to generate the final answer.

[0132] The final answer: "Quantitative trading is a method that uses mathematical models and computer programs to develop and execute investment strategies."

[0133] Unlike traditional trading methods that rely on intuition and experience, quantitative trading uses data analysis, statistics, and algorithms to quantify market behavior into models, automatically determining buying and selling opportunities, thereby achieving more efficient and rational investment operations. This approach is particularly common in high-frequency trading and large institutional investing.

[0134] Example 2

[0135] like Figure 2 As shown, in a second aspect, the present invention discloses a financial concept question-and-answer system 20 based on a dynamically expanded personal wealth management standard concept library, including a financial concept extraction module 21, a financial concept retrieval module 22, a financial concept addition module 23, and a question-and-answer module 24.

[0136] The financial concept extraction module 21 is responsible for identifying potential financial concepts from input documents or user-submitted questions; that is, extracting financial concepts from user questions through the financial concept extraction module; including:

[0137] The financial concept recognition unit 211 uses prompts and a language model to identify financial concepts in user input, such as "which investors are suitable for stocks?" Here, "stocks" and "investors" are financial concepts.

[0138] The financial concept screening unit 212 receives the recognition results from the financial concept identification unit and filters out concepts that truly belong to the financial field, ensuring that the identified concepts belong to the financial field and preventing non-financial concepts such as flow, risk, and variance from being included when the user's question is too long.

[0139] The financial concept retrieval module 22 receives concept information extracted by the financial concept extraction module and performs concept definition retrieval. The structured data retrieval unit 221 of this module checks whether each concept in the concept list described in step 2 exists in the database. The concept semantic expansion unit 222 of this module determines whether synonyms exist in the database as described in step 3. The network retrieval unit 224 of this module retrieves concept definitions online as described in step 4. Specifically:

[0140] The structured data retrieval unit 221 is specifically designed to perform precise retrieval in structured databases (such as relational databases) to identify known definitions related to the input concept;

[0141] The concept semantic extension unit 222 is used to determine whether there are synonyms corresponding to the input concept in the structured database. Specifically, it determines whether there are words that are different in language, have different translations, are abbreviations, or can be substituted for each other.

[0142] The unstructured data retrieval unit 223 searches the document input by the user to find the definition corresponding to the input concept, so as to ensure that the concept definition is fully extracted from text information from different sources.

[0143] Network retrieval unit 224, when the aforementioned retrieval unit cannot provide a concept definition, accesses the internet to search for more concept information. If the internet resources still cannot find a clear definition, the system will generate a reasonable concept definition based on the existing information. After successfully retrieving a concept definition, all concepts retrieved by related modules or units will be removed from the search list and will no longer participate in subsequent search processes, thereby effectively reducing duplicate queries and improving the overall response speed of the system.

[0144] The newly added financial concept module 23 is used in step 4.3 to add concepts that were not found in the relational database in the financial concept retrieval module and the definitions of the concepts obtained by online retrieval in step 4 to the database.

[0145] The question-answering module 24 receives the definitions of the concepts retrieved by the above modules, constructs prompt words, and finally generates the answer to the question.

[0146] The modules in the aforementioned financial concept question-and-answer system based on a dynamically expanding personal wealth management standard concept base can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, allowing the processor to invoke and execute the corresponding operations. By using this system, the aforementioned computational processing methods can be executed, and the corresponding technical effects can be achieved.

[0147] Example 3

[0148] This invention also provides a computer device, including a memory and a processor. Specifically, the memory stores data and related computer programs required for executing a method, and the processor can call the data and programs in the memory to implement all the steps of the aforementioned method. The computer device also includes input / output interfaces (I / O), communication interfaces, an audio processor, a display, a power supply, and other components. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces; these are well known to those skilled in the art and therefore will not be described further. The processor of this computer device provides computational and control capabilities. The memory of this computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of this computer device stores data such as financial concepts.

[0149] Example 4

[0150] In a fourth aspect, this invention discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements a method for answering financial concept questions. This storage medium can be any readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk. For example, the computer-readable storage medium can be the aforementioned memory including program instructions, which can be executed by the processor of the financial concept question-and-answer system 20 to complete the aforementioned financial concept question-and-answer method.

[0151] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0152] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A financial concept question-and-answer method based on a dynamically expanding personal wealth management standard concept base, characterized in that... Includes the following steps: Step 1: User question preprocessing, including question segmentation and concept extraction; Step 2: For each preprocessed concept To check if it already exists in the database; Step 3: For those still in Step 2 Each concept in It performs synonym comparison with concepts in the database; Step 4: For those still in step 3 Each concept in To retrieve the corresponding explanations of the concepts through online searches; Step 5: Finally, The final prompt is generated by embedding prompts into the question and then handed over to the deepseek-v3 model to generate the final answer.

2. The financial concept question-and-answer method based on a dynamically expanded personal wealth management standard concept library according to claim 1, characterized in that, Step 1 includes: Step 1.1: Receive user questions. To prevent excessively long user input, the user input is split into segments, generating a list of text fragments: contents; Step 1.2: Traverse the text fragment list `contents`, embed the `content` into the prompt words, and call the deepseek-v3 model to extract concepts, obtaining a concept list. ; Step 1.3: List the concepts The suggestion words are embedded and the deepseek-v3 model is called to determine which concepts belong to the financial field, resulting in a filtered list of concepts. .

3. The financial concept question-and-answer method based on a dynamically expanded personal wealth management standard concept library according to claim 1, characterized in that, Step 2 includes: Step 2.1: If concept c belongs to a concept stored in the database or is a synonym of a concept stored in the database, then remove c from the database. Remove and add to dictionary ,in Explanation of concepts retrieved from the database; Step 2.2: If If there are still concepts that have not been removed, proceed to step 3; if If the value is empty, proceed to step 5.

4. The financial concept question-and-answer method based on a dynamically expanded personal wealth management standard concept library according to claim 1, characterized in that, Step 3 includes: Step 3.1: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] Split into multiple lists The number of partitions n is determined by The total number of concepts and the maximum number of concepts k1 after the hyperparameter segmentation are determined by the total number of concepts in the dataset. The specific formula is as follows: ; Step 3.2: Divide all concepts in the database into multiple lists. The number of lists *m* to be split is determined by the number of concepts *N* in the database and the maximum number of concepts *k2* after splitting, as set by the hyperparameter. The specific formula is as follows: ; Step 3.3: [The text appears to be incomplete and contains several grammatical errors. A more accurate translation would require the full context.] and The permutations and combinations of embedded prompts form n * m reasoning tasks, which are then fed into the deepseek-v3 model to generate a thesaurus. ; Step 3.4: Remove all concepts c that have synonyms from... Remove and add to dictionary And append the synonym c to the corresponding synonym set in the database of c'; Step 3.5: If If there are still concepts that have not been removed, proceed to step 4; if If empty, proceed to step 5.

5. The financial concept question-and-answer method based on a dynamically expanded personal wealth management standard concept library according to claim 1, characterized in that, Step 4 includes: Step 4.1: For each filtered concept The system uses an online search tool to retrieve the content W of the webpage that ranks first in relevance among the relevant webpages. It then recursively divides the content W into a list of texts of size W_max. To prevent the size of the prompt word from exceeding the window size of the large language model; Step 4.2: Traversal The deepseek-v3 model determines whether it contains [the specified content]. The definition and output of the result in JSON format includes whether it was generated by a large language model and the definition of the concept; if it was not generated by a large language model, the iteration ends, because this means the model is in W. i Found in Definition; Step 4.3: Put all Add the corresponding explanations obtained in step 4.2 to the dictionary. And update the database.

6. A financial concept question-and-answer system based on a dynamically expanding personal wealth management standard concept library, characterized in that... It includes a financial concept extraction module, a financial concept retrieval module, a financial concept addition module, and a question and answer module; The financial concept extraction module is responsible for identifying potential financial concepts from input documents or user-submitted questions. The financial concept retrieval module receives concept information extracted by the financial concept extraction module and performs a retrieval of concept definitions. The aforementioned financial concept addition module is used to add concepts from the financial concept retrieval module that are not found in the relational database to the database. The question-and-answer module receives the definitions of concepts retrieved by the above module and generates answers to the questions.

7. The financial concept question-and-answer system based on a dynamically expanding personal wealth management standard concept library according to claim 6, characterized in that, The financial concept extraction module also includes: The financial concept recognition unit uses prompts and a language model to identify financial concepts in user input. The financial concept screening unit receives the recognition results from the financial concept identification unit and filters out concepts that truly belong to the financial field, ensuring that the identified concepts belong to the financial field.

8. The financial concept question-and-answer system based on a dynamically expanding personal wealth management standard concept library according to claim 6, characterized in that, The financial concept retrieval module includes: The structured data retrieval unit is specifically designed for precise retrieval in structured databases to identify known definitions related to the input concept; The concept semantic extension unit is mainly responsible for determining whether there are synonyms for the input concept in the structured database; The unstructured data retrieval unit is responsible for retrieving the definitions corresponding to concepts from the documents input by the user; The network retrieval unit is responsible for accessing the Internet to retrieve more conceptual information when the above retrieval units cannot provide a conceptual definition. If the Internet resources still cannot find a clear definition, the system will generate a reasonable conceptual definition based on the existing information.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 5.

Images