Personal financial service intelligent question and answer method, device, equipment, medium and product

Abstract

The invention provides a personal financial service intelligent question-answering method and device, equipment, a medium and a product, and relates to the field of big data. The method comprises the steps that keywords in user questions are determined, context information of the user questions and the keywords is extracted, the context information comprises semantic features and time sequence features, classification of the user questions is determined according to the context information, the classification of the user questions is personal financial services or non-personal financial services, and the user questions are classified according to the classification of the user questions. Under the condition that the user question is classified as the personal financial business, extracting a first number of pieces of personal financial business knowledge related to the user question from a preset personal financial business knowledge base, and inputting the user question and the first number of pieces of personal financial business knowledge into a personal financial business large model, obtaining an answer output by the personal financial business large model; according to the method, the efficiency and accuracy of obtaining related information under the personal financial business by the user are improved.

Description

Technical Field

[0001] The present application relates to the field of big data, and in particular to a method, device, equipment, medium and product for intelligent question answering of personal financial services. Background Art

[0002] In recent years, personal financial services have shown a booming trend, and its diverse products and services have attracted the attention and active participation of a large number of users. As users gain a deeper understanding of personal financial services and their demand increases, the demand for inquiries related to personal finance has also increased significantly. These issues may involve multiple aspects such as financial management, loans, insurance, etc. Users hope to obtain accurate and timely information through inquiries to guide their financial decisions.

[0003] In the prior art, when users query questions related to personal finance, they mainly rely on two traditional methods: one is to manually read through thick financial books, reports or policy documents and other materials; the other is to manually enter keywords in the search engine to search and get answers to questions.

[0004] However, both methods have obvious shortcomings. Manual browsing of materials is not only time-consuming and laborious, but also inefficient, and it is difficult to find the required information in a short time. Although manual search engine retrieval is relatively convenient, it is limited by the search engine algorithm and the user's keyword selection, and the search results are often not accurate enough, and may even be misleading due to subjective judgment and uncertainty. These problems not only affect the efficiency of users in obtaining information, but also reduce the accuracy of business and customer satisfaction. Summary of the invention

[0005] The present application provides a method, device, equipment, medium and product for intelligent question answering of personal financial services, which are used to solve the technical problems of low efficiency and insufficient accuracy in users obtaining relevant information under personal financial services.

[0006] In a first aspect, the present application provides a personal financial service intelligent question-answering method, comprising:

[0007] Determine keywords in the user question, and extract context information of the user question and the keywords, wherein the context information includes semantic features and temporal features;

[0008] Determining a classification of the user question according to the context information, wherein the classification of the user question is personal financial service or non-personal financial service;

[0009] In the case where the user question is classified as a personal financial service, extracting a first number of pieces of personal financial service knowledge related to the user question from a preset personal financial service knowledge base;

[0010] The user question and the first number of pieces of personal financial business knowledge are input into a personal financial business big model to obtain an answer output by the personal financial business big model, wherein the personal financial business big model is obtained by pre-training a big language model with personal financial business question and answer samples.

[0011] Optionally, determining keywords in the user's question includes:

[0012] Using a language technology platform model to segment the user's question and perform part-of-speech tagging on the segmentation results;

[0013] The words marked as nouns in the word segmentation results are determined as the keywords.

[0014] Optionally, extracting context information of the user question and the keyword includes:

[0015] Using a word frequency-inverse text frequency algorithm to determine the criticality of each of the keywords, and sorting the keywords in descending order of criticality;

[0016] Splicing the user question and the sorted keywords into a question description text;

[0017] Inputting the problem description text into a bidirectional encoder language model to obtain hidden layer information;

[0018] The hidden layer information is input into a bidirectional long short-term memory network model to obtain the context information.

[0019] Optionally, extracting a first number of pieces of personal financial service knowledge related to the user's question from a preset personal financial service knowledge base includes:

[0020] Determine a first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base using a hash similarity algorithm, and determine a second number of pieces of personal financial service knowledge in the personal financial service knowledge base that are ranked first according to the first similarity;

[0021] Determining first semantic features of the user question and the keyword, and second semantic features of each piece of personal financial business knowledge in the second number of pieces of personal financial business knowledge;

[0022] According to the second similarity between the first semantic feature and each of the second semantic features, the first number of pieces of personal financial service knowledge that are ranked first according to the second similarity among the second number of pieces of personal financial service knowledge are determined.

[0023] Optionally, the using a hash similarity algorithm to determine the first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base includes:

[0024] Determine a first hash code value of the user question and a second hash code value of each piece of personal financial business knowledge, wherein the number of digits of the first hash code value and the second hash code value is a preset number of digits;

[0025] An XOR operation is performed on the first Hash code value and the second Hash code value, and the number of 1s in the XOR operation result is divided by the preset number of bits to obtain the first similarity.

[0026] Optionally, each piece of personal financial service knowledge in the personal financial service knowledge base includes a title and a knowledge point, and the method further includes:

[0027] Determining a first hash code value of the user question includes:

[0028] Hash-encode the keywords of the user question to obtain a first hash-encoded value;

[0029] Determine the second hash code value of each piece of personal financial business knowledge, including:

[0030] Keywords are extracted from the titles of each piece of personal financial business knowledge, and hash codes are performed on the keywords of the titles of each piece of personal financial business knowledge to obtain second hash code values ​​corresponding to each piece of personal financial business knowledge.

[0031] Optionally, the method further includes:

[0032] In the case where the user question is classified as non-personal financial business, the user question is input into the large language model to obtain an answer output by the large language model.

[0033] In a second aspect, the present application provides an intelligent question-answering device for personal financial services, comprising:

[0034] A determination module is used to determine the keywords in the user's question;

[0035] A processing module, used to extract context information of the user question and the keyword, wherein the context information includes semantic features and temporal features;

[0036] The determination module is further used to determine the classification of the user question according to the context information, wherein the classification of the user question is personal financial service or non-personal financial service;

[0037] The processing module is further configured to extract a first number of pieces of personal financial service knowledge related to the user question from a preset personal financial service knowledge base when the user question is classified as a personal financial service;

[0038] The processing module is further used to input the user question and the first number of personal financial business knowledge into a personal financial business big model to obtain an answer output by the personal financial business big model, wherein the personal financial business big model is obtained by pre-training a big language model using personal financial business question and answer samples.

[0039] Optionally, the processing module is further used to segment the user question using a language technology platform model and to perform part-of-speech tagging on the segmentation results;

[0040] The determination module is further configured to determine the words marked as nouns in the word segmentation results as the keywords.

[0041] Optionally, the determination module is further configured to determine the criticality of each of the keywords using a word frequency-inverse text frequency algorithm;

[0042] The processing module is further used to sort the keywords in descending order of criticality;

[0043] The processing module is further used to splice the user question and the sorted keywords into a question description text;

[0044] The determination module is further used to input the problem description text into a bidirectional encoder language model to obtain hidden layer information;

[0045] The determination module is also used to input the hidden layer information into a bidirectional long short-term memory network model to obtain the context information.

[0046] Optionally, the determination module is further configured to use a hash similarity algorithm to determine a first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base;

[0047] The determination module is further used to determine a second number of pieces of personal financial service knowledge ranked first according to the first similarity in the personal financial service knowledge base;

[0048] The determination module is further used to determine the first semantic features of the user question and the keyword, and the second semantic features of each piece of personal financial business knowledge in the second number of pieces of personal financial business knowledge;

[0049] The determination module is further configured to determine, according to a second similarity between the first semantic feature and each of the second semantic features, the first number of pieces of personal financial service knowledge that are ranked first according to the second similarity among the second number of pieces of personal financial service knowledge.

[0050] Optionally, the determination module is further used to determine a first hash code value of the user question and a second hash code value of each piece of personal financial business knowledge, wherein the number of digits of the first hash code value and the second hash code value is a preset number of digits;

[0051] The processing module is further configured to perform an XOR operation on the first hash code value and the second hash code value;

[0052] The determination module is further configured to divide the number of 1s in the XOR operation result by the preset number of bits to obtain the first similarity.

[0053] Optionally, the determining module is further configured to determine the first hash code value of the user question, including:

[0054] Hash-encode the keywords of the user question to obtain a first hash-encoded value;

[0055] The determination module is further used to determine the second hash code value of each piece of personal financial business knowledge, including:

[0056] Keywords are extracted from the titles of each piece of personal financial business knowledge, and hash codes are performed on the keywords of the titles of each piece of personal financial business knowledge to obtain second hash code values ​​corresponding to each piece of personal financial business knowledge.

[0057] Optionally, the determination module is further configured to, when the user question is classified as non-personal financial business, input the user question into the large language model to obtain an answer output by the large language model.

[0058] In a third aspect, an embodiment of the present application provides a personal financial service intelligent question-answering device, including: a memory, a processor;

[0059] The memory stores computer-executable instructions;

[0060] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.

[0061] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementations of the first aspect.

[0062] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0063] The present application provides an intelligent question-answering method for personal financial services, which determines keywords in user questions and extracts context information of the user questions and keywords, wherein the context information includes semantic features and temporal features, and determines the classification of the user questions according to the context information, wherein the user questions are classified as personal financial services or non-personal financial services. When the user questions are classified as personal financial services, a first number of pieces of personal financial services knowledge related to the user questions are extracted from a preset personal financial services knowledge base, and the user questions and the first number of pieces of personal financial services knowledge are input into a personal financial services big model to obtain answers output by the personal financial services big model, wherein the personal financial services big model is obtained by pre-training a big language model using personal financial services question-answering samples. The method improves the efficiency and accuracy of users in obtaining relevant information under personal financial services. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0065] Figure 1 Schematic diagram of the process of the intelligent question-answering method for personal financial services provided in this application Figure 1 ;

[0066] Figure 2 Schematic diagram of the process of the intelligent question-answering method for personal financial services provided in this application Figure 2 ;

[0067] Figure 3 A framework diagram for determining the classification of user questions provided in this embodiment;

[0068] Figure 4 Schematic diagram of the process of the intelligent question-answering method for personal financial services provided in this application Figure 3 ;

[0069] Figure 5 The overall process structure diagram of the intelligent question-answering method for personal financial services provided in this application;

[0070] Figure 6 A schematic diagram of the structure of the intelligent question-answering device for personal financial services provided in this application;

[0071] Figure 7 This is a schematic diagram of the structure of the intelligent question-and-answer device for personal financial services provided in this application.

[0072] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0073] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0074] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with the relevant laws, regulations and standards of the relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0075] In addition, this application involves conducting big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.), and using artificial intelligence technology to make automated decisions, and making technical solutions that have a significant impact on personal rights and interests based on the results of automated decisions. The application provides users with corresponding operation entrances for them to choose to agree or reject the results of automated decisions; if the user chooses to reject, the expert decision-making process will be entered.

[0076] It should be noted that the intelligent question-answering method, device, equipment, medium and product for personal financial services provided in this application can be used in the big data field or the financial technology field, and can also be used in any field other than the big data field or the financial technology field. The application field of the intelligent question-answering method, device, equipment, medium and product for personal financial services in this application is not limited.

[0077] In recent years, personal financial services have flourished, attracting the attention and participation of a large number of users, resulting in a significant increase in the demand for inquiries related to personal finance.

[0078] In the prior art, when users inquire about issues related to personal finance, they mainly rely on manually reading through thick materials or manually entering keywords into search engines to search for answers to their questions.

[0079] However, the manual method of searching for answers to questions not only consumes a lot of time and energy, is inefficient, and makes it difficult to quickly locate the required information, but also the subjective judgment and uncertainty of manual retrieval greatly reduce the accuracy of the retrieval results, which in turn affects the accuracy of the business and customer satisfaction.

[0080] In response to the above problems, the personal financial business intelligent question and answer method provided by the present application determines the keywords in the user's question and extracts the context information of the user's question and the keywords, determines the classification of the user's question based on the context information, and when the user's question is classified as personal financial business, extracts the top N personal financial business knowledge related to the user's question from a preset personal financial business knowledge base, inputs the user's question and the top N personal financial business knowledge into the personal financial business big model, and obtains the answer output by the personal financial business big model; this method improves the efficiency and accuracy of users in obtaining relevant information under personal financial business.

[0081] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0082] Figure 1 Schematic diagram of the process of the intelligent question-answering method for personal financial services provided in this application Figure 1 The execution subject of this embodiment is, for example, an intelligent question-answering management system. Figure 1 As shown, the personal financial service intelligent question-answering method shown in this embodiment includes:

[0083] S101: Determine keywords in a user question and extract context information of the user question and the keywords, wherein the context information includes semantic features and temporal features.

[0084] Among them, user questions are used to indicate the questions, demands or feedback raised by users during the use of products or services. They contain the core problems that users hope to solve, the answers they expect or the specific content of improvements they expect. They are the basis for understanding user intentions, optimizing product functions and improving user experience.

[0085] Keywords in user questions refer to core words that can reflect the user's concerns, questions or needs in the user questions.

[0086] The context information of user questions and keywords indicates that in addition to the directly identified keywords, the specific context of the keywords should also be considered when parsing user questions or needs, including the semantic association between words (i.e. semantic features, such as synonyms, antonyms, hyponyms, and hyponyms) and the time sequence and periodicity of events or behaviors (i.e. time sequence features, such as sequence, time interval, periodic changes, etc.). Together, this information constitutes a comprehensive understanding of user questions and keywords, which helps to obtain accurate user intentions.

[0087] Keywords in user questions can directly reflect the core points of user questions, while contextual information provides richer background knowledge, which helps the intelligent question-answering management system to accurately understand user intentions. In particular, semantic features and temporal features can reveal the association between words and the order of events, which are of great significance for understanding and answering complex questions. By accurately extracting keywords in user questions and the contextual information of user questions and keywords, the intelligent question-answering management system can answer user questions more effectively and improve user experience and satisfaction.

[0088] S102: Determine the classification of the user question according to the context information, wherein the classification of the user question is personal financial service or non-personal financial service.

[0089] Among them, personal financial services include: personal savings, loans, insurance and other personal financial management matters.

[0090] Non-personal financial services include: corporate finance, financial market analysis, macroeconomic policies and other economic and financial fields.

[0091] By analyzing user questions, keywords in user questions, the contextual information relationship between user questions and keywords, as well as background information such as user identity and user historical behavior, user questions can be accurately classified as personal financial services or non-personal financial services.

[0092] First, analyze user questions and identify the keywords of user questions. The keywords of user questions directly reflect the core needs and concerns of users. For example, if a user mentions "housing loan", "credit card repayment" or "personal finance", these keywords are likely to point to personal financial services. On the contrary, if the keywords involve "corporate loans", "financial market analysis" or "economic policy interpretation", they may fall into the category of non-personal financial services.

[0093] Secondly, the specific meaning and relationship of the contextual information of the user's question and keywords. Keywords may not be sufficient to determine the classification of the question, but combined with contextual information, the user's intention can be understood more accurately. For example, if the user mentions "I want to learn how to apply for a loan for my business", although "loan" is a common financial term, combined with the word "business" in the context, it can be judged that this is a non-personal financial business question.

[0094] In addition, user identity and historical behavior are also the basis for classification. For example, if the user is a known individual consumer, the questions they ask are more likely to involve personal financial services. If the user is a business owner, financial analyst, or economist, their questions may be more inclined to non-personal financial services. At the same time, by analyzing the user's historical behavior data, their financial needs and preferences can be understood, so as to more accurately classify their questions.

[0095] Personal financial services and non-personal financial services are significantly different in terms of demand characteristics, service processes, and professional knowledge requirements. Accurate classification can ensure that users are provided with more personalized, professional, and efficient services. At the same time, it can also help financial institutions allocate resources more effectively, optimize service processes, and improve overall operational efficiency.

[0096] S103: When the user question is classified as personal financial service, extract a first number of pieces of personal financial service knowledge related to the user question from a preset personal financial service knowledge base.

[0097] Among them, the preset personal financial business knowledge base is used to indicate a large amount of regulatory knowledge, operating procedures and guidance documents related to personal financial business.

[0098] The first number of pieces of personal financial business knowledge may be, for example, the first three pieces of personal financial business knowledge related to the user's question in a preset personal financial business knowledge base.

[0099] For example, if a user asks about "interest rates and repayment methods for personal housing loans", then "personal housing loans", "interest rates" and "repayment methods" will become keywords. Subsequently, the intelligent question and answer management system will automatically start the knowledge base search engine, enter these keywords, and set search conditions, such as relevance sorting, knowledge type (such as regulatory knowledge, operating procedures, guidance documents, etc.) and the newness of knowledge (the latest or most relevant knowledge is displayed first). Based on these conditions, the intelligent question and answer management system will automatically filter out the knowledge points that are most relevant to the user's question from the knowledge base. Next, the intelligent question and answer management system will automatically select the top three personal financial business knowledge that are most closely related to the user's question based on the sorting of the search results. These knowledge points may include the interest rate policy for personal housing loans, specific repayment method introductions, and related operating guidelines or regulatory interpretations.

[0100] S104: Input the user question and the first number of pieces of personal financial business knowledge into a personal financial business big model to obtain an answer output by the personal financial business big model, wherein the personal financial business big model is obtained by pre-training a big language model with personal financial business question and answer samples.

[0101] Among them, the Large Language Models (LLM) is used as a guiding framework to build and optimize the intelligent question and answer model for financial business. The personal financial business large model is a specialized model obtained by further training samples using a large number of personal financial question and answer pairs based on the large language model. It can more accurately understand and answer relevant questions in the field of personal finance.

[0102] First, the user question and the first number of personal financial business knowledge are input into the personal financial business big model; then, the personal financial business big model will use the knowledge and skills it has learned in the training phase to deeply understand and analyze the input information; finally, the personal financial business big model will generate one or more possible answer options, and select one or more answers that best match the user question and the first number of personal financial business knowledge as output.

[0103] The personal financial business intelligent question-answering method provided in this embodiment determines the keywords in the user's question and extracts the context information of the user's question and the keywords, wherein the context information includes semantic features and temporal features, and determines the classification of the user's question according to the context information, wherein the user's question is classified as personal financial business or non-personal financial business. When the user's question is classified as personal financial business, a first number of personal financial business knowledge related to the user's question is extracted from a preset personal financial business knowledge base, and the user's question and the first number of personal financial business knowledge are input into a personal financial business big model to obtain an answer output by the personal financial business big model, wherein the personal financial business big model is obtained by pre-training a big language model using personal financial business question and answer samples; the method improves the efficiency and accuracy of users in obtaining relevant information under personal financial business.

[0104] Figure 2 Schematic diagram of the process of the intelligent question-answering method for personal financial services provided in this application Figure 2 .like Figure 2 As shown, in this embodiment Figure 1 Based on the embodiment, how to determine the classification of user problems is described in detail. The method for determining the classification of user problems shown in this embodiment includes:

[0105] S201: Use the language technology platform model to segment the user's questions and perform part-of-speech tagging on the segmentation results.

[0106] Among them, the Language Technology Platform model (LTP) is a model that integrates multiple Chinese language processing functions such as word segmentation, part-of-speech tagging, syntactic analysis, named entity recognition, word sense disambiguation and semantic role labeling.

[0107] The word segmentation result is the output of the process of dividing user questions into independent and meaningful word units, which is expressed as a word sequence.

[0108] First, the user's question is input into the intelligent question and answer management system; then, the intelligent question and answer management system calls the language technology platform model to preprocess the text, including removing stop words, punctuation marks, etc.; then, the language technology platform model uses the built-in word segmentation algorithm to cut the text into independent words; then, each word is tagged with the part of speech, and its part of speech is determined based on the context information and the characteristics of the vocabulary itself; finally, the language technology platform model outputs the word segmentation and its part of speech tagging results.

[0109] The word segmentation and part-of-speech tagging of the language technology platform model are the basic links in natural language processing, and are crucial to improving the accuracy and efficiency of text analysis. Word segmentation can cut continuous strings into words with clear meanings, which is the first step to understand the meaning of the text. Part-of-speech tagging further gives grammatical functions to each word, which helps to identify sentence structure, extract key information, and conduct deeper language understanding and analysis. Especially in complex information processing tasks, such as intelligent question and answer, sentiment analysis, machine translation, etc., the accuracy of word segmentation and part-of-speech tagging is directly related to the performance of the entire system and user experience.

[0110] S202: Determine the words marked as nouns in the word segmentation results as keywords.

[0111] Understandably, the language technology platform model is used to tag each word in the word segmentation results with a part-of-speech tag; all words marked as nouns are screened out and used as keywords. In this process, the accuracy of word segmentation and part-of-speech tagging needs to be ensured to ensure the effectiveness and reliability of the extracted keywords.

[0112] Nouns usually carry information about specific entities or abstract concepts in a language, and they are often the core objects or topics discussed in a text. By extracting nouns as keywords, we can more accurately grasp the theme and key points of a text, which is helpful for tasks such as text analysis, information retrieval, or knowledge extraction. In addition, nouns as keywords have good stability and versatility, and are not easily affected by contextual changes. Therefore, they have high practicality and reliability in practical applications.

[0113] S203: Using a word frequency-inverse text frequency algorithm to determine the criticality of each keyword, and sorting the keywords in descending order of criticality.

[0114] Among them, the Term Frequency-Inverse Document Frequency (TF-IDF) algorithm measures the importance of each keyword in the user's question by calculating the product of the frequency of each keyword in the current user's question and the rarity of the keyword in the historical user's question. In other words, keywords that appear frequently in user questions and rarely appear in other user questions have higher importance.

[0115] The formula for calculating the criticality of keywords in user questions using the word frequency-inverse text frequency algorithm is:

[0116]

[0117] in, For keywords User issues The criticality in For keywords User issues The word frequency in For keywords In all historical user questions The rarity of For keywords In user questions The number of times it appears in For user questions The total number of words, For all historical user questions The total number of For all historical user questions Contains keywords The number of user questions.

[0118] The higher the value, the more keywords In user questions and less frequently in other user questions, so it is important for this user question. The importance is also higher.

[0119] The word frequency-inverse text frequency algorithm uses the key degree calculation formula of the keywords in the user's question to calculate the key degree of each keyword in the user's question in turn, and sorts each keyword in descending order according to the key degree of each keyword.

[0120] S204: The user question and the sorted keywords are concatenated into a question description text.

[0121] S205: Input the question description text into the bidirectional encoder language model to obtain hidden layer information.

[0122] S206: Input the hidden layer information into the bidirectional long short-term memory network model to obtain context information.

[0123] Among them, the bidirectional encoder language model (for example: RoBERTa-WWM, Robustly Optimized BERTPretraining Approach-Whole Word Masking) is an enhanced version of the BERT model (Bidirectional Encoder Representations from Transformers), which improves language understanding capabilities by training for a longer time on large-scale datasets and adopting a whole-word masking strategy.

[0124] The Bidirectional Long Short-Term Memory (BiLSTM) model is an improved recurrent neural network specifically designed for processing and predicting sequence data. It consists of two independent long short-term memory networks, one that processes the sequence from front to back and the other that processes the sequence from back to front. This bidirectional structure allows the BiLSTM to consider both the contextual information at each time step, thereby more comprehensively capturing the dependencies and semantic features in the sequence.

[0125] First, the user question is concatenated with the sorted keywords to form a question description text to better capture the user's intention. Then, the question description text is input into the bidirectional encoder language model, which generates hidden layer information through its deep Transformer architecture. The hidden layer information contains the deep semantics and contextual features of the text. Subsequently, the hidden layer information is input into the bidirectional long short-term memory network, which further analyzes the question description text through its bidirectional processing capabilities, combines the contextual information, and generates richer contextual information.

[0126] S207: Determine the classification of the user question according to the context information, wherein the classification of the user question is personal financial service or non-personal financial service.

[0127] Among them, step S207 is similar to step S102 and will not be repeated here.

[0128] Figure 3 A framework diagram for determining the classification of user questions provided in this embodiment. Figure 3To explain in detail:

[0129] First, the language technology platform model is used to segment and tag user questions, and the words marked as nouns are determined as keywords. Then, the word frequency-inverse text frequency algorithm is used to sort the keywords to obtain the sorted keywords. The user questions and the sorted keywords are concatenated to obtain the question description text. Then, the question description text is input into the neural network model (the neural network model includes: bidirectional encoder language model and bidirectional long short-term memory network model) to obtain the classification of user questions.

[0130] The personal financial service intelligent question-answering method provided in this embodiment uses a language technology platform model to segment user questions and perform part-of-speech tagging on the segmentation results, determines the words marked as nouns in the segmentation results as keywords, uses a word frequency-inverse text frequency algorithm to determine the criticality of each keyword, sorts the keywords in descending order of criticality, concatenates the user question and the sorted keywords into a question description text, inputs the question description text into a bidirectional encoder language model to obtain hidden layer information, inputs the hidden layer information into a bidirectional long short-term memory network model to obtain context information, and determines the classification of the user question based on the context information, wherein the user question is classified into personal financial service or non-personal financial service; the method improves the intelligence and accuracy of user question processing.

[0131] Figure 4 Schematic diagram of the process of the intelligent question-answering method for personal financial services provided in this application Figure 3 .like Figure 3 As shown, in this embodiment Figure 1 Based on the embodiment, the question-answering process of the personal financial service is described in detail. The question-answering process of the personal financial service shown in this embodiment includes:

[0132] S301: When the user question is classified as personal financial service, a hash similarity algorithm is used to determine a first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base, and a second number of pieces of personal financial service knowledge in the personal financial service knowledge base that are ranked first according to the first similarity are determined.

[0133] Among them, the hash similarity algorithm is a calculation method that determines the similarity between the user question and each piece of personal financial business knowledge in the personal financial business knowledge base by comparing the hash value of the user question with the hash value of each piece of personal financial business knowledge in the personal financial business knowledge base.

[0134] When the classification of the user's question is determined to be personal financial business, a hash similarity algorithm can be used to efficiently and accurately find the personal financial business knowledge that is most relevant to the user's question. By comparing the first similarity between the hash value of the user's question and the hash value of each piece of personal financial business knowledge in the personal financial business knowledge base, the higher the first similarity, the higher the correlation between the user's question and each piece of personal financial business knowledge in the personal financial business knowledge base. After sorting the personal financial business knowledge by the first similarity, a second quantity value is further determined, which represents the number of the top N most similar personal financial business knowledge items selected from the sorted personal financial business knowledge. For example, the top 128 personal financial business knowledge items. By selecting these top-ranked personal financial business knowledge items, a targeted and highly relevant personal financial business knowledge set can be provided to the user, thereby helping the user find answers or solutions to the problem faster.

[0135] Optionally, the specific implementation steps of using a hash similarity algorithm to determine the first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base are:

[0136] Determine a first hash code value of the user's question and a second hash code value of each piece of personal financial business knowledge, wherein the number of digits of the first hash code value and the second hash code value is a preset number of digits;

[0137] An XOR operation is performed on the first hash code value and the second hash code value, and the number of 1s in the XOR operation result is divided by a preset number of bits to obtain a first similarity.

[0138] The first similarity calculation formula is:

[0139]

[0140] in, is the similarity between the user's question and each piece of personal financial business knowledge in the personal financial business knowledge base, that is, the first similarity; The first hash code value of the user's question; is the second hash code value of each piece of personal financial business knowledge, The number of bits of the first hash code value and the second hash code value, that is, the preset number of bits; is an exclusive OR operation of the first hash code value and the second hash code value; is the number of 1s in the XOR operation result of the first hash code value and the second hash code value.

[0141] Optionally, each piece of personal financial business knowledge in the personal financial business knowledge base includes a title and a knowledge point, and the method further includes:

[0142] Determine the first hash code value of the user's question, including:

[0143] Hash-encode the keywords of the user's question to obtain a first hash-encoded value;

[0144] Determine the second hash code value of each piece of personal financial business knowledge, including:

[0145] Keywords are extracted from the titles of each piece of personal financial business knowledge, and hash codes are performed on the keywords of the titles of each piece of personal financial business knowledge to obtain second hash code values ​​corresponding to each piece of personal financial business knowledge.

[0146] S302: Determine the first semantic features of the user question and the keyword, and the second semantic features of each piece of personal financial business knowledge in the second number of pieces of personal financial business knowledge.

[0147] The S-BERT model (Sentence-Bidirectional Encoder Representations from Transformers) is used to map sentences or phrases into a high-dimensional vector space, where each vector represents the semantic features of a sentence or phrase. These feature vectors capture key information in a sentence or phrase, such as the topic, sentiment, and contextual relationships, so that they can accurately reflect the core intent of the user's question and the semantic connotation of the keyword.

[0148] Therefore, user questions and keywords can be input into the S-BERT model to obtain the first semantic feature; and each piece of personal financial business knowledge in the second number of pieces of personal financial business knowledge can be input into the S-BERT model to obtain the second semantic feature.

[0149] S303: Determine, according to the second similarity between the first semantic feature and each second semantic feature, a first number of pieces of personal financial service knowledge ranked first according to the second similarity among the second number of pieces of personal financial service knowledge.

[0150] The similarity calculation method in the vector space (such as cosine similarity, Euclidean distance, etc.) can be used to calculate the similarity or distance value between the first semantic feature and each second semantic feature. This value is the second similarity, which reflects the semantic proximity between the user question and keyword and each piece of personal financial business knowledge in the second number of pieces of personal financial business knowledge.

[0151] The second number of pieces of personal financial business knowledge may be sorted according to the second similarity. The principle of sorting is that the personal financial business knowledge with higher similarity is ranked higher, because they have a higher semantic relevance to the user's question and are more likely to be the content that the user wants to know or query, and the first number of pieces of personal financial business knowledge ranked higher in the second similarity are selected.

[0152] S304: Input the user question and the first number of pieces of personal financial business knowledge into the personal financial business big model to obtain an answer output by the personal financial business big model, wherein the personal financial business big model is obtained by pre-training a big language model with personal financial business question and answer samples.

[0153] The construction of the personal finance business model is based on a large number of personal finance business question and answer samples. The personal finance business question and answer samples are collected and annotated to cover various questions and answers that users may encounter in historical operations, ensuring that the model training can be close to the real application scenario.

[0154] When user questions and the first number of pieces of personal financial business knowledge are input into the personal financial business big model, the personal financial business big model will use its powerful computing power and understanding of personal financial knowledge to quickly analyze and generate the most appropriate answer. This process not only improves the response speed, but also ensures the accuracy and professionalism of the answer, greatly improving the user experience and service quality.

[0155] For example, the user's question is: What are the personal financial services of Branch A? The answer output by the personal financial service model is: Branch A's personal financial services include the following: 1. Personal housing loans; 2. Personal commercial housing loans; 3. Personal housing provident fund loans; 4. Personal comprehensive consumption and pledge loans; 5. Personal automobile consumption loans.

[0156] The personal financial service intelligent question-answering method provided in this embodiment uses a hash similarity algorithm to determine the first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base when the user question is classified as personal financial service, and determines a second number of personal financial service knowledge in the personal financial service knowledge base that are ranked first according to the first similarity, determines the first semantic feature of the user question and the keyword, and the second semantic feature of each piece of personal financial service knowledge in the second number of pieces of personal financial service knowledge, determines the first number of personal financial service knowledge in the second number of pieces of personal financial service knowledge that are ranked first according to the second similarity according to the second similarity between the first semantic feature and each second semantic feature, inputs the user question and the first number of pieces of personal financial service knowledge into a personal financial service big model, and obtains an answer output by the personal financial service big model, wherein the personal financial service big model is obtained by pre-training a big language model using a personal financial service question-answering sample; the method improves the efficiency and accuracy of users in obtaining relevant information under personal financial services.

[0157] Optionally, in this embodiment Figure 1Based on the embodiment, the question-answering process of the non-personal financial service is described in detail. The question-answering process of the non-personal financial service shown in this embodiment includes:

[0158] When the user question is classified as non-personal financial business, the user question is input into the large language model to obtain the answer output by the large language model.

[0159] When the user's question is classified as non-personal financial business, it indicates that the user's question goes beyond the scope of personal financial services and may involve a broader knowledge field or general information query. At this time, when the user's question is identified as non-personal financial business, submitting it to the large language model means that the model's mastery of a wide range of knowledge fields can be used to find and generate the most relevant answers to the user's question.

[0160] Optionally, in this embodiment Figure 1 Based on the embodiment, the training process of the personal financial service model is described in detail. The training process of the personal financial service model shown in this embodiment includes:

[0161] A large number of personal finance business question and answer sample pairs are obtained, and each personal finance business question and answer sample pair is annotated. The question in each personal finance business question and answer sample pair provides three pieces of knowledge information and a standard answer, forming a one-three structure style. After data processing and segmentation, each piece of knowledge information can be segmented into knowledge information including titles and referenced knowledge piece by piece. At the same time, during the segmentation process, abnormal information of the knowledge information is processed, including the removal of pictures and captions, and the processing of document punctuation.

[0162] For example, a question in a personal finance business question and answer sample is: What is the source of the personal finance cockpit? Take one of the three pieces of knowledge as an example, the title in the knowledge information is: "Personal Finance Business" Chapter 1, Article 3, and the referenced knowledge is: Data information refers to data that is legally acquired, integrated, processed, and stored through various channels such as data warehouses and flow platforms inside and outside the bank according to specific calibers; a standard answer is: data warehouse, flow platform.

[0163] A large number of personal financial business question and answer samples are used to train the large language model, and the parameters of the large language model are fine-tuned using the fine-tuned large language model (LoRA, Low-Rank Adaptation of Large Language Models) to obtain the large model of personal financial business.

[0164] Figure 5 This is the overall process structure diagram of the intelligent question-answering method for personal financial services provided in this application. Figure 5 The figure shows the training and use process of the personal financial business big model, including:

[0165] First, personal financial business question and answer samples are input into the big model for training to obtain the personal financial business big model. Next, user questions are classified, and user questions classified as personal financial business are matched with the personal financial business knowledge base for similarity, and the second number of personal financial business knowledge in the personal financial business knowledge base that are ranked at the top in similarity with the user are extracted. Finally, the user questions and the second number of personal financial business knowledge are input into the personal financial business big model, and the personal financial business big model outputs the answer.

[0166] Figure 6 The structural diagram of the intelligent question-answering device for personal financial services provided in this application is as follows: Figure 6 As shown, the personal financial service intelligent question-answering device 400 provided in this embodiment includes:

[0167] Determination module 401, used to determine keywords in the user's question;

[0168] Processing module 402, used to extract context information of user questions and keywords, wherein the context information includes semantic features and temporal features;

[0169] The determination module 401 is further used to determine the classification of the user question according to the context information, wherein the classification of the user question is personal financial service or non-personal financial service;

[0170] The processing module 402 is further configured to extract a first number of pieces of personal financial service knowledge related to the user's question from a preset personal financial service knowledge base when the user's question is classified as personal financial service;

[0171] The processing module 402 is also used to input the user question and the first number of personal financial business knowledge into the personal financial business big model to obtain the answer output by the personal financial business big model, wherein the personal financial business big model is obtained by pre-training the big language model with personal financial business question and answer samples.

[0172] Optionally, the processing module 402 is further used to segment the user's question using the language technology platform model and perform part-of-speech tagging on the segmentation results;

[0173] The determination module 401 is further configured to determine the words marked as nouns in the word segmentation results as keywords.

[0174] Optionally, the determination module 401 is further configured to determine the criticality of each keyword using a word frequency-inverse text frequency algorithm;

[0175] The processing module 402 is also used to sort the keywords in descending order of criticality;

[0176] The processing module 402 is further used to combine the user question and the sorted keywords into a question description text;

[0177] The determination module 401 is further used to input the question description text into the bidirectional encoder language model to obtain hidden layer information;

[0178] The determination module 401 is also used to input the hidden layer information into the bidirectional long short-term memory network model to obtain context information.

[0179] Optionally, the determination module 401 is further configured to use a hash similarity algorithm to determine a first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base;

[0180] The determination module 401 is further used to determine a second number of pieces of personal financial service knowledge ranked first according to the first similarity in the personal financial service knowledge base;

[0181] The determination module 401 is further used to determine the first semantic features of the user question and the keyword, and the second semantic features of each piece of personal financial business knowledge in the second number of pieces of personal financial business knowledge;

[0182] The determination module 401 is further configured to determine, according to the second similarity between the first semantic feature and each second semantic feature, a first number of pieces of personal financial service knowledge ranked first according to the second similarity among the second number of pieces of personal financial service knowledge.

[0183] Optionally, the determination module 401 is further used to determine a first hash code value of the user question and a second hash code value of each piece of personal financial business knowledge, wherein the number of digits of the first hash code value and the second hash code value is a preset number of digits;

[0184] The processing module 402 is further configured to perform an XOR operation on the first hash code value and the second hash code value;

[0185] The determination module 401 is further configured to divide the number of 1s in the XOR operation result by a preset number of bits to obtain a first similarity.

[0186] Optionally, the determination module 401 is further configured to determine a first hash code value of the user question, including:

[0187] Hash-encode the keywords of the user's question to obtain a first hash-encoded value;

[0188] The determination module 401 is further used to determine the second hash code value of each piece of personal financial service knowledge, including:

[0189] Keywords are extracted from the titles of each piece of personal financial business knowledge, and hash codes are performed on the keywords of the titles of each piece of personal financial business knowledge to obtain second hash code values ​​corresponding to each piece of personal financial business knowledge.

[0190] Optionally, the determination module 401 is further configured to input the user question into the large language model to obtain an answer output by the large language model when the user question is classified as non-personal financial business.

[0191] Figure 7 This is a schematic diagram of the structure of the intelligent question-answering device for personal financial services provided in this application. Figure 7 As shown, the electronic device 500 provided in this embodiment includes: at least one processor 501 and a memory 502. Optionally, the device 500 also includes a communication component 503. The processor 501, the memory 502 and the communication component 503 are connected via a bus 504.

[0192] In a specific implementation process, at least one processor 501 executes the computer-executable instructions stored in the memory 502, so that at least one processor 501 executes the above method.

[0193] The specific implementation process of the processor 501 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0194] In the above embodiments, it should be understood that the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the invention can be directly implemented as a hardware processor, or can be implemented by a combination of hardware and software modules in the processor.

[0195] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk storage.

[0196] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.

[0197] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0198] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0199] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0200] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0201] The division of units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0202] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0203] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0204] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0205] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0206] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0207] It should be further noted that, although the various steps in the flowchart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowchart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0208] It should be understood that the above-mentioned device embodiments are only illustrative, and the device of the present application can also be implemented in other ways. For example, the division of units / modules in the above-mentioned embodiments is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units, modules or components can be combined, or can be integrated into another system, or some features can be ignored or not executed.

[0209] In addition, unless otherwise specified, each functional unit / module in each embodiment of the present application may be integrated into one unit / module, each unit / module may exist physically separately, or two or more units / modules may be integrated together. The above-mentioned integrated unit / module may be implemented in the form of hardware or in the form of a software program module.

[0210] If the integrated unit / module is implemented in the form of hardware, the hardware may be a digital circuit, an analog circuit, etc. The physical implementation of the hardware structure includes but is not limited to transistors, memristors, etc. Unless otherwise specified, the processor may be any appropriate hardware processor, such as a CPU, a GPU, an FPGA, a DSP, an ASIC, etc. Unless otherwise specified, the storage unit may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory (RRAM), a dynamic random access memory (DRAM), a static random access memory (SRAM), an enhanced dynamic random access memory (EDRAM), a high-bandwidth memory (HBM), a hybrid memory cube (HMC), etc.

[0211] If the integrated unit / module is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a memory, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0212] In the above embodiments, the description of each embodiment has its own emphasis. For the part not described in detail in a certain embodiment, please refer to the relevant description of other embodiments. The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0213] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0214] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A personal financial business intelligent question-answering method, characterized in that: include: Determine keywords in the user question, and extract context information of the user question and the keywords, wherein the context information includes semantic features and temporal features; Determining a classification of the user question according to the context information, wherein the classification of the user question is personal financial service or non-personal financial service; In the case where the user question is classified as a personal financial service, extracting a first number of pieces of personal financial service knowledge related to the user question from a preset personal financial service knowledge base; The user question and the first number of pieces of personal financial business knowledge are input into a personal financial business big model to obtain an answer output by the personal financial business big model, wherein the personal financial business big model is obtained by pre-training a big language model with personal financial business question and answer samples.

2. The method according to claim 1, characterized in that Determining the keywords in the user's question includes: Using a language technology platform model to segment the user's question and perform part-of-speech tagging on the segmentation results; The words marked as nouns in the word segmentation results are determined as the keywords.

3. The method according to claim 1, characterized in that The extracting the context information of the user question and the keyword includes: Using a word frequency-inverse text frequency algorithm to determine the criticality of each of the keywords, and sorting the keywords in descending order of criticality; Splicing the user question and the sorted keywords into a question description text; Inputting the problem description text into a bidirectional encoder language model to obtain hidden layer information; The hidden layer information is input into a bidirectional long short-term memory network model to obtain the context information.

4. The method according to any one of claims 1 to 3, characterized in that The extracting a first number of pieces of personal financial service knowledge related to the user's question from a preset personal financial service knowledge base includes: Determine a first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base using a hash similarity algorithm, and determine a second number of pieces of personal financial service knowledge in the personal financial service knowledge base that are ranked first according to the first similarity; Determining first semantic features of the user question and the keyword, and second semantic features of each piece of personal financial business knowledge in the second number of pieces of personal financial business knowledge; According to the second similarity between the first semantic feature and each of the second semantic features, the first number of pieces of personal financial service knowledge that are ranked first according to the second similarity among the second number of pieces of personal financial service knowledge are determined.

5. The method according to claim 4, characterized in that The step of using a hash similarity algorithm to determine a first similarity between the user question and each piece of personal financial service knowledge in the personal financial service knowledge base includes: Determine a first hash code value of the user question and a second hash code value of each piece of personal financial business knowledge, wherein the number of digits of the first hash code value and the second hash code value is a preset number of digits; An XOR operation is performed on the first Hash code value and the second Hash code value, and the number of 1s in the XOR operation result is divided by the preset number of bits to obtain the first similarity.

6. The method according to claim 4, characterized in that Each piece of personal financial business knowledge in the personal financial business knowledge base includes a title and a knowledge point, and the method further includes: Determining a first hash code value of the user question includes: Hash-encode the keywords of the user question to obtain a first hash-encoded value; Determine the second hash code value of each piece of personal financial business knowledge, including: Keywords are extracted from the titles of each piece of personal financial business knowledge, and hash codes are performed on the keywords of the titles of each piece of personal financial business knowledge to obtain second hash code values ​​corresponding to each piece of personal financial business knowledge.

7. The method according to claim 1, characterized in that The method further comprises: In the case where the user question is classified as non-personal financial business, the user question is input into the large language model to obtain an answer output by the large language model.

8. An intelligent question-answering device for personal financial services, comprising: A determination module is used to determine the keywords in the user's question; A processing module, used to extract context information of the user question and the keyword, wherein the context information includes semantic features and temporal features; The determination module is further used to determine the classification of the user question according to the context information, wherein the classification of the user question is personal financial service or non-personal financial service; The processing module is further configured to extract a first number of pieces of personal financial service knowledge related to the user question from a preset personal financial service knowledge base when the user question is classified as a personal financial service; The processing module is further used to input the user question and the first number of personal financial business knowledge into a personal financial business big model to obtain an answer output by the personal financial business big model, wherein the personal financial business big model is obtained by pre-training a big language model using personal financial business question and answer samples.

9. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

11. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 7 when being executed by a processor.

Images