Question and answer interaction method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a question and answer interaction method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the steps of obtaining a to-be-learned text; inputting a first prompt word containing the to-be-learned text into a preset large language model to obtain question content; the first prompt word is used for indicating a preset large language model to generate question content according to the input text; displaying the question content, and obtaining answer content given by the first user for the question content; determining text content matched with the question content in the to-be-learned text; and based on the determined similarity between the text content and the answer content, obtaining an evaluation result representing the understanding degree of the first user for the to-be-learned text. The occupation of labor cost and time cost in the question and answer interaction process can be reduced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a question-answering interaction method, device, electronic device, and storage medium. Background Art

[0002] In recent years, the financial market has developed rapidly and financial products have become increasingly diversified, which requires employees in the financial industry to quickly master knowledge of a large number of new products.

[0003] In existing technology, during employee training, instructors can pre-design a question bank containing questions and standard answers based on the information about each financial product. Employees then answer the questions accordingly. The instructor then determines whether the employee understands the financial product based on their answers and the standard answers. If the employee's understanding of the financial product is insufficient, they can then study the relevant information further.

[0004] However, there is a wide variety of information on financial products. If employees are trained using existing technology, instructors will need to manually design a question bank based on a large amount of information. The efficiency of generating the question bank is low, and instructors will have to manually judge whether employees understand the financial products. The efficiency of generating the judgment results is also low, which requires a lot of labor and time costs. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide a question-and-answer interaction method, device, electronic device, and storage medium to reduce the labor and time costs during the question-and-answer interaction process. The specific technical solution is as follows:

[0006] In a first aspect, an embodiment of the present invention provides a question-answering interaction method, the method comprising:

[0007] Get the text to be learned;

[0008] Inputting a first prompt word containing the text to be learned into a preset large language model to obtain question content; wherein the first prompt word is used to instruct the preset large language model to generate question content based on the input text;

[0009] Displaying the question content and obtaining the answer content given by the first user to the question content; and determining the text content that matches the question content in the text to be learned;

[0010] Based on the determined similarity between the text content and the answer content, an evaluation result indicating the first user's understanding of the text to be learned is obtained.

[0011] Optionally, the text to be learned is a text associated with a specified product;

[0012] Before obtaining the text to be learned, the method further includes:

[0013] Display the configuration page;

[0014] In response to a configuration operation, obtaining customer information indicated by the configuration operation in the configuration page as customer information to be processed;

[0015] Determining, from among the preset products, a preset product whose associated text matches the customer information to be processed as the designated product;

[0016] The first prompt word also includes the customer information to be processed, and the question content conforms to the questioning method of the second user represented by the customer information to be processed.

[0017] Optionally, the customer information to be processed includes attribute values ​​of multiple user attributes;

[0018] The step of determining, from among the preset products, a preset product whose associated text matches the to-be-processed customer information as the designated product includes:

[0019] For each preset product, if the attribute values ​​of each product attribute represented by the text associated with the preset product are the same as the attribute values ​​of each user attribute in the customer information to be processed, the preset product is used as the designated product.

[0020] Optionally, for each preset product, if the attribute values ​​of each product attribute represented by the text associated with the preset product are the same as the attribute values ​​of each user attribute in the customer information to be processed, then the preset product is used as the designated product, including:

[0021] Obtain a pre-constructed decision tree; wherein the decision tree comprises multiple layers, the multiple layers corresponding one-to-one to multiple preset product attributes, and the nodes contained in any layer corresponding one-to-one to the attribute values ​​of the product attributes corresponding to the layer; a node represents a preset product having a corresponding attribute value; a child node represents a preset product belonging to the preset product represented by the parent node of the child node;

[0022] According to the order of the layers of the decision tree, for each node in the first layer of the decision tree, if the attribute value corresponding to the node exists in the customer information to be processed, the node is used as the node to be used in the first layer; wherein the order of the layers in the decision tree is: the order of the parent node pointing to the child node in the decision tree;

[0023] According to the order of the layers, the next layer of the first layer in the decision tree is used as the current layer;

[0024] Determine the child nodes of the node to be used in the previous layer from the current layer;

[0025] For each determined child node, if the attribute value corresponding to the child node exists in the client information to be processed, the child node is used as a node to be utilized in the current layer;

[0026] According to the order of the layers, the layer below the current layer is used as the new current layer, and the process of determining the child nodes of the node to be used in the previous layer from the current layer is returned until the node to be used in the last layer is determined;

[0027] The preset product represented by the determined node to be utilized in the last layer is used as the designated product.

[0028] Optionally, before inputting the first prompt word containing the to-be-learned text into a preset large language model to obtain question content, the method further includes:

[0029] Obtaining scene information of the current scene; wherein the scene information of a scene indicates the type of content that the second user needs to know in the scene;

[0030] The first prompt word also includes scene information of the current scene; the first prompt word is used to instruct the preset large language model to generate question content based on the input text and the type of content that the second user needs to know in the current scene; the evaluation result obtained represents the first user's degree of understanding of the part corresponding to the current scene in the text to be learned.

[0031] Optionally, there are multiple questions, and the obtained answers include answers to each question;

[0032] The step of determining text content that matches the question content in the to-be-learned text includes:

[0033] For each question content, determining text content in the to-be-learned text whose similarity to the question content is greater than a first similarity threshold;

[0034] Obtaining an evaluation result indicating the first user's understanding of the to-be-learned text based on the determined similarity between the text content and the answer content includes:

[0035] Calculating the similarity between the determined text content and the answer content corresponding to the question content;

[0036] If the calculated similarity is greater than the second similarity threshold, the answer content corresponding to the question content is determined to be correct;

[0037] Count the number of correct answers and get the test results.

[0038] Optionally, the current scene is one of a plurality of preset scenes arranged in a preset order;

[0039] After obtaining an evaluation result indicating the first user's understanding of the to-be-learned text based on the determined similarity between the text content and the answer content, the method further includes:

[0040] If the evaluation result indicates that the first user's understanding of the portion corresponding to the current scene in the text to be learned is lower than a first preset threshold, then, according to the preset order, the previous scene is used as the new current scene, and the step of obtaining the scene information of the current scene is returned;

[0041] If the evaluation result obtained indicates that the first user's understanding of the part corresponding to the current scene in the text to be learned is higher than the second preset threshold, then according to the preset order, the next scene is used as the new current scene, and the process of obtaining the scene information of the current scene is returned to execute until the target evaluation result is obtained, and the understanding represented by the target evaluation result is not less than the first preset threshold; wherein, the target evaluation result is: the first user's understanding of the part corresponding to the last scene in the text to be learned.

[0042] In a second aspect, an embodiment of the present invention provides a question-answer interaction device, the device comprising:

[0043] A first acquisition module is used to acquire the text to be learned;

[0044] An input module, configured to input a first prompt word containing the text to be learned into a preset large language model to obtain question content; wherein the first prompt word is used to instruct the preset large language model to generate question content based on the input text;

[0045] A first determination module is configured to display the question content and obtain the answer content given by the first user to the question content; and determine text content in the text to be learned that matches the question content;

[0046] An evaluation module is configured to obtain an evaluation result indicating the first user's understanding of the text to be learned based on the determined similarity between the text content and the answer content.

[0047] An embodiment of the present invention further provides an electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0048] Memory for storing computer programs;

[0049] The processor is configured to implement the above-mentioned question-and-answer interaction method when executing the program stored in the memory.

[0050] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned question-and-answer interaction method is implemented.

[0051] Beneficial effects of the embodiments of the present invention:

[0052] In an embodiment of the present invention, the first prompt word contains the text to be learned, and the first prompt word is used to instruct the preset large language model to generate question content based on the input text, so that the first prompt word is input into the preset large language model to obtain the question content. Then, the text content in the text to be learned that matches the question content is used as the standard answer, and the evaluation result can be obtained based on the similarity between the standard answer and the answer content of the first user to the question content. In the above process, the machine generates the question content with the help of the preset large language model, and automatically determines the standard answer, and obtains the evaluation result based on the similarity between the standard answer and the answer content. This can avoid the manual design of the question bank and the manual judgment of the result based on the answer content, thereby reducing the labor cost and time cost occupied in the question and answer interaction process.

[0053] Of course, it is not necessary to achieve all of the advantages described above simultaneously in order to implement any product or method of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0055] Figure 1 A schematic diagram of a flow chart of a first question-answer interaction method provided in an embodiment of the present invention;

[0056] Figure 2 A schematic diagram of a flow chart of a second question-answer interaction method provided in an embodiment of the present invention;

[0057] Figure 3 A schematic diagram of a process for determining a designated product through a decision tree provided in an embodiment of the present invention;

[0058] Figure 4 A schematic diagram of the structure of a decision tree involved in a question-answering interactive method provided by an embodiment of the present invention;

[0059] Figure 5 A schematic diagram of a flow chart of a third question-answer interaction method provided in an embodiment of the present invention;

[0060] Figure 6 A schematic diagram illustrating the principle of processing a text to be learned in a question-answering interactive method provided by an embodiment of the present invention;

[0061] Figure 7 A schematic diagram of the principle of obtaining question content in a question-answering interactive method provided by an embodiment of the present invention;

[0062] Figure 8 A schematic diagram illustrating the principles of a question-and-answer interaction method provided by an embodiment of the present invention;

[0063] Figure 9 A schematic diagram of the structure of a question-answering interactive device provided by an embodiment of the present invention;

[0064] Figure 10 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field based on the present invention are within the scope of protection of the present invention.

[0066] In order to reduce the labor cost and time cost in the question-answering interaction process, embodiments of the present invention provide a question-answering interaction method, device, electronic device, and storage medium.

[0067] The following first introduces a question-answer interaction method provided by an embodiment of the present invention.

[0068] Among them, a question-answering interaction method provided by an embodiment of the present invention can be applied to electronic devices. Exemplarily, the electronic device can be a smart phone, a tablet computer or a desktop computer. The question-answering interaction method provided by an embodiment of the present invention can be applied to scenarios such as employee training or student education. The electronic device can display the content of the question, and the first user answers the content of the question, wherein the first user can be an employee participating in training or a student learning knowledge, etc. The electronic device can also simulate questioners of various roles (i.e., the second user) based on a preset large language model, and ask questions to the first user to fit the real question-answering scenario. For example, if the first user is an employee, the second user can be the customer served by the employee; if the first user is a student, the second user can be the teacher who teaches the student.

[0069] An embodiment of the present invention provides a question-answer interaction method, which may include the following steps:

[0070] Get the text to be learned;

[0071] Inputting a first prompt word containing the text to be learned into a preset large language model to obtain question content; wherein the first prompt word is used to instruct the preset large language model to generate question content based on the input text;

[0072] Displaying the question content and obtaining the answer content given by the first user to the question content; and determining the text content that matches the question content in the text to be learned;

[0073] Based on the determined similarity between the text content and the answer content, an evaluation result indicating the first user's understanding of the text to be learned is obtained.

[0074] In an embodiment of the present invention, the first prompt word contains the text to be learned, and the first prompt word is used to instruct the preset large language model to generate question content based on the input text, so that the first prompt word is input into the preset large language model to obtain the question content. Then, the text content in the text to be learned that matches the question content is used as the standard answer, and the evaluation result can be obtained based on the similarity between the standard answer and the answer content of the first user to the question content. In the above process, the machine generates the question content with the help of the preset large language model, and automatically determines the standard answer, and obtains the evaluation result based on the similarity between the standard answer and the answer content. This can avoid the manual design of the question bank and the manual judgment of the result based on the answer content, thereby reducing the labor cost and time cost occupied in the question and answer interaction process.

[0075] The following combination Figure 1 Introducing a question-answering interactive method provided by an embodiment of the present invention, such as Figure 1 As shown, the method may include steps S101-S104.

[0076] S101, obtaining the text to be learned.

[0077] It is understandable that the text to be learned is a text that the first user needs to learn. In order to detect the first user's understanding of the text to be learned, the electronic device needs to generate question content for the text to be learned.

[0078] For example, if the first user is an employee of a company, the text to be studied may be information about products sold by the company, such as product usage tutorials and product purchase procedures. If the first user is a student, the text to be studied may be teaching materials, such as textbooks and workbooks.

[0079] Among them, the text to be learned can be data pre-stored in a knowledge database. The electronic device can obtain the text to be learned from the knowledge database storing the text to be learned. In one implementation, the electronic device can obtain the text to be learned associated with a specified product to generate question content for the specified product and detect the first user's understanding of the information about the specified product. This implementation will be introduced in subsequent embodiments. In another implementation, the electronic device can obtain all the texts to be learned to generate question content for all the texts to be learned and detect the first user's understanding of all the texts to be learned. In this regard, the embodiments of the present invention are only illustrative and not specifically limited.

[0080] The text to be learned can be a text obtained by extracting the content of various types of data. Among them, various types of data can be information about the product associated with the text to be learned. For example, in the financial industry, the data source database stores various types of data that comply with financial industry standards, such as introduction videos of financial products, policy documents related to financial products, and purchase contract documents for amount products. When an update occurs in the data source database, the device where the data source database is located can write the updated various types of data into the message queue. The device where the knowledge database is located can obtain various types of data from the message queue, and for each type of data, parse the data of that type according to the parsing method corresponding to that type to extract the text to be learned.

[0081] Specifically, for data of non-editable document types, optical character recognition (OCR) technology can be used to identify the text information in the data of this type to obtain the text to be learned; wherein, the non-editable document type data can be Portable Document Format (PDF) data or image data. For video type data, OCR technology or automatic speech recognition (ASR) technology can be used to identify the text information in the video type data to obtain the text to be learned. For data of editable document types, the text information can be directly extracted to obtain the text to be learned.

[0082] It can be understood that the text to be learned can exist in a knowledge database. In order to facilitate the retrieval of the required text in the text to be learned, each part of the text to be learned can be converted into a semantic vector, and the semantic vector of each part of the text and the text identifier of the part can be stored in the semantic vector database.

[0083] In one implementation, after extracting the text to be learned, the device where the knowledge database is located can segment the text to be learned according to the specified dimension to obtain each segment data, and then vectorize each segment data to obtain the semantic vector of each segment data, and store the semantic vector of each segment data in the semantic vector database. The specified dimension can be a sentence, a paragraph, or a chapter, etc. If the electronic device needs to obtain the text to be learned of a specified product, the electronic device can first obtain the semantic vector representing the specified product (for example, a product attribute vector representing the product portrait of the specified product), and then determine the semantic vector that matches the obtained semantic vector in the semantic vector database, and then obtain the text represented by the text identifier corresponding to the matching semantic vector to obtain the text to be learned of the specified product.

[0084] It is understandable that the text to be learned may also include product portraits. In one implementation, the device where the knowledge database is located can identify various types of data in the text to be learned, determine the attribute values ​​of each attribute of the preset product associated with the text to be learned, obtain the product portrait of each preset product, and then add the product portrait to the text to be learned. For example, identify the risk level, intended investment type, and intended investment period of the preset product associated with the text to be learned. The product portrait can enrich the content of the text to be learned, so that the preset large language model can output question content that is more in line with the text to be learned, and improve the accuracy of evaluating the first user's understanding of the learning text.

[0085] S102: Input the first prompt word containing the text to be learned into a preset large language model to obtain question content.

[0086] The first prompt word is used to instruct the preset large language model to generate question content based on the input text.

[0087] It is understandable that the electronic device can use the text to be learned to generate a first prompt word containing the text to be learned. Specifically, the electronic device can splice the text to be learned and the preset task instructions to obtain the first prompt word. The preset task instructions are in text form and are used to indicate the operations to be performed by the preset large language model. For example, the content of the preset task instructions can be "Generate 100 questions according to the content of the text to be learned."

[0088] Among them, the preset large language model can be a language model trained using large-scale text data. The preset large language model has rich language knowledge, comprehension ability and generation ability, can recognize the semantics of the input content, and generate text to respond to the input content.

[0089] Since the first prompt word can instruct the preset large language model to generate question content based on the input text, the electronic device inputs the first prompt word into the preset large language model. The preset large language model can understand the text to be learned in the first prompt word based on its own understanding ability and generate question content for the text to be learned.

[0090] S103: Display the question content, obtain the answer content given by the first user to the question content, and determine the text content that matches the question content in the text to be learned.

[0091] It is understood that after obtaining the question content, the electronic device can display the question content. For example, the electronic device can display the question content on the question page, and the first user can answer the question content in the input box on the question page, and then the electronic device can obtain the answer content given by the first user to the question content.

[0092] After obtaining the question content, the electronic device can determine the text content that matches the question content in the text to be learned, and use the matched text content as the standard answer. Subsequently, the similarity between the standard answer and the first user's answer content can be compared to determine the first user's understanding of the text to be learned.

[0093] Illustratively, in order to determine the text content matching the question content in the text to be learned, the electronic device may determine the text content with the greatest similarity to the question content in each segment data in the text to be learned as the text content matching the question content.

[0094] It should be noted that step S103 includes two operations: the first operation is to display the question content and obtain the answer content, and the second operation is to determine the text content that matches the question content. Both operations are performed after obtaining the question content.

[0095] S104 : Based on the determined similarity between the text content and the answer content, an evaluation result indicating the first user's understanding of the text to be learned is obtained.

[0096] It is understandable that the determined text content is the content in the text to be learned, the text to be learned is the content that the first user needs to learn, and the content in the text to be learned can be used as a standard answer to be compared with the answer content of the first user.

[0097] In one implementation, there is one question and one corresponding answer. If the similarity between the determined text content and the answer content is greater than a preset threshold, it can be determined that the first user's answer content is close to the content of the text to be learned, and an evaluation result can be obtained indicating that the first user has a high level of understanding of the text to be learned. If the similarity between the determined text content and the answer content is not greater than the preset threshold, it can be determined that the first user's answer content is not close to the content of the text to be learned, and an evaluation result can be obtained indicating that the first user has a low level of understanding of the text to be learned.

[0098] In another implementation, there are multiple questions, and the obtained answers include answers to each question; step S103 includes step A1, and step S104 includes steps A2-A4.

[0099] A1. For each question content, determine text content in the to-be-learned text that has a similarity with the question content greater than a first similarity threshold.

[0100] A2, calculating the similarity between the determined text content and the answer content corresponding to the question content.

[0101] A3: If the calculated similarity is greater than a second similarity threshold, it is determined that the answer content corresponding to the question content is correct.

[0102] A4 counts the number of correct answers and obtains the test results.

[0103] It is understood that the preset large language model can output multiple questions, the first user can answer each question, and the electronic device can obtain multiple answers. The multiple questions output by the preset large language model can be questions about different parts of the text to be learned. The electronic device can comprehensively evaluate the first user's understanding of the learning content based on the questions and answers for different parts.

[0104] For each question content, the electronic device can determine the text content in the text to be learned that has a similarity with the question content greater than a first similarity threshold. Specifically, the electronic device can convert the question content into a semantic vector to obtain the semantic vector of the question content; then, for the semantic vector of each fragment of the text to be learned in the semantic vector database, calculate the similarity between the semantic vector of the question content and the semantic vector of each fragment (for example, the cosine similarity of the two semantic vectors can be calculated), and determine the semantic vector whose similarity with the semantic vector of the question content is greater than a specified threshold. The text corresponding to the determined semantic vector is then used as the text content that matches the question content to accurately determine the text content that matches the question content.

[0105] The electronic device may use the text content that matches each question as a criterion to determine the answer content corresponding to the question content. If the similarity between the text content that matches the question content and the answer content is greater than a second similarity threshold, the answer content may be determined to be correct. The first similarity threshold and the second similarity threshold may be the same or different.

[0106] The electronic device can count the number of correct answers to obtain a detection result. For example, if the number of correct answers exceeds a preset threshold, an evaluation result can be determined indicating that the first user has a high level of understanding of the text to be learned; if the number of correct answers does not exceed the preset threshold, an evaluation result can be determined indicating that the first user has a low level of understanding of the text to be learned. The evaluation result can be the number of correct answers; the evaluation result can also be whether the first user understands the text content of the text to be learned, for example, the evaluation result can be the field content of "the first user understands the text to be learned" or "the first user does not understand the text to be learned."

[0107] In one embodiment, Figure 1 Based on the question-answering interactive method shown in , the text to be learned is the text associated with the specified product, such as Figure 2 As shown, the method further includes steps S201-S203.

[0108] S201, displaying a configuration page.

[0109] S202 : In response to the configuration operation, obtain the customer information indicated by the configuration operation in the configuration page as the customer information to be processed.

[0110] S203 : Determine, from among the preset products, a preset product whose associated text matches the customer information to be processed as the designated product.

[0111] It is understood that the to-be-learned text is text associated with a specified product. For example, the to-be-learned text may include the specified product's operating procedures, purchasing guide, and product introduction. The specified product is a product from a plurality of preset products, and the plurality of preset products are all products that the first user can provide to the second user. For example, if the first user is an employee of a financial company and the second user is a client of the financial company, the plurality of preset products are all products that the financial company's employees can provide to the client.

[0112] The electronic device may display a configuration page, where the first user may perform configuration operations. The configuration operations may indicate the customer information to be processed. For example, the configuration page may include a selection box, where the first user may select customer information. The customer information may represent a profile of the customer. For example, the customer information may include gender, age, education level, asset status, risk assessment level, and health status.

[0113] The electronic device can determine, from among the preset products, a designated product whose associated text matches the customer information to be processed. In other words, the text to be learned associated with the designated product matches the customer information to be processed.

[0114] It is understandable that the electronic device can also add the customer information to be processed in the first prompt word, and the preset large language model can output the question content according to the text to be learned in the first prompt word and the questioning method of the second user represented by the customer information to be processed, that is, simulate the second user to ask questions. If the first user is an employee and the second user is a customer, by simulating the second user's questioning, it is possible to model the scenario of the customer asking questions to the employee. Among them, the questioning method of the second user represented by the customer information to be processed can be a questioning method that conforms to the user portrait of the second user. For example, if the second user's education level is junior high school, the second user's questioning method will use spoken language to ask questions; if the second user's education level is college, the second user's questioning method will use written language to ask questions. The preset large language model can simulate the second user's questioning method to ask questions, which is more in line with real business scenarios and can improve the real experience of the first user's answer.

[0115] In one implementation, the customer information to be processed includes attribute values ​​of multiple user attributes, and step S203 includes the following steps:

[0116] For each preset product, if the attribute values ​​of each product attribute represented by the text associated with the preset product are the same as the attribute values ​​of each user attribute in the customer information to be processed, the preset product is regarded as the designated product.

[0117] It is understood that the customer information to be processed includes attribute values ​​for multiple user attributes, where the multiple user attributes may include risk assessment level, investment type preference, and investment period preference. The attribute value for each user attribute may be the specific content of that attribute. For example, if the user attribute is risk assessment level, the attribute value of that user attribute may indicate low risk; if the user attribute is investment type preference, the attribute value of that user attribute may indicate a preference for investing in equity products; and if the user attribute is investment period preference, the attribute value of that user attribute may indicate a preference for short-term investments of 0-1 years.

[0118] The electronic device can determine, for each preset product, the attribute value of the product attribute that belongs to the same dimension as the attributes of each user in the customer information to be processed. For example, the customer information to be processed contains the risk assessment level, investment type preference and investment period preference, and the electronic device can determine the risk level, intended investment type and intended investment period of the preset product. The electronic device can then compare the attribute values ​​of the product attributes and customer attributes of the same dimension. If the attribute values ​​of each dimension are the same, the preset product is used as the designated product. For example, the risk assessment level in the customer information to be processed is low risk, the investment type preference is equity products, and the investment period preference is 0-1 years; in the product attributes of the preset product, the risk level of the preset product is low risk, the intended investment type is equity, and the intended investment period is 0-1 years, then it can be determined that the preset product matches the customer information to be processed, and the preset product is the designated product.

[0119] To facilitate comparison of the attribute values ​​of user attributes and product attributes, in one implementation, the electronic device may convert the attribute values ​​of the user attributes into user attribute vectors, convert the attribute values ​​of the product attributes into product attribute vectors, and then calculate the similarity between the product attribute vector and the user attribute vector for each preset product. If the similarity is greater than a preset threshold, the preset product may be determined to be the designated product. Exemplarily, the electronic device may use a one-hot encoding method to convert the attribute value of each attribute into an independent binary feature, that is, convert the attribute value of the user attribute into a user attribute vector, and convert the attribute value of the product attribute into a product attribute vector.

[0120] In another implementation, the electronic device may determine the designated product through a decision tree. Figure 3 The flowchart of determining a specified product through a decision tree provided by an embodiment of the present invention is as follows. Figure 3 As shown, the process of determining a designated product through a decision tree may include steps S301-S307.

[0121] S301, obtaining a pre-built decision tree.

[0122] Among them, the decision tree contains multiple layers, and the multiple layers correspond one-to-one to multiple preset product attributes. The nodes contained in any layer correspond one-to-one to the attribute values ​​of the product attributes corresponding to the layer; a node represents a preset product with a corresponding attribute value; the preset product represented by a child node belongs to the preset product represented by the parent node of the child node.

[0123] It is understandable that the electronic device can construct a decision tree for each preset product according to a plurality of preset product attributes, and then store the constructed decision tree. When the electronic device needs to determine a specific product, it can obtain the pre-constructed decision tree.

[0124] The decision tree contains multiple layers, and the multiple layers correspond one-to-one to multiple preset product attributes, that is, each layer in the decision tree corresponds to a product attribute, and the product attributes corresponding to each layer are different. The nodes contained in any layer correspond one-to-one to the attribute values ​​of the product attribute corresponding to the layer, that is, each node contained in any layer corresponds to an attribute value, and the attribute value belongs to the product attribute corresponding to the layer. A node represents a preset product with a corresponding attribute value. For example, a node can represent at least one preset product with a corresponding attribute value. The preset product represented by a child node belongs to the preset product represented by the parent node of the child node, that is, the preset product represented by a child node is a subset of the preset product represented by the parent node of the child node, and in the order of the layers of the decision tree, the layer where a child node is located is the next layer of the layer where the parent node of the child node is located, that is, the order of the layers in the decision tree is: the order of the parent node pointing to the child node in the decision tree.

[0125] The following will introduce the decision tree in conjunction with the decision tree that represents financial products. Figure 4 A schematic diagram of the structure of a decision tree involved in a question-answering interactive method provided in an embodiment of the present invention.

[0126] like Figure 4 As shown, the decision tree contains multiple layers and a root node, with node 1 being the root node, and node 1 representing each preset product; the product attribute corresponding to the first layer in the decision tree is the risk level, and the attribute values ​​belonging to the risk level include: low risk, medium-low risk, medium risk, medium-high risk, and high risk; the product attribute corresponding to the second layer in the decision tree is the intended investment type, and the attribute values ​​belonging to the intended investment type include: principal protection and currency type, fixed income type, equity type, credit trading-leverage trading type, and complex or high-risk type; the product attribute corresponding to the third layer in the decision tree is the intended investment period, and the attribute values ​​belonging to the intended investment period include: 0-1 year, 0-3 years, 0-5 years, and no fixed period.

[0127] The first level of the decision tree includes: Node 2 for low risk, Node 3 for medium-low risk, Node 4 for medium risk, Node 5 for medium-high risk, and Node 6 for high risk. The second level of the decision tree includes Nodes 7, 11, 16, and 31. Nodes 7 through 11 are child nodes of Node 2. Node 7 corresponds to the principal protection and currency type, while Node 11 corresponds to the complex or high-risk type. The pre-set products represented by Nodes 7 through 11 belong to the pre-set products in Node 2. In other words, the attribute values ​​of the pre-set products represented by Nodes 7 through 11 include the attribute values ​​corresponding to the node itself and the attribute values ​​corresponding to Node 2. For example, the pre-set product represented by Node 7 is a low-risk product that belongs to the principal protection and currency type. Nodes 16 through 31 are child nodes of Node 6.

[0128] The third level of the decision tree includes nodes 32 corresponding to 0-1 years, 33 corresponding to 0-3 years, 34 corresponding to 0-5 years, and 35 corresponding to an unfixed term. Nodes 32, 33, 34, and 35 are child nodes of node 7, and the pre-set products represented by nodes 32, 33, 34, and 35 belong to the pre-set product represented by node 7. The attribute values ​​of the pre-set products represented by nodes 32 through 35 include their own corresponding attribute values ​​and the attribute values ​​corresponding to node 7. For example, the pre-set product represented by node 32 is a low-risk, principal-guaranteed currency product with an investment period of 0-1 years. The third layer of the decision tree also includes node n corresponding to 0-1 years, node n+1 corresponding to 0-3 years, node n+2 corresponding to 0-5 years, and node n+3 corresponding to no fixed term, among which node n, node n+1, node n+2 and node n+3 are child nodes of node 31, and the preset products represented by node n, node n+1, node n+2 and node n+3 belong to the preset products represented by node 31.

[0129] The decision tree ends at the third level. If the decision tree is used to determine a specific product, the preset product represented by the node in the third level is the specific product. Subsequently, obtaining the to-be-learned text associated with the specific product is equivalent to obtaining the product data package indicated by the node in the third level. Node 32 can instruct the electronic device to obtain the data for the specific product represented by that node, namely, product data package 1. Node 33 can instruct the electronic device to obtain product data package 2. Node 34 can instruct the electronic device to obtain product data package 3. Node 35 can instruct the electronic device to obtain product data package 4.

[0130] S302 , according to the order of each layer of the decision tree, for each node in the first layer of the decision tree, if the attribute value corresponding to the node exists in the customer information to be processed, the node is used as a node to be used in the first layer.

[0131] S303: According to the order of each layer, the next layer of the first layer in the decision tree is used as the current layer.

[0132] S304: Determine the child nodes of the node to be used in the previous layer from the current layer.

[0133] S305 , for each determined child node, if the attribute value corresponding to the child node exists in the customer information to be processed, the child node is used as a node to be utilized in the current layer.

[0134] S306 , according to the order of each layer, the layer below the current layer is used as the new current layer, and the process returns to determine the child nodes of the node to be used in the previous layer from the current layer, until the node to be used in the last layer is determined.

[0135] S307 , taking the preset product represented by the determined node to be utilized in the last layer as the designated product.

[0136] It is understood that the electronic device can determine the nodes to be used in sequence starting from the first layer according to the order of the layers of the decision tree, and traverse the child nodes in each layer of the decision tree in the direction from the parent node to the child node, finally reaching the last layer, and quickly locate the preset product that matches the customer information to be processed. There is at least one designated product.

[0137] Specifically, the electronic device can determine, for each node in the first layer of the decision tree, whether the attribute value corresponding to the node exists in the customer information to be processed. If so, it is determined that the attribute value corresponding to the node is the same as the attribute value in the customer information to be processed, and the node is used as the node to be used in the first layer; then the next layer of the first layer in the decision tree is used as the current layer; the child nodes of the node to be used in the previous layer are determined from the current layer, that is, the nodes connected to the node to be used in the previous layer are determined from the current layer; for each determined child node, it is determined whether the attribute value corresponding to the child node exists in the customer information to be processed. If so, the next layer of the current layer is used as the new current layer, the child nodes are determined, and then it is determined whether the attribute value corresponding to the child node exists in the customer information to be processed, until the node to be used in the last layer is determined.

[0138] For example, the attribute values ​​included in the customer information to be processed are low risk, principal protection currency type and 0-1 year. Figure 4 The designated product is determined based on the decision tree shown. Specifically, in the first layer, node 2, corresponding to low risk, is identified as the first-level node to be utilized. In the second layer, among the child nodes of node 2 (i.e., nodes 7 through 11), node 7, corresponding to the principal-protected currency type, is identified as the second-level node to be utilized. In the third layer, among the child nodes of node 7 (i.e., nodes 32 through 35), node 32, corresponding to the 0-1 year period, is identified as the third-level node to be utilized. The third layer is the final layer of the decision tree, so node 32 can be determined as the designated product.

[0139] In an embodiment of the present invention, the electronic device can match preset products based on a traversal method of a decision tree from a parent node to a child node. It can transform a complex customer information matching process into a structured judgment path through hierarchical condition screening, narrow the search scope by orderly dividing attributes, avoid searching massive product information in the database, and improve the efficiency of determining a specified product.

[0140] In one embodiment, Figure 1 Based on the method shown, Figure 5 As shown, the method further includes steps S501-S503.

[0141] S501: Acquire scene information of the current scene.

[0142] The scenario information of a scenario represents the type of content that the second user needs to know in the scenario.

[0143] It is understandable that different scenarios exist for pre-set products, and in different scenarios, the types of content that the second user needs to understand vary. For example, in the scenario of understanding financial products, customers need to understand the product features and user reviews of the financial products; in the scenario of purchasing financial products, customers need to understand the purchase guide and operation process of the financial products.

[0144] To more accurately simulate a second user asking a question to the first user, the electronic device can obtain contextual information about the current scene, use this contextual information as part of a first prompt, and input the first prompt into a preset large language model. The preset large language model can then generate a question based on the input text and the type of content the second user needs to understand in the current scene. In other words, the model generates a question tailored to the type of content the second user needs to understand in the current scene. Therefore, the resulting evaluation result can indicate the first user's level of understanding of the portion of the text to be learned that corresponds to the current scene.

[0145] For example, the current scenario is a product understanding scenario. The scenario information of the current scenario may include product features and user reviews. The preset large language model can identify the text representing product features and user reviews in the text to be learned, and use the text representing product features and user reviews to generate question content.

[0146] In one implementation, the scene information of the current scene may be configured on a scene configuration page. Specifically, a first user may configure the scene information of the current scene, and the electronic device obtains the scene information of the current scene configured by the first user. In another implementation, the scene information of the current scene may be one of multiple preset scenes. The electronic device may use each preset scene as the current scene and obtain the scene information of the preset scene as the scene information of the current scene. This implementation will be described in detail in subsequent steps.

[0147] S502: If the evaluation result obtained indicates that the first user's understanding of the part corresponding to the current scene in the text to be learned is lower than the first preset threshold, the previous scene is used as the new current scene in the preset order, and the execution is returned to obtain the scene information of the current scene.

[0148] S503. If the evaluation result obtained indicates that the first user's understanding of the part corresponding to the current scene in the text to be learned is higher than the second preset threshold, then according to the preset order, the next scene is used as the new current scene, and the execution returns to obtain the scene information of the current scene until the target evaluation result is obtained, and the understanding level represented by the target evaluation result is not less than the first preset threshold.

[0149] The target evaluation result is: the first user's understanding of the part corresponding to the last scene in the text to be learned.

[0150] It is understandable that in order to better fit the real scene and improve the flexibility of questions and answers, the electronic device can switch the current scene among multiple preset scenes. The multiple preset scenes are arranged in a preset order, and the processing flow of the real business can be used as the arrangement order of the multiple preset scenes. For example, from the customer's perspective, financial products have the following scenario stages: understanding, decision-making, purchase, holding and exit. In the understanding scenario, the customer needs to understand the characteristics of the financial product; in the decision-making scenario, the customer needs to understand the risks of the financial product; in the purchase scenario, the customer needs to understand the purchase process of the financial product; in the holding scenario, the customer needs to understand the operating status of the financial product; in the exit scenario, the customer needs to understand the sales process of the financial product.

[0151] The evaluation result for the current scenario represents the first user's understanding of the portion of the learning text corresponding to the current scenario. If there are multiple questions, the number of correct answers can represent this understanding; if there is only one question, the similarity between the answer and the text in the learning text can represent this understanding.

[0152] The electronic device can make a judgment for each of the multiple preset scenarios until the level of understanding of the portion corresponding to the last scenario exceeds a second preset threshold, thereby determining that the first user has understanding of each scenario. Because the scenarios are interdependent in the order in which the preset scenarios are arranged, if the level of understanding of the current scenario is lower than the first preset threshold, it is determined that the first user has a low level of understanding not only of the current scenario, but also of the scenario preceding the current scenario. For example, if, in a purchase scenario, the first user does not understand the purchase process for a specified product, the first user may also not understand the product information of the specified product. In other words, the first user does not understand the understanding scenario preceding the purchase scenario.

[0153] The current scene is one of multiple preset scenes arranged in a preset order. For example, the current scene may be the first scene among the multiple preset scenes, or the last scene among the current scenes, or any other scene except the first and last scene among the current scenes.

[0154] If the current scene is the last scene, only step S502 can be executed, and step S503 cannot be executed. If the current scene is the first scene, only step S503 can be executed, and step S502 cannot be executed. If the current scene is another scene, the first user can determine whether to execute step S502 or S503 based on the obtained evaluation results representing the relationship between the first user's understanding of the portion of the text to be learned corresponding to the current scene and the first preset threshold, and the relationship between the understanding and the second preset threshold.

[0155] It can be understood that the first preset threshold is smaller than the second preset threshold. For example, the first preset threshold is 70 and the second preset threshold is 90. The number of correct answers in the current scenario can indicate the degree of understanding. If the number of correct answers is 60, questions will be asked about the previous scenario (previous scenario) of the current scenario; if the number of correct answers is 98, questions will be asked about the next scenario (subsequent scenario) of the current scenario.

[0156] In one implementation, the electronic device can record the first user's answer progress. For example, the electronic device can record the scene identifier of the current scene, the customer identifier of the customer information to be processed, the product identifier of the designated product, the user identifier of the first user, and the evaluation results of the previous scene to facilitate the first user to understand his or her own training progress.

[0157] In an embodiment of the present invention, the electronic device can dynamically adjust the current scene according to the first user's understanding of each scene, realize personalized question and answer, improve the flexibility of question and answer, and be more in line with the real scene. It not only avoids repeating questions and answers for scenes that have been understood, but also automatically advances to questions and answers for the next scene, and automatically matches the first user with questions that are appropriate to his or her knowledge level, thereby improving the first user's experience.

[0158] Figure 6 A schematic diagram of the principle of processing a text to be learned in a question-answering interactive method provided in an embodiment of the present invention.

[0159] like Figure 6 As shown, the device housing the data source database can write the updated data to a message queue when an update occurs in the data source database. That is, when a database change occurs, the changed data is written to the message queue. The device housing the knowledge database can utilize message queue services to retrieve various types of data from the message queue. This device can also be referred to as a financial product intelligent training system. The knowledge database can store the various types of data it acquires in the database and determine the attribute values ​​of each attribute for the preset product to which each type of data belongs. This generates a product profile for each preset product, which is then stored in the database to configure the product profile.

[0160] The knowledge database device can then parse various data types based on the product information type to obtain the text to be learned. After extracting the text to be learned, the knowledge database device can segment the text according to the specified dimensions to obtain each segmented data, i.e., perform knowledge segmentation. Each segmented data is then vectorized to obtain a semantic vector for each segmented data, which is then stored in the semantic vector database.

[0161] Figure 7 A schematic diagram of the principle of obtaining question content in a question-answering interactive method provided by an embodiment of the present invention.

[0162] like Figure 7 As shown, the electronic device can display a configuration page; in response to a configuration operation, obtain the customer information indicated in the configuration page as the customer information to be processed (i.e., configure a customer profile); and from each preset product, determine the preset product whose associated text matches the customer information to be processed as the designated product. Then, obtain the text associated with the designated product as the text to be learned (i.e., match the product profile and recommend a product information package). In one implementation, the electronic device can use a custom selected product information package.

[0163] After confirming the product data package, the electronic device can configure the scenario elements and obtain the context information of the current scenario. It then generates a first prompt word containing the text to be learned, the customer information to be processed, and the context information of the current scenario (i.e., generates a comprehensive role construction prompt word). The electronic device can then input the first prompt word into a preset large language model to obtain the question content.

[0164] Figure 8 A schematic diagram illustrating the principles of a question-and-answer interaction method provided in an embodiment of the present invention.

[0165] like Figure 8 As shown, the electronic device can display a configuration page; in response to a configuration operation, obtain the customer information indicated in the configuration page as the customer information to be processed (i.e., configuring a customer profile); and from each preset product, determine the preset product whose associated text matches the customer information to be processed as the designated product. Then, obtain the text associated with the designated product as the text to be learned (i.e., the associated knowledge base). In one implementation, the electronic device can use a custom selection product information package. That is, the product information package is determined based on matching recommendations or customization.

[0166] After determining the product information package, the electronic device can obtain the scene information of the current scene (i.e., scene elements). Then generate a first prompt word containing the text to be learned, the customer information to be processed, and the scene information of the current scene (i.e., generate a comprehensive role construction prompt word). The electronic device can then input the first prompt word into the preset large language model, so that the preset large language model constructs the second user (i.e., constructs a sparring role) to obtain the question content. The first user can interact with the second user constructed by the preset large language model in sparring, evaluate and analyze the answer content of the first user, and obtain an evaluation result. The current scene is one of multiple preset scenes arranged in a preset order. For example, the multiple preset scenes include scene 1, scene 2, and scene 3.

[0167] If the evaluation result obtained indicates that the first user's understanding of the part corresponding to the current scene in the text to be learned is lower than the first preset threshold, the previous scene is used as the new current scene in the preset order, and the scene information of the new current scene is used to construct the prompt word.

[0168] If the evaluation result obtained indicates that the first user's understanding of the part corresponding to the current scene in the text to be learned is higher than the second preset threshold, then the next scene is used as the new current scene in the preset order, and the scene information of the new current scene is used to construct prompt words until the target evaluation result is obtained, and the understanding represented by the target evaluation result is not less than the first preset threshold; wherein, the target evaluation result is: the first user's understanding of the part corresponding to the last scene in the text to be learned.

[0169] Figure 9 A structural diagram of a question-answering interactive device provided by an embodiment of the present invention is shown in FIG. Figure 9 As shown, the device includes:

[0170] A first acquisition module 910 is used to acquire a text to be learned;

[0171] Input module 920, configured to input a first prompt word containing the text to be learned into a preset large language model to obtain a question content; wherein the first prompt word is used to instruct the preset large language model to generate a question content based on the input text;

[0172] A first determination module 930 is configured to display the question content and obtain the answer content given by the first user to the question content; and determine text content in the text to be learned that matches the question content;

[0173] The evaluation module 940 is configured to obtain an evaluation result indicating the first user's understanding of the text to be learned based on the determined similarity between the text content and the answer content.

[0174] Optionally, the text to be learned is a text associated with a specified product;

[0175] The device further comprises:

[0176] Display module, used to display the configuration page;

[0177] a response module, configured to, in response to a configuration operation, obtain the customer information indicated by the configuration operation in the configuration page as customer information to be processed;

[0178] A second determining module is configured to determine, from among the preset products, a preset product whose associated text matches the customer information to be processed, as the designated product;

[0179] The first prompt word also includes the customer information to be processed, and the question content conforms to the questioning method of the second user represented by the customer information to be processed.

[0180] Optionally, the customer information to be processed includes attribute values ​​of multiple user attributes;

[0181] The first determination module includes:

[0182] The determination unit is configured to, for each preset product, determine that if the attribute values ​​of each product attribute represented by the text associated with the preset product are the same as the attribute values ​​of each user attribute in the customer information to be processed, the preset product is used as the designated product.

[0183] Optionally, determine the unit, including:

[0184] An acquisition subunit is configured to acquire a pre-constructed decision tree; wherein the decision tree comprises multiple layers, each of which corresponds one-to-one to a plurality of preset product attributes, and each node in any layer corresponds one-to-one to each attribute value of the product attribute corresponding to the layer; a node represents a preset product having a corresponding attribute value; a preset product represented by a child node belongs to the preset product represented by the parent node of the child node;

[0185] a first determining subunit configured to, according to the order of the layers of the decision tree, for each node in the first layer of the decision tree, determine, if the attribute value corresponding to the node exists in the customer information to be processed, that node as a node to be utilized in the first layer; wherein the order of the layers in the decision tree is the order of parent nodes pointing to child nodes in the decision tree;

[0186] A second determining subunit is configured to use, according to the order of the layers, a layer next to the first layer in the decision tree as the current layer;

[0187] A third determining subunit is used to determine, from the current layer, the child nodes of the node to be utilized in the previous layer;

[0188] a fourth determining subunit, configured to, for each determined child node, determine, if the attribute value corresponding to the child node exists in the client information to be processed, to use the child node as a node to be utilized in the current layer;

[0189] An execution subunit, configured to, in accordance with the order of the layers, take the next layer of the current layer as the new current layer, and return to execute the process of determining the child nodes of the node to be utilized in the previous layer from the current layer, until the node to be utilized in the last layer is determined;

[0190] The fifth determining subunit is configured to use the preset product represented by the determined node to be utilized in the last layer as the designated product.

[0191] Optionally, the device further includes:

[0192] A second acquisition module is configured to acquire scene information of a current scene; wherein the scene information of a scene indicates the type of content that the second user needs to know in the scene;

[0193] The first prompt word also includes scene information of the current scene; the first prompt word is used to instruct the preset large language model to generate question content based on the input text and the type of content that the second user needs to know in the current scene; the evaluation result obtained represents the first user's degree of understanding of the part corresponding to the current scene in the text to be learned.

[0194] Optionally, there are multiple questions, and the obtained answers include answers to each question;

[0195] A first determination module is specifically configured to determine, for each question content, text content in the to-be-learned text having a similarity with the question content greater than a first similarity threshold;

[0196] Evaluation modules include:

[0197] a calculation unit, configured to calculate a similarity between the determined text content and the answer content corresponding to the question content;

[0198] a determining unit, configured to determine that the answer content corresponding to the question content is correct if the calculated similarity is greater than a second similarity threshold;

[0199] The statistical unit is used to count the number of correct answers and obtain the test results.

[0200] Optionally, the current scene is one of a plurality of preset scenes arranged in a preset order;

[0201] The device further comprises:

[0202] a first execution module configured to, if the obtained evaluation result indicates that the first user's understanding of the portion corresponding to the current scene in the to-be-learned text is lower than a first preset threshold, use the previous scene as a new current scene according to the preset order, and return to executing the step of obtaining scene information of the current scene;

[0203] The second execution module is used to, if the obtained evaluation result indicates that the first user's understanding of the part corresponding to the current scene in the text to be learned is higher than a second preset threshold, then, in accordance with the preset order, use the next scene as the new current scene, and return to execute the scene information of the current scene until the target evaluation result is obtained, and the understanding represented by the target evaluation result is not less than the first preset threshold; wherein, the target evaluation result is: the first user's understanding of the part corresponding to the last scene in the text to be learned.

[0204] In an embodiment of the present invention, the first prompt word contains the text to be learned, and the first prompt word is used to instruct the preset large language model to generate question content based on the input text, so that the first prompt word is input into the preset large language model to obtain the question content. Then, the text content in the text to be learned that matches the question content is used as the standard answer, and the evaluation result can be obtained based on the similarity between the standard answer and the answer content of the first user to the question content. In the above process, the machine generates the question content with the help of the preset large language model, and automatically determines the standard answer, and obtains the evaluation result based on the similarity between the standard answer and the answer content. This can avoid the manual design of the question bank and the manual judgment of the result based on the answer content, thereby reducing the labor cost and time cost occupied in the question and answer interaction process.

[0205] The embodiment of the present invention further provides an electronic device, such as Figure 10 As shown, it includes a processor 1001 , a communication interface 1002 , a memory 1003 and a communication bus 1004 , wherein the processor 1001 , the communication interface 1002 , and the memory 1003 communicate with each other via the communication bus 1004 .

[0206] Memory 1003, used for storing computer programs;

[0207] The processor 1001 is configured to implement the question-answering interaction method when executing the program stored in the memory 1003 .

[0208] The communication bus mentioned in the electronic device mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. This communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is used in the figure, but this does not mean that there is only one bus or only one type of bus.

[0209] The communication interface is used for communication between the above electronic device and other devices.

[0210] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage. Alternatively, the memory may be at least one storage device located away from the processor.

[0211] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0212] In another embodiment provided by the present invention, a computer-readable storage medium is further provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned question-and-answer interaction methods are implemented.

[0213] In another embodiment provided by the present invention, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any one of the question-answer interaction methods in the above embodiments.

[0214] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0215] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0216] Each embodiment in this specification is described in a related manner. Similar portions between the embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences from other embodiments. In particular, the device embodiments are generally similar to the method embodiments, so their description is relatively simple. For related portions, refer to the description of the method embodiments.

[0217] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A question-answering interactive method, characterized in that: The method comprises: Get the text to be learned; Inputting a first prompt word containing the text to be learned into a preset large language model to obtain question content; wherein the first prompt word is used to instruct the preset large language model to generate question content based on the input text; Displaying the question content and obtaining the answer content given by the first user to the question content; and determining the text content that matches the question content in the text to be learned; Based on the determined similarity between the text content and the answer content, an evaluation result indicating the first user's understanding of the text to be learned is obtained.

2. The method according to claim 1, characterized in that The text to be learned is a text associated with a specified product; Before obtaining the text to be learned, the method further includes: Display the configuration page; In response to a configuration operation, obtaining customer information indicated by the configuration operation in the configuration page as customer information to be processed; Determining, from among the preset products, a preset product whose associated text matches the customer information to be processed as the designated product; The first prompt word also includes the customer information to be processed, and the question content conforms to the questioning method of the second user represented by the customer information to be processed.

3. The method according to claim 2, characterized in that The client information to be processed includes attribute values ​​of multiple user attributes; The step of determining, from among the preset products, a preset product whose associated text matches the to-be-processed customer information as the designated product includes: For each preset product, if the attribute values ​​of each product attribute represented by the text associated with the preset product are the same as the attribute values ​​of each user attribute in the customer information to be processed, the preset product is used as the designated product.

4. The method according to claim 3, characterized in that For each preset product, if the attribute values ​​of each product attribute represented by the text associated with the preset product are the same as the attribute values ​​of each user attribute in the customer information to be processed, then the preset product is used as the designated product, including: Obtain a pre-constructed decision tree; wherein the decision tree comprises multiple layers, the multiple layers corresponding one-to-one to multiple preset product attributes, and the nodes contained in any layer corresponding one-to-one to the attribute values ​​of the product attributes corresponding to the layer; a node represents a preset product having a corresponding attribute value; a child node represents a preset product belonging to the preset product represented by the parent node of the child node; According to the order of the layers of the decision tree, for each node in the first layer of the decision tree, if the attribute value corresponding to the node exists in the customer information to be processed, the node is used as the node to be used in the first layer; wherein the order of the layers in the decision tree is: the order of the parent node pointing to the child node in the decision tree; According to the order of the layers, the next layer of the first layer in the decision tree is used as the current layer; Determine the child nodes of the node to be used in the previous layer from the current layer; For each determined child node, if the attribute value corresponding to the child node exists in the client information to be processed, the child node is used as a node to be utilized in the current layer; According to the order of the layers, the layer below the current layer is used as the new current layer, and the process of determining the child nodes of the node to be used in the previous layer from the current layer is returned until the node to be used in the last layer is determined; The preset product represented by the determined node to be utilized in the last layer is used as the designated product.

5. The method according to claim 1, wherein Before inputting the first prompt word containing the to-be-learned text into a preset large language model to obtain question content, the method further includes: Obtaining scene information of the current scene; wherein the scene information of a scene indicates the type of content that the second user needs to know in the scene; The first prompt word also includes scene information of the current scene; the first prompt word is used to instruct the preset large language model to generate question content based on the input text and the type of content that the second user needs to know in the current scene; the evaluation result obtained represents the first user's degree of understanding of the part corresponding to the current scene in the text to be learned.

6. The method according to claim 1, characterized in that There are multiple questions, and the obtained answers include answers to each question; The step of determining text content that matches the question content in the to-be-learned text includes: For each question content, determining text content in the to-be-learned text whose similarity to the question content is greater than a first similarity threshold; Obtaining an evaluation result indicating the first user's understanding of the to-be-learned text based on the determined similarity between the text content and the answer content includes: Calculating the similarity between the determined text content and the answer content corresponding to the question content; If the calculated similarity is greater than the second similarity threshold, the answer content corresponding to the question content is determined to be correct; Count the number of correct answers and get the test results.

7. The method according to claim 5, characterized in that The current scene is one of multiple preset scenes arranged in a preset order; After obtaining an evaluation result indicating the first user's understanding of the to-be-learned text based on the determined similarity between the text content and the answer content, the method further includes: If the evaluation result indicates that the first user's understanding of the portion corresponding to the current scene in the text to be learned is lower than a first preset threshold, then, according to the preset order, using the previous scene as the new current scene, and returning to execute the step of obtaining the scene information of the current scene; If the evaluation result obtained indicates that the first user's understanding of the part corresponding to the current scene in the text to be learned is higher than the second preset threshold, then according to the preset order, the next scene is used as the new current scene, and the process of obtaining the scene information of the current scene is returned to until the target evaluation result is obtained, and the understanding represented by the target evaluation result is not less than the first preset threshold; wherein, the target evaluation result is: the first user's understanding of the part corresponding to the last scene in the text to be learned.

8. A question-answering interactive device, characterized in that: The device comprises: A first acquisition module is used to acquire the text to be learned; An input module, configured to input a first prompt word containing the text to be learned into a preset large language model to obtain question content; wherein the first prompt word is used to instruct the preset large language model to generate question content based on the input text; A first determination module is configured to display the question content and obtain the answer content given by the first user to the question content; and determine text content in the text to be learned that matches the question content; An evaluation module is configured to obtain an evaluation result indicating the first user's understanding of the text to be learned based on the determined similarity between the text content and the answer content.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory for storing computer programs; A processor, configured to implement the method according to any one of claims 1 to 7 when executing a program stored in a memory.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images