Auxiliary learning method, device, equipment, system and program product
By acquiring test questions, answers, and supplementary learning information, task instructions are generated to guide the interaction between the large language model and the user, solving the problem of not being able to understand the solution process in existing technologies and improving learning effectiveness.
Patent Information
- Application Number
- CN202410848408.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-27
- Publication Date
- 2024-11-12
AI Technical Summary
Existing AI-assisted question-answering systems can only provide answers to questions, but cannot help users understand the solution process, thus offering limited learning assistance.
By acquiring test questions, answers, and supplementary learning information, task instructions are generated to guide the large language model in generating interactive information, assisting users in answering test questions step by step. This includes teaching and research guidance logic, similar question supplementary learning records, and related knowledge, thereby achieving interaction with users.
It enhances users' understanding of the test question answering process and methods, and promotes knowledge learning and mastery of answering skills.
Smart Images

Figure CN118942295A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence technology, and in particular to an auxiliary learning method, device, equipment, system and program product. Background Art
[0002] Under the current circumstances, the application of artificial intelligence in assisted answering scenarios is mainly to assist users in knowledge query and test question answering. For example, users can conveniently retrieve and query knowledge through devices equipped with artificial intelligence technology, or answer test questions through the devices to obtain the correct answers.
[0003] However, the above-mentioned question-answering assistance function realized by artificial intelligence can only enable users to know the answers to the test questions, but cannot enable users to truly understand the ideas and methods of solving the test questions, and its role in helping users' learning is limited. Summary of the invention
[0004] In view of this, the present application provides an assisted learning method, which can improve the effectiveness of information in the assisted learning process.
[0005] According to a first aspect of an embodiment of the present application, an auxiliary learning method is provided, which can be applied to a system or program including an auxiliary learning function in a terminal device, and specifically includes:
[0006] Acquire a test question, an answer to the test question, and supplementary learning information of the test question, wherein the supplementary learning information includes at least one of teaching and research guidance logic, supplementary learning records of similar questions, and related knowledge;
[0007] Based on the test question, the answer to the test question, and the auxiliary learning information of the test question, a first task instruction is generated, where the first task instruction is used to instruct the large language model to generate interactive information to guide the user to correctly answer the test question;
[0008] The first task instruction is input into a pre-trained first large language model, so that the first large language model generates interactive information for guiding the user to correctly answer the test question by executing the first task instruction.
[0009] Optionally, in some possible implementations of the present application, obtaining the answer to the test question includes:
[0010] Solve the test question by using the pre-trained second largest language model to obtain the answer to the test question;
[0011] The second largest language model is obtained by training test questions and answer data.
[0012] Optionally, in some possible implementations of the present application, answering the test question by using a pre-trained second language model to obtain an answer to the test question includes:
[0013] generating a second task instruction based on the test question, wherein the second task instruction is used to instruct the large language model to answer the test question and output the answer result to the test question according to the set answer output requirement;
[0014] The second task instruction is input into a pre-trained second largest language model, so that the second largest language model obtains the answer to the test question by executing the second task instruction.
[0015] Optionally, in some possible implementations of the present application, when the auxiliary learning information includes teaching and research guidance logic, obtaining the auxiliary learning information of the test question includes:
[0016] Determine the test point category corresponding to the test question;
[0017] Based on the test point category corresponding to the test question, the teaching and research guidance logic corresponding to the test point category is determined from a preset guidance logic knowledge base; wherein the guidance logic knowledge base stores the teaching and research guidance logic corresponding to various test point categories.
[0018] Optionally, in some possible implementations of the present application, when the supplementary learning information includes a supplementary learning record of a similar question, obtaining the supplementary learning information of the question includes:
[0019] Based on the test question, the supplementary learning records of the test questions similar to the test question are retrieved from a preset supplementary learning record database; wherein the supplementary learning record database stores supplementary learning record data of a plurality of test questions.
[0020] Optionally, in some possible implementations of the present application, when the supplementary learning information includes relevant knowledge, obtaining the supplementary learning information of the test question includes:
[0021] The knowledge related to the test question is retrieved from a preset knowledge base.
[0022] Optionally, in some possible implementations of the present application, the first task instruction includes:
[0023] The test questions, answers to the test questions, supplementary learning information of the test questions, introduction information of supplementary learning tasks, and introduction information of interactive information output requirements;
[0024] Among them, the auxiliary learning task is used to interact with the user according to the set interaction process and interaction requirements, and guide the user to correctly answer the test questions; the interaction information output requirements include requirements for the interaction information generated by the large language model output.
[0025] Optionally, in some possible implementations of the present application, the introduction information of the auxiliary learning task includes: at least one of a description of the auxiliary learning task and an introduction to the auxiliary learning link;
[0026] and / or,
[0027] The interactive information output requirements include at least one of the following: interactive information highlighting requirements, material marking requirements, interactive dialogue bottom line requirements, interactive information semantic unit block definitions, and interactive information semantic unit block output requirements.
[0028] Optionally, in some possible implementations of the present application, the first task instruction also includes tag introduction information and / or interaction record information;
[0029] The label introduction information includes introduction information of the set identification label;
[0030] The interaction record information includes historical interaction content with the user and the user's current input content.
[0031] According to a second aspect of an embodiment of the present application, there is provided an auxiliary learning device, specifically comprising:
[0032] An acquisition unit, used to acquire the test questions, the answers to the test questions, and supplementary learning information of the test questions, wherein the supplementary learning information includes at least one of the teaching and research guidance logic, supplementary learning records of similar questions, and related knowledge;
[0033] A processing unit, configured to generate a first task instruction based on the test question, the answer to the test question, and the auxiliary learning information of the test question, wherein the first task instruction is used to instruct the large language model to generate interactive information for guiding the user to correctly answer the test question;
[0034] The learning unit is used to input the first task instruction into a pre-trained first large language model, so that the first large language model generates interactive information to guide the user to correctly answer the test question by executing the first task instruction.
[0035] Optionally, in some possible implementations of the present application, the acquisition unit is specifically configured to answer the test question by using a pre-trained second language model to obtain an answer to the test question;
[0036] The second largest language model is obtained by training test questions and answer data.
[0037] Optionally, in some possible implementations of the present application, the acquisition unit is specifically used to generate a second task instruction based on the test question, and the second task instruction is used to instruct the large language model to answer the test question and output the answer result to the test question according to the set answer output requirement;
[0038] The acquisition unit is specifically used to input the second task instruction into a pre-trained second largest language model, so that the second largest language model obtains the answer to the test question by executing the second task instruction.
[0039] Optionally, in some possible implementations of the present application, the acquisition unit is specifically used to determine the test point category corresponding to the test question;
[0040] The acquisition unit is specifically used to determine the teaching and research guidance logic corresponding to the test point category from a preset guidance logic knowledge base based on the test point category corresponding to the test question; wherein the guidance logic knowledge base stores the teaching and research guidance logic corresponding to various test point categories.
[0041] Optionally, in some possible implementations of the present application, the acquisition unit is specifically used to retrieve, based on the test question, supplementary learning records of test questions similar to the test question from a preset supplementary learning record database; wherein the supplementary learning record database stores supplementary learning record data for multiple test questions.
[0042] Optionally, in some possible implementations of the present application, the acquisition unit is specifically used to retrieve knowledge related to the test question from a preset knowledge base.
[0043] Optionally, in some possible implementations of the present application, the learning unit is specifically used to obtain the test question, the answer to the test question, the auxiliary learning information of the test question, the introduction information of the auxiliary learning task, and the introduction information of the interactive information output requirement;
[0044] Among them, the auxiliary learning task is used to interact with the user according to the set interaction process and interaction requirements, and guide the user to correctly answer the test questions; the interaction information output requirements include requirements for the interaction information generated by the large language model output.
[0045] Optionally, in some possible implementations of the present application, the learning unit is specifically used to obtain the introduction information of the auxiliary learning task, including: at least one of the description of the auxiliary learning task and the introduction of the auxiliary learning link;
[0046] and / or,
[0047] The interactive information output requirement is obtained, including at least one of the following: interactive information highlighting requirement, material marking requirement, interactive dialogue bottom line requirement, interactive information semantic unit block definition, and interactive information semantic unit block output requirement.
[0048] Optionally, in some possible implementations of the present application, the learning unit is specifically used to determine that the first task instruction also includes label introduction information and / or interaction record information;
[0049] The label introduction information includes introduction information of the set identification label;
[0050] The interaction record information includes historical interaction content with the user and the user's current input content.
[0051] According to a third aspect of an embodiment of the present application, there is provided an auxiliary learning device, specifically comprising:
[0052] An input-output component, and a processor connected to the input-output component; wherein the input-output component is used to receive data input by a user, and output data sent by the processor;
[0053] The processor is used to obtain the test questions input by the user through the input-output component, and generate interactive information to guide the user to correctly answer the test questions by executing the auxiliary learning method described in any one of the above aspects, and output the interactive information through the input-output component.
[0054] According to the fourth aspect of the embodiment of the present application, an auxiliary learning system is provided, which specifically includes: a human-computer interaction client, and a server communicatively connected to the human-computer interaction client; the human-computer interaction client is used to receive data input by the user, and output data sent by the processor; the server is used to obtain test questions input by the user through the input-output component, and by executing the auxiliary learning method described in any one of the above aspects, generate interactive information to guide the user to correctly answer the test questions, and output the interactive information through the input-output component.
[0055] According to a fifth aspect of an embodiment of the present application, a computer program product is provided, which specifically includes computer program instructions, and when the computer program instructions are executed by a processor, the processor executes the assisted learning method as described in any one of the above aspects.
[0056] The assisted learning method proposed in this application can, on the basis of obtaining test questions, test answers, and auxiliary learning information of the test questions, further trigger the large language model to generate interactive information to guide users to correctly answer the test questions. By triggering the large language model to interact with the user, it can guide the user to answer the test questions step by step, so that the user can understand the test answering process and answering methods, which helps users to fundamentally learn knowledge and master answering methods, and improves the effectiveness of information in the assisted learning process. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0058] Figure 1 A diagram of the network architecture that runs the assisted learning system;
[0059] Figure 2 A flow chart of an auxiliary learning process provided in an embodiment of the present application;
[0060] Figure 3 A flowchart of an auxiliary learning method provided in an embodiment of the present application;
[0061] Figure 4 A schematic diagram of a scenario of an auxiliary learning method provided in an embodiment of the present application;
[0062] Figure 5 A schematic diagram of another scenario of an auxiliary learning method provided in an embodiment of the present application;
[0063] Figure 6 A schematic diagram of the structure of an auxiliary learning device provided in an embodiment of the present application;
[0064] Figure 7 A schematic diagram of the structure of a terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0065] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0066] It should be understood that the auxiliary learning method provided in the present application can be applied to a system or program containing an auxiliary learning function in a terminal device, such as a teaching auxiliary application. Specifically, the auxiliary learning system can be run on Figure 1 In the network architecture shown in Figure 1 As shown in FIG. 1 , it is a network architecture diagram of the auxiliary learning system. As can be seen from the figure, the auxiliary learning system can provide an auxiliary learning process with multiple information sources, that is, obtaining information related to the test questions through the terminal side, thereby triggering the server side to output auxiliary learning information related to the test questions; it can be understood that Figure 1 A variety of terminal devices are shown in FIG. 1 . The terminal devices may be computer devices. In actual scenarios, more or fewer types of terminal devices may participate in the assisted learning process. The specific number and type depend on the actual scenario and are not limited here. In addition, Figure 1 One server is shown in the figure, but in actual scenarios, multiple servers may be involved, especially in multidisciplinary solution scenarios, and the specific number of servers depends on the actual scenario.
[0067] In this embodiment, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, tablet computer, laptop computer, desktop computer, smart speaker, smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected through wired or wireless communication, and the terminal and the server can be connected to form a blockchain network, which is not limited in this application.
[0068] It can be understood that the above-mentioned assisted learning system can run on a personal mobile terminal, for example: as an application such as educational tutoring, it can also run on a server, and it can also run on a third-party device to provide assisted learning to obtain the assisted learning processing results of the information source; the specific assisted learning system can be run in the above-mentioned device in the form of a program, or it can be run as a system component in the above-mentioned device, and it can also be used as a cloud service program. The specific operation mode depends on the actual scenario and is not limited here.
[0069] Under the current circumstances, the application of artificial intelligence in assisted answering scenarios is mainly to assist users in knowledge query and test question answering. For example, users can conveniently retrieve and query knowledge through devices equipped with artificial intelligence technology, or answer test questions through the devices to obtain the correct answers.
[0070] However, the above-mentioned question-answering assistance function realized by artificial intelligence can only enable users to know the answers to the test questions, but cannot enable users to truly understand the ideas and methods of solving the test questions, and its role in helping users' learning is limited.
[0071] In order to solve the above problems, this application proposes an auxiliary learning method, which is applied to Figure 2 In the process framework of assisted learning shown in Figure 2 As shown, it is a process architecture diagram of assisted learning provided in an embodiment of the present application. The user performs interactive operations based on assisted learning through the terminal, and the server solves the questions displayed by the terminal and generates answers to the questions; and generates auxiliary learning content from different dimensions, thereby combining the answers to the questions and the auxiliary learning content to generate task instructions of a large language model to support the interaction process with the terminal.
[0072] It is understandable that the auxiliary learning method provided by the present application can be written as a program as a processing logic in a hardware system, or as an auxiliary learning device, which implements the above processing logic in an integrated or external manner. As an implementation method, the auxiliary learning device further triggers the large language model to generate interactive information that guides users to correctly answer the test questions based on the acquisition of test questions, test answers, and auxiliary learning information of the test questions. By triggering the large language model to interact with the user, it can guide the user to answer the test questions step by step, so that the user can understand the test answering process and answering method, which helps the user to fundamentally learn knowledge and master the answering method.
[0073] Combined with the above process architecture, the following will introduce the auxiliary learning method in this application. Please refer to Figure 3 , Figure 3 A flowchart of an auxiliary learning method provided in an embodiment of the present application, the embodiment of the present application at least includes the following steps:
[0074] 301. Obtain test questions, answers to the test questions, and supplementary learning information for the test questions, where the supplementary learning information includes at least one of the teaching and research guidance logic, supplementary learning records of similar questions, and related knowledge.
[0075] In this embodiment, the process of obtaining test questions is to determine the questions to be performed for assisted learning. Specifically, the questions may be obtained by the user using an image acquisition device (for example, the user takes photos of teaching materials), or by the user's selection (for example, selection in an electronic textbook), or by the use of various sensor acquisition methods (for example, a handheld sensing acquisition device identifies homework). The specific method depends on the actual scenario.
[0076] Furthermore, when the user specifies the questions that need to be supplemented, the triggering conditions for supplementary learning are met, so as to further determine the answers to the questions and the supplementary learning information of the questions. For example, when a user is practicing Chinese questions in an offline exercise book and there are questions that need to be supplemented during the practice, the user can take a photo and upload it to the supplementary learning application. The supplementary learning application automatically performs optical character recognition (OCR) in the background to convert the questions specified by the user into text.
[0077] It can be understood that the process of converting a specified question into text can be obtained by identifying the symbols carried in the text when converting the question image into text, such as underscores, underscores, bold, italics, etc., so as to achieve accurate question division and completely restore the original state of the question.
[0078] Below, the answers to the test questions involved in the auxiliary learning process and the auxiliary learning information of the test questions are explained.
[0079] The answer to the test question can be obtained through a large language model, that is, the test question is first answered through the pre-trained second largest language model to obtain the answer to the test question; wherein, the second largest language model is trained to answer the test question using the test question and answer data pairs, so it can be called the large answer model. The large answer model is the main tool for obtaining the answer, and is trained by collecting a large number of Chinese question-answer data pairs, so that the model can have a strong answering ability.
[0080] It can be understood that the big answering model uses the question text corresponding to the test question as input, uses the big answering model to predict the answer to the test question, and establishes structured information for the answer, so as to accurately distinguish the answer to each question in the subsequent question-answering process.
[0081] Specifically, the configuration of structured information is to generate task instructions that instruct the large model to answer each question. Therefore, for the process of obtaining the answer to the test question, a second task instruction can be first generated based on the test question, and the second task instruction is used to instruct the large language model to answer the test question and output the answer result to the test question according to the set answer output requirements; and the second task instruction is input into the pre-trained second large language model, so that the second large language model obtains the answer to the test question by executing the second task instruction.
[0082] It is understandable that the second task instruction is a type of prompt information, which is used to trigger the answering ability of the large model. Specifically, since the number of input questions can be multiple, it is necessary to splice the prompts. For this splicing process, first use regular expression rules to mark the symbols such as "()" and "______" that need to be answered in the question stem text with serial numbers, and then splice the question stem of the marked test questions with the prompt template.
[0083] In a possible scenario, the prompt template example is as follows:
[0084] ## Please understand the question stem according to the test question symbol convention and correctly answer this primary school Chinese test question.
[0085] ## Test question symbol convention
[0086] 1. <point>< / point> For the under-dot symbol.
[0087] For example: sky <point> null< / point> , indicating that the character "sky" has no subscript dot, and only the character "empty" has a subscript dot.
[0088] 2. For the underscore symbol.
[0089] For example: vast of the vast sea, indicating that the two characters "vast" have an underscore, and the three characters "of the sea" have no underscore.
[0090] 3. <wavy>< / wavy> For the under-wavy line symbol.
[0091] For example: <wavy> Azure< / wavy> of the blue sky, indicating that the two characters "blue" have an under-wavy line, and the three characters "of the sky" have no under-wavy line.
[0092] 4. <udouble>< / udouble> For the double-underscore symbol.
[0093] For example: Bees are in <udouble> Dance< / udouble> , indicating that the three characters "Bees are in" have no double-underscore, and the two characters "dancing" have a double-underscore.
[0094] 5. For the italic symbol.
[0095] For example: smart bright , indicating that the character "smart" has no italic, and only the character "wise" is italic.
[0096] 6. For the bold symbol.
[0097] For example: tall prosperIt means that the character "高" is not in bold, only the character "兴" is in bold.
[0098] 7. <blank answer area number> is the symbol for the answer area.
[0099] For example: <blank 0-1>______, watching the rising tide in the wind and rain across the river. It means that this question has one answer area with the number 0-1, and the answer needs to be filled in <blank 0-1>______.
[0100] ## Output format convention
[0101] Please complete the answer in the answer area of the question stem and retain the answer area symbols in the question stem, and output the completed text content.
[0102] It can be understood that for the symbol convention of the test questions in the above example, when the OCR recognizes the question text, it will carry some symbols in the text. Therefore, it is necessary to tell the large model what the symbol tags in the text represent respectively to ensure the feasibility of the recognition process.
[0103] In a possible scenario, based on the above prompt template example, the following test question examples can be obtained:
[0104] Question stem: Write the homophone according to the pinyin.
[0105] [fēng]
[0106] <blank 0-1>______ sharp
[0107] crazy <blank 0-2>______
[0108] mountain <blank 0-3>______
[0109] Answer: <blank 0-1> sharp
[0110] crazy <blank 0-2> wind
[0111] mountain <blank 0-3> peak
[0112] ## Question stem
[0113] {Test question stem}
[0114] ## Task restatement
[0115] Refer to the test question symbol convention, understand the question stem, and then answer. Please note that for each <blank>< / blank> included answer area, it must be completed.
[0116] It can be understood that based on the above configuration of the answering large model, the expected output of the answering large model is a string, which is the question stem with answers filled in each answering area.
[0117] In addition, in order to adapt to the subsequent splicing with supplementary learning content, regular expression rules can be used to extract the answers in each answering area and form a json structure. An example is as follows:
[0119] {"idx":"0-1","cont":"front"},
[0120] {"idx":"0-2","cont":"wind"},
[0121] {"idx":"0-3","cont":"peak"}
[0123] Among them, idx represents the number of the answering area, and "0-1" represents the first answering area of the "0"th sub-question. Cont represents the standard answer of this answering area.
[0124] Therefore, through this data structure, the structured results of the answers for each sub-question and each answering area can be obtained, providing a prerequisite for the next supplementary learning and answering questions.
[0125] Next, the possible composition of the supplementary learning information will be described.
[0126] In this embodiment, the supplementary learning information can also be determined using a model similar to the above answering large model, that is, the determination of the supplementary learning information can also use a large language model. This large language model can share the underlying model structure with the answering model or can be trained separately, and the specific method depends on the actual scenario.
[0127] Therefore, for the determination of the supplementary learning information, multiple possible compositions of the supplementary learning information can be spliced for the prompt, so as to construct the model input, where the supplementary learning information can include at least one of the teaching and research guiding logic, similar question supplementary learning records, and relevant knowledge, and thus be jointly input into the supplementary learning and answering large language model as supplementary information, enabling the supplementary learning and answering model to have stronger supplementary learning capabilities and overcoming problems such as overly divergent guidance and disloyalty to facts in the supplementary learning and answering model.
[0128] Specifically, considering that the teaching and research guidance logic comes from the experience summary of teachers in teaching activities, according to the different test points and different question types of Chinese test questions, the explanation framework is pre-defined, which contains the teacher's rich teaching experience and can be regarded as the easiest way for students to understand and learn. In order to be close to the real teaching scenario, this embodiment pre-constructs the guidance logic of common test points of Chinese test questions, which is used to guide the large model to select and adjust teaching strategies according to different test questions, so as to better control the teaching rhythm and teaching focus. Therefore, in the case where the auxiliary learning information includes the teaching and research guidance logic, for the process of obtaining the auxiliary learning information of the test questions. First, the test point category corresponding to the test question can be determined; then, based on the test point category corresponding to the test question, the teaching and research guidance logic corresponding to the test point category is determined from the preset guidance logic knowledge base; wherein, the guidance logic knowledge base stores the teaching and research guidance logic corresponding to each of the various test point categories.
[0129] In a possible scenario, taking the Chinese language test as an example, the Chinese language test entered by the user can be classified to find the test point category with the highest probability for this test point. Among them, one test point category can correspond to multiple sets of different guidance logics, and different guidance logics will be input into the auxiliary learning question answering model. For the above process of obtaining the most suitable guidance logic for this test point through retrieval, the following examples can be used:
[0130] "Evaluating Characters" test point guidance logic
[0131] Explanation: The point of this question is to evaluate the character's image.
[0132] Explanation: To evaluate a person, we need to combine descriptions of the person's appearance, language, actions, psychology, and expression, and analyze the person's personality / quality and other characteristics.
[0133] Heuristic question: Now, please tell me where in the text there is a detailed description of the character’s appearance, language, actions, psychology, and expression?
[0134] Discussion dialogue: (student response + teacher evaluation)
[0135] Explanation: Explain the question in detail
[0136] Heuristic question: Please think about what characteristics of *** can be seen through the description of ***'s appearance / language / actions / psychology / expression.
[0137] Discussion dialogue: (student response + teacher evaluation)
[0138] Explanation: Explain the question in detail
[0139] Explanation: After analyzing the characteristics of the character, you can follow the example of “Through the article’s *** description of ***, you can see that he / she is a ***** person.
[0140] Among them, a classification model can be used to determine the test point category corresponding to the test questions. The classification model can use but is not limited to common statistical probability models, deep neural network models, large language models, etc., which are not limited here.
[0141] Furthermore, in addition to the more generalized teaching and research guidance logic, this embodiment also designs a database of high-quality supplementary learning records annotated by a number of senior teaching and research teachers. This database selects a small number of high-quality test questions commonly used in teaching, and stores the real tutoring process carried out by teachers based on these typical test questions. The tutoring process involves real student reactions and teacher replies, so it is highly professional and suitable as a reference input for the large supplementary learning model. Therefore, in the case where the supplementary learning information includes supplementary learning records for similar questions, the process of obtaining the supplementary learning information of the test question can be based on the test question, and the supplementary learning records of similar test questions of the test question can be retrieved from the preset supplementary learning record database; wherein, the supplementary learning record data of multiple test questions are stored in the supplementary learning record database.
[0142] Specifically, in the process of retrieving similar supplementary learning records of test questions from the preset supplementary learning record database, the test questions to be tutored can be vectorized using a vectorization tool, and then the vector similarity calculation is performed with the test questions in the high-quality supplementary learning record database to select similar tutoring records. The examples that can be used to implement this process are as follows:
[0143]
[0144]
[0145] It is understandable that the above examples are a structuring and presentation of test question tutoring records, which focus on the teacher's output content and, in some possible scenarios, the students' responses, in order to reflect the interaction process between teachers and students and the tutoring process.
[0146] Furthermore, considering that during the auxiliary learning process, students' free questions may include various types, such as subject knowledge questions, encyclopedia knowledge questions, and chat questions, in order to enhance the large model's ability to respond to subject knowledge, encyclopedia knowledge, and other questions that require background knowledge. Therefore, this embodiment designs a knowledge base retrieval module, uses user questions as retrieval items, and uses a pre-built background knowledge base as a retrieval source to match the background knowledge that may be involved in the user's questions, and also splices it into the input of the auxiliary learning question-answering large model as background knowledge. That is, when the auxiliary learning information includes relevant knowledge, for the process of obtaining the auxiliary learning information of the test questions, the knowledge related to the test questions can be retrieved from the preset knowledge base.
[0147] In a possible scenario, examples of retrieving knowledge related to test questions from a preset knowledge base are as follows:
[0148] User input: How to write the stroke order of the character "回"?
[0149] Match hit: (回, stroke order)
[0150] Retrieval result: The stroke order of the character "回" is: vertical, horizontal fold, vertical, horizontal fold, horizontal, horizontal.
[0151] By determining the supplementary learning information that combines the test point classification, guided logical retrieval, and knowledge base retrieval above, it can be adapted to the configuration of the subsequent supplementary learning Q&A large model.
[0152] 302. Generate a first task instruction based on the test question, the answer to the test question, and the supplementary learning information of the test question. The first task instruction is used to instruct the large language model to generate interactive information for guiding the user to correctly answer the test question.
[0153] In this embodiment, the first task instruction is a type of prompt information, which is used to instruct the large language model (supplementary learning Q&A large model) to generate interactive information for guiding the user to correctly answer the test question. Combining the descriptions of the test question, the answer to the test question, and the supplementary learning information of the test question in step 301 above, the first task instruction is composed of the prompts corresponding to the test question, the answer to the test question, and the supplementary learning information of the test question spliced together.
[0154] Specifically, for the setting of the prompts for the test question, the answer to the test question, and the supplementary learning information, refer to the description in step 301 and will not be elaborated here. Since the supplementary learning information can include the introduction information of the supplementary learning task and the introduction information of the requirements for outputting interactive information, the composition of the first task instruction can include the test question, the answer to the test question, the supplementary learning information of the test question, the introduction information of the supplementary learning task, and the introduction information of the requirements for outputting interactive information; among them, the supplementary learning task is used to interact with the user according to the set interactive process and interactive requirements to guide the user to correctly answer the test question; and the requirements for outputting interactive information include the requirements for the interactive information generated by the large language model. For example, the stem of the major question and the stem of the minor question selected by the user, the user's voice or keyboard input, supplemented by the intermediate result of the system, namely the standard answer, the teaching and research guiding logic, the supplementary learning record of similar questions, and the knowledge base retrieval result, are jointly spliced into the prompt template as the input of the supplementary learning Q&A large model.
[0155] Specifically, the introduction information for the supplementary learning tasks may include at least one of the description of the supplementary learning tasks and the introduction of the supplementary learning links. Among them, the supplementary learning links can be composed of multiple semantic unit blocks. This is because one semantic unit block can correspond to a sentence or a paragraph output by the large model, and it is a semantic-based unit block division of the large model output. Therefore, in one link, there can be one semantic unit block or multiple semantic unit blocks. For the requirements for outputting interaction information, it may include at least one of the requirements for highlighting the interaction information, the requirements for marking the materials, the requirements for the bottom line of the interaction dialogue, the definition of the semantic unit block of the interaction information, and the requirements for outputting the semantic unit block of the interaction information, thus ensuring the comprehensiveness of the interaction information.
[0156] In summary, for the first task instruction, the following example can be adopted:
[0157] #Given test questions
[0158]
Stem of the major question
[0159] {Stem of the major question}
[0160]
Stem
[0161] {Stem of the sub-question}
[0162]
Answer
[0163] {Standard answer}
[0164]
Reference label
[0165] 1. <point>< / point> Is the subscript dot symbol.
[0166] For example: sky <point> null< / point> , indicating that the character "sky" has no subscript dot, and only the character "empty" has a subscript dot.
[0167] 2. Is the underlined symbol.
[0168] For example: vast of the vast sea, indicating that the two characters "vast" have an underline, and the three characters "of the sea" have no underline.
[0169] 3. <wavy>< / wavy> Is the wavy underline symbol.
[0170] For example: <wavy> Azure< / wavy> of the blue sky, indicating that the two characters "blue" have a wavy underline, and the three characters "of the sky" have no wavy underline.
[0171] 4. <udouble>< / udouble> Is the double underline symbol.
[0172] For example: Bees are in <udouble> Dance< / udouble>, indicating that the three characters "bee is here" do not have double underlines, while the two characters "dancing" have double underlines.
[0173] 5. is an italic symbol.
[0174] For example: Cong bright , indicating that the character "Cong" is not italic, only the character "Ming" is italic.
[0175] 6. is a bold symbol.
[0176] For example: Gao prosper , indicating that the character "Gao" is not bold, only the character "Xing" is bold.
[0177] # Task description
[0178] You are a professional primary school Chinese teacher, tutoring students in solving problems and constructing problem-solving ideas. Please follow the "student-teacher" dialogue flow form to simulate the entire tutoring process.
[0179] Please explain according to the four steps of "reading the question" → "analysis" → "answer" → "summary and extension". In each step, there are multiple semantic unit blocks. At the same time, the semantic unit blocks of "heuristic questions" and "discussion dialogues" are interspersed during the tutoring process, requiring the semantic unit blocks to be logically clear, smooth, coherent, and not verbose.
[0180] If reference information is given, please give a reply based on the reference information.
[0181] # Link task
[0182] "Reading the question" task: Read out the current sub-question [question stem] to be tutored and mark the key content in the [question stem].
[0183] "Analysis" task: ① Give the examination point. ② Explain and introduce the problem-solving methods and ideas around the examination point. ③ Highly combine the material with the sub-question and integrate the methods and steps just explained, and lead the students to analyze carefully. Through thinking guidance and heuristic guiding questions at key points, gradually guide out the possible thinking difficulties and key derivation points in the process of deriving the answer in the form of explanation and heuristic questions, and help the students construct a similar derivation logic problem-solving idea and thinking chain.
[0184] "Answer" task: After the analysis step, give the answer.
[0185] "Summary and extension" task: Briefly review the content and methods just explained and conduct appropriate extension of knowledge and methods.
[0186] # Highlight identifier
[0187] <mark id=$number>< / mark> Flag: During the tutoring process, words, fragments, analysis objects, etc. that students need to focus on and understand should be highlighted by adding this tag.
[0188] Note that this tag can only appear in the "cont" tutoring content and "display_cont" display content fields.
[0189] It must be noted that within the same semantic unit block, a group of tags with the same number <mark id=$number>< / mark> must appear in pairs in the "cont" tutoring content and the corresponding "display_cont" display content, and the words marked by this pair may not be exactly the same but must have similar meanings. Simulate the logic of highlighting the corresponding semantic fragment in "display_cont" when the teacher explains the highlighted tag in "cont". $number represents an int type value, starting from 0 and incrementing sequentially throughout the conversation flow.
[0190] It must be noted that in the "analysis section", "answer section", and "summary and expansion section", as many <mark id=$number>< / mark> highlight tags as possible should be used.
[0191] # Material marking symbol
[0192] <material - sentence>< / material - sentence> and <material - word>< / material - word> symbols: During the tutoring process, as much material as possible should be combined for explanation. Simulate the action of the teacher circling and marking in the
Material
[0193] Note that this tag can only appear in the "display_material" material marking content field and cannot appear in the "cont" tutoring content field.
[0194] It must be noted that material marking is divided into two types: sentence type and word type. The <material - sentence>< / material - sentence> tag is used to mark sentence fragments in the
Material
Material
[0195] # Bottom - line requirements
[0196] In the "discussion dialogue", when the teacher asks an "inspiring question", if the student responds to the teacher's question, the teacher responds according to the following words and guides the student back to the main process of auxiliary learning:
[0197] Student: "I don't know" Teacher: "It's okay, kids. Let's continue learning."
[0198] Student: The student's answer to the teacher's question is correct. Teacher: "Little friend, you are great. Let's continue learning."
[0199] Student: The student's answer to the teacher's question is half correct. Teacher: "My dear, you have thought carefully. We can think about it again. Next, we will continue to study."
[0200] Student: The student's answer to the teacher's question is wrong. Teacher: "My dear, you have thought about this question very carefully. We can think about it again. Next, we will continue to study."
[0201] When the teacher asks an "inspiring question", if the student does not actively answer / the student actively interrupts the question, the teacher responds as follows and guides the student back to the main learning process:
[0202] Student: Political and sensitive issues, common application control commands for learning machines such as "open an application, adjust the volume, etc.", insults, small talk, popular science, and other issues not related to the tutoring of this question. Teacher: "Little friend, I am Xiaofei. I am only responsible for tutoring you on the content related to this question. Let's continue learning.
[0203] "
[0204] Students: "Who are you?" and other personality-related questions, the teacher: "Hello, little friend! I am Xiaofei, let's continue learning."
[0205] Student: Review / skip / specify a part / end. Teacher: Just follow the student’s instructions.
[0206] Students: The questions asked by students are related to the materials and the answers can be found in the materials. Teachers: Explain and respond to the questions asked by students.
[0207] Student: The student's question is related to the material, but the answer cannot be found in the material. Teacher: "My dear, I am Xiaofei. I am only responsible for tutoring you on the content related to this question. Let's continue studying.
[0208] "
[0209] #Semantic unit block definition
[0210] The explanation content of a semantic unit block can constitute the basic particle of semantic segmentation. The format of the semantic block is as follows:
[0211] <json>
[0212] {{
[0213] "type":"", / / semantic unit block type. Optional: ["read the question", "analysis", "answer", "summary and expansion", "heuristic question", "discussion and dialogue"]
[0214] "cont":"", / / Explain the content. If there are numbers, symbols, etc. in the transcript, they need to be converted into text that can be read directly.
[0215] "display_cont":"", / / Display content. The blackboard content on the screen is a summary of the "explained content", which requires to be extremely concise. If the type is "discussion dialogue", the display content is empty.
[0216] "display_material":[] / / Material labeling content. Only when the type is "analysis", there is material labeling content, and other types are empty.
[0217] "judge":"" / / Whether the user's response is correct.
[0218] }}
[0219] < / json>
[0220] #Interaction History
[0221] {Interaction History}
[0222] #Teaching and Research Guidance Logic
[0223] {Teaching and Research Guidance Logic}
[0224] #Similar question guide example
[0225] {Similar question guidance example}
[0226] #Reference Information
[0227] {Reference Information}
[0228] #Student current input
[0229] {Student current input}
[0230] #Task Recap
[0231] Please refer to "Segment Tasks", "Highlight Identifiers", "Material Marking Symbols", and "Semantic Unit Block Definitions" to tutor on the "Given Reading Test Questions". Please start tutoring.
[0232] Furthermore, considering that the prompt of the first round of input does not need to be spliced with user input, that is, the first round of input of the large language model is triggered by the test questions, and the model starts the first round of explanation. For non-first round input, it is necessary to splice the interaction history record, that is, considering that the user can interact with the large language model during the question-answering process, such as asking questions, etc., at this time, the changes in the user's learning ideas can be shown, so the first task instruction can also include label introduction information and / or interaction record information; wherein, the label introduction information includes introduction information of the set identification label, which is used to enhance the display of information, such as indicating the highlighted content and the timing of the highlight; and the interaction record information includes the historical interaction content with the user, as well as the user's current input content.
[0233] Specifically, in non-first-round dialogues, in addition to the prompt template in the first task instruction example above, the responses of the big model and the user in each round of Q&A will be spliced together as the input of the big model. For example:
[0234] Teacher: {Model first round response}
[0235] Student: {User's first round of response}
[0236] It is used to generate the second round of responses from the model, and so on. That is, after each user speaks, that is, after each round of obtaining the student's current input, the above prompt will be generated once and input into the big model, so that the big model can generate the output content corresponding to the user input according to the user's latest input and historical interaction in the prompt. Therefore, in each round of inputting the prompt of the auxiliary learning question-answering model, both the interaction history and the student's current input are updated, realizing the dynamic update process of the auxiliary learning question-answering model and improving the accuracy of the auxiliary learning information.
[0237] 303. Input the first task instruction into the pre-trained first large language model, so that the first large language model generates interactive information for guiding the user to correctly answer the test question by executing the first task instruction.
[0238] In this embodiment, based on the configuration of the first task instructions in the above steps, the corresponding auxiliary learning tasks are performed to generate interactive information to guide users to correctly answer the test questions. The specific configuration of the first task instructions may include part or all of the above examples. The specific composition depends on the actual scenario and is not limited here.
[0239] Next, we will explain the scenario of using the first language model as a learning and question-answering model. Figure 4 As shown, Figure 4 A scenario diagram of an auxiliary learning method provided for an embodiment of the present application; the figure shows the module setting of this embodiment based on the auxiliary learning information configuration, that is, to enhance the auxiliary learning ability of the large language model for different questions, in addition to the basic question stem, answer, and user answer information, this embodiment also adds a teaching and research guidance logic extraction sub-module, a similar question retrieval sub-module, and an external knowledge base retrieval sub-module, which are input as supplementary information into the auxiliary learning question-answering large model, so that the model has stronger auxiliary learning ability and overcomes problems such as excessive divergence of model guidance and unfaithfulness to facts.
[0240] It is understandable that the auxiliary learning and answering model responds to the user's (student's) input each time and generates output based on the input prompt, and achieves the purpose of answering questions and guiding students by realizing the interaction between the user and the auxiliary learning and answering model. Among them, each generated output may include one or more semantic unit blocks, and the output of the large model is in the form of semantic unit blocks. The semantic unit blocks are then subjected to speech synthesis, text display, image display, etc., to form interactive content with students, so that students can hear the answer voice of the large model, see the text output by the large model, and see the marking and highlighting of the large model through images.
[0241] Therefore, based on the input corresponding to the first task instruction, the large model returns each semantic unit block one by one, and the explanation content of a semantic unit block can constitute the basic particle of semantic segmentation. Each semantic unit block is an independent JSON structure, and the specific field interpretation is shown in Table 1.
[0242] Table 1 Field interpretation comparison table
[0243]
[0244] Among them, type is the semantic unit block type, which is used to mark the current auxiliary learning link at the front end. The first is the material summary link. If you encounter a reading question, the model will first summarize the content of the reading material and quickly introduce the scenario. Next is the question reading link. The model reads out the current auxiliary learning question word by word and reminds you of the key requirements. Then there is the analysis link, which mainly explains the question in detail, interspersed with heuristic questions and discussion dialogues to control the guidance rhythm. Next is the answer link. The model gradually guides the answer through detailed guidance and shows the standard answer to students. Finally, there is the summary and expansion link. The model summarizes the methods and steps in the guidance process and appropriately expands and elevates them.
[0245] cont is the explanation content, which is used for voice broadcast, leading students to analyze the questions step by step and start the explanation.
[0246] display_cont is the content on the screen, which is used to display key information on the screen while the voice is broadcasting, to help students understand and remember.
[0247] display_material is the marked content of the original picture, which mainly marks the key words in the question stem and the key sentences in the original material that students need to pay attention to.
[0248] In particular, during the learning process, the model will ask some heuristic questions to help students focus and follow the rhythm of the explanation. At this time, the type is "heuristic question". After the user answers, the model will output the judge field to evaluate the user's reply. According to the correctness of the user's answer, the corresponding label is assigned to provide intuitive feedback to the user. For example, if the student answered the previous round of questions completely correctly, it should be rated as correct; or if the student does not cooperate with the communication and replies with other irrelevant content, it should be rated as wrong.
[0249] In addition, the keywords of the upper screen content are highlighted, that is, during the explanation process, sometimes the key content of the upper screen needs to be highlighted. <mark>Label (label introduction information) is used to indicate the highlighted content and the timing of highlighting. The following is an example:
[0250] Original question: The following options are incorrect ()
[0251] Transcript: Classmates, let's look at this question together. The question asks the following options <mark id=0> Incorrect< / mark> Which one is it?
[0252] Upper screen content: The following options <mark id=0> Incorrect< / mark> is()
[0253] Among them, the transcript <mark>< / mark> The marked part is the time to highlight. <mark>< / mark> The marked part is the content that needs to be highlighted when the corresponding timing is reached. That is, the label introduction information will be directly carried in the specific content of the semantic unit block json structure display_cont output by the large model, so that when the specific content of display_cont is output or rendered, it can be determined according to this label which content to highlight.
[0254] Furthermore, the original diagram can be marked, that is, during the explanation, sometimes it is necessary to mark the question information that students need to pay extra attention to on the original diagram. <material>Label (identifying introduction information), used to indicate the material content that needs to be marked in the original figure when the current semantic unit is broadcast. The example is as follows:
[0255] Verbatim transcript: In the original text, the description of Sima Zhao's language shows his intelligent characteristics.
[0256] Material marking content: <material-sentence>The second son, Sima Zhao, said, "Could it be that Zhuge Liang has no army and is just putting on this show? Why does father want to retreat the troops?"< / material-sentence>
[0257] Therefore, this label introduction information is carried in the specific content of the semantic unit block json structure display_material output by the large model. Thus, when outputting or rendering the specific content of display_material, it is possible to determine which content to mark according to this label.
[0258] Combined with the above embodiments, the content corresponding to the auxiliary learning information can be displayed in Figure 5 the interface shown, Figure 5 which is a schematic diagram of the scenario of another auxiliary learning method provided by the embodiments of this application; that is, while showing the questions and answers, the auxiliary learning content is shown. This auxiliary learning content is obtained by the auxiliary learning Q&A large model based on any one of the above embodiments. That is, the content shown in the figure is implemented by two modules: answer acquisition and auxiliary learning Q&A. The answer acquisition module is responsible for using the large model to automatically answer questions and obtain the standard answers for each sub-question and each answering area. This way can accurately obtain the standard answers while establishing structured information for the answers, laying a foundation for the subsequent auxiliary learning Q&A module. The auxiliary learning Q&A module mainly has three capabilities: material general idea summarization, auxiliary learning Q&A, and user free interruption reply. Among them, material general idea summarization mainly involves reading questions, and the model will summarize the main writing content and writing theme of the material to help students quickly establish a preliminary understanding of the reading questions. Auxiliary learning Q&A is the main function, and the model conducts auxiliary learning Q&A in the form of semantic unit blocks through comprehensive teaching and research guiding logic. Specifically in the application scenario, the auxiliary learning process of a real teacher will mainly be voice explanations, and at the same time, the summary content of the explanation process will be written on the blackboard, and key points that students need to pay attention to will also be marked by ticking and circling in the original questions. Imitating the explanation process of a real teacher, we define a semantic unit block as a semantic unit that includes a verbatim transcript to be read out word by word, display content to be shown on the screen, and material content to be marked in the original figure. The model generates a complete semantic unit block each time, so as to achieve real-time interaction with students. In addition to the teacher-led auxiliary learning Q&A, students may also freely interrupt the tutoring process and ask questions actively. We use a large model enhanced by combining plugins to reply to the user's questions and guide the students back to the normal auxiliary learning process after the reply.
[0259] The above-mentioned assisted learning process based on the auxiliary question-answering model starts from the user's test questions, guides through teaching and research logic, introduces heuristic questions, helps students learn why they answer in this way, and learns to draw inferences through knowledge expansion, thereby improving their learning ability.
[0260] In combination with the above embodiments, it can be seen that by further triggering the large language model to generate interactive information to guide users to correctly answer the test questions based on obtaining the test questions, the test answers, and the auxiliary learning information of the test questions, and by triggering the large language model to interact with the user, the user can be guided to answer the test questions step by step, so that the user can understand the test answering process and answering methods, which helps the user to fundamentally learn knowledge and master the answering methods.
[0261] In order to better implement the above solution of the embodiment of the present application, the following also provides related devices for implementing the above solution. Figure 6 , Figure 6 The structure diagram of an auxiliary learning device provided in an embodiment of the present application is as follows. The auxiliary learning device 600 includes:
[0262] An acquisition unit 601 is used to acquire a test question, an answer to the test question, and supplementary learning information of the test question, wherein the supplementary learning information includes at least one of a teaching and research guidance logic, supplementary learning records of similar questions, and related knowledge;
[0263] The processing unit 602 is used to generate a first task instruction based on the test question, the answer to the test question and the auxiliary learning information of the test question, wherein the first task instruction is used to instruct the large language model to generate interactive information for guiding the user to correctly answer the test question;
[0264] The learning unit 603 is used to input the first task instruction into a pre-trained first large language model, so that the first large language model generates interactive information for guiding the user to correctly answer the test question by executing the first task instruction.
[0265] Optionally, in some possible implementations of the present application, the acquisition unit 601 is specifically configured to answer the test question by using a pre-trained second language model to obtain an answer to the test question;
[0266] The second largest language model is obtained by training test questions and answer data.
[0267] Optionally, in some possible implementations of the present application, the acquisition unit 601 is specifically used to generate a second task instruction based on the test question, and the second task instruction is used to instruct the large language model to answer the test question and output the answer result to the test question according to the set answer output requirement;
[0268] The acquisition unit 601 is specifically configured to input the second task instruction into a pre-trained second largest language model, so that the second largest language model obtains the answer to the test question by executing the second task instruction.
[0269] Optionally, in some possible implementations of the present application, the acquisition unit 601 is specifically used to determine the test point category corresponding to the test question;
[0270] The acquisition unit 601 is specifically used to determine the teaching and research guidance logic corresponding to the test point category from a preset guidance logic knowledge base based on the test point category corresponding to the test question; wherein the guidance logic knowledge base stores the teaching and research guidance logic corresponding to various test point categories.
[0271] Optionally, in some possible implementations of the present application, the acquisition unit 601 is specifically used to retrieve, based on the test question, supplementary learning records of similar test questions to the test question from a preset supplementary learning record database; wherein the supplementary learning record database stores supplementary learning record data for multiple test questions.
[0272] Optionally, in some possible implementations of the present application, the acquisition unit 601 is specifically used to retrieve knowledge related to the test question from a preset knowledge base.
[0273] Optionally, in some possible implementations of the present application, the learning unit 603 is specifically used to obtain the test question, the answer to the test question, the auxiliary learning information of the test question, the introduction information of the auxiliary learning task, and the introduction information of the interactive information output requirement;
[0274] Among them, the auxiliary learning task is used to interact with the user according to the set interaction process and interaction requirements, and guide the user to correctly answer the test questions; the interaction information output requirements include requirements for the interaction information generated by the large language model output.
[0275] Optionally, in some possible implementations of the present application, the learning unit 603 is specifically used to obtain the introduction information of the auxiliary learning task, including: at least one of the description of the auxiliary learning task and the introduction of the auxiliary learning link;
[0276] and / or,
[0277] The interactive information output requirement is obtained, including at least one of the following: interactive information highlighting requirement, material marking requirement, interactive dialogue bottom line requirement, interactive information semantic unit block definition, and interactive information semantic unit block output requirement.
[0278] Optionally, in some possible implementations of the present application, the learning unit 603 is specifically configured to determine that the first task instruction also includes label introduction information and / or interaction record information;
[0279] The label introduction information includes introduction information of the set identification label;
[0280] The interaction record information includes historical interaction content with the user and the user's current input content.
[0281] By obtaining the test questions, answers, and auxiliary learning information of the test questions, the large language model is further triggered to generate interactive information to guide users to correctly answer the test questions. By triggering the large language model to interact with the user, the user can be guided to answer the test questions step by step, so that the user can understand the test answering process and answering methods, which helps users to fundamentally learn knowledge and master answering methods.
[0282] The auxiliary learning device provided in this embodiment belongs to the same application concept as the method provided in the above embodiments of this application, can execute the auxiliary learning method provided in any of the above embodiments of this application, and has the corresponding functional modules and beneficial effects of executing the auxiliary learning method. For technical details not fully described in this embodiment, please refer to the specific processing content of the auxiliary learning method provided in the above embodiments of this application, which will not be repeated here.
[0283] The functions implemented by the above acquisition unit 601, processing unit 602 and learning unit 603 can be implemented by the same or different processors respectively, and the embodiments of the present application are not limited thereto.
[0284] It should be understood that each unit in the above device can be implemented in the form of a processor calling software. For example, the device includes a processor, the processor is connected to a memory, and instructions are stored in the memory. The processor calls the instructions stored in the memory to implement any of the above methods or realize the functions of each unit of the device, wherein the processor can be a general-purpose processor, such as a CPU or a microprocessor, etc., and the memory can be a memory in the device or a memory outside the device. Alternatively, the unit in the device can be implemented in the form of a hardware circuit, and the functions of some or all units can be realized by designing the hardware circuit. The hardware circuit can be understood as one or more processors; for example, in one implementation, the hardware circuit is an ASIC, and the functions of some or all units above are realized by designing the logical relationship of the components in the circuit; for another example, in another implementation, the hardware circuit can be implemented by PLD, taking FPGA as an example, which can include a large number of logic gate circuits, and the connection relationship between the logic gate circuits is configured by a configuration file, so as to realize the functions of some or all units above. All units of the above device can be implemented in the form of a processor calling software, or in the form of a hardware circuit, or in the form of a processor calling software, and the remaining part can be implemented in the form of a hardware circuit.
[0285] In an embodiment of the present application, a processor is a circuit with the ability to process signals. In one implementation, the processor may be a circuit with the ability to read and run instructions, such as a CPU, a microprocessor, a GPU, or a DSP; in another implementation, the processor may implement certain functions through the logical relationship of a hardware circuit, and the logical relationship of the hardware circuit is fixed or reconfigurable, such as a hardware circuit implemented by an ASIC or PLD, such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document to implement the hardware circuit configuration can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as an NPU, TPU, DPU, etc.
[0286] It can be seen that each unit in the above device can be one or more processors (or processing circuits) configured to implement the above method, such as: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.
[0287] In addition, all or part of the units in the above device can be integrated together, or can be implemented independently. In one implementation, these units are integrated together and implemented in the form of a SOC. The SOC may include at least one processor for implementing any of the above methods or implementing the functions of each unit of the device. The type of the at least one processor may be different, for example, including a CPU and an FPGA, a CPU and an artificial intelligence processor, a CPU and a GPU, etc.
[0288] The present application also provides an auxiliary learning device, which includes a processor and an input-output component. The input-output component is used to obtain an image to be processed;
[0289] The processor is used to answer the target question in the image to be processed obtained by the input-output component by executing any one of the auxiliary learning methods in the above embodiments.
[0290] The above-mentioned interface circuit can be any interface circuit that can realize the data communication function, for example, it can be a USB interface circuit, a Type-C interface circuit, a serial port circuit, a PCIE circuit, etc.
[0291] Optionally, the embodiment of the present application further provides a system, wherein the interactive client is used to obtain a question and send the question to the server, and to display the auxiliary learning content output by the server;
[0292] The server is used to answer the questions obtained by the interactive client and generate auxiliary learning content by executing the auxiliary learning method described in any one of the above embodiments.
[0293] Another embodiment of the present application also provides an auxiliary learning device, see Figure 7 As shown, the device includes:
[0294] Memory 700 and processor 710;
[0295] The memory 700 is connected to the processor 710 and is used to store programs;
[0296] The processor 710 is used to implement the auxiliary learning method disclosed in any of the above embodiments by running the program stored in the memory 700.
[0297] Specifically, the auxiliary learning device may further include: a bus, a communication interface 720 , an input device 730 and an output device 740 .
[0298] The processor 710, the memory 700, the communication interface 720, the input device 730 and the output device 740 are connected to each other via a bus.
[0299] A bus may include a pathway that transfers information between components of a computer system.
[0300] Processor 710 may be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the scheme of the present invention. It may 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.
[0301] The processor 710 may include a main processor, and may also include a baseband chip, a modem, and the like.
[0302] The memory 700 stores a program for executing the technical solution of the present invention, and may also store an operating system and other key services. Specifically, the program may include a program code, and the program code includes a computer operation instruction. More specifically, the memory 700 may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk storage, a flash, and the like.
[0303] The input device 730 may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.
[0304] Output device 740 may include devices that allow information to be output to a user, such as a display screen, printer, speaker, etc.
[0305] The communication interface 720 may include any transceiver or the like to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.
[0306] The processor 710 executes the program stored in the memory 700 and calls other devices, which can be used to implement each step of any auxiliary learning method provided in the above embodiments of the present application.
[0307] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the assisted learning method according to various embodiments of the present application described in any of the above-mentioned embodiments of this specification.
[0308] The computer program product may be written in any combination of one or more programming languages to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user computing device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.
[0309] In addition, the embodiment of the present application may also be a storage medium on which a computer program is stored. The computer program is executed by a processor to execute the steps of the auxiliary learning method according to various embodiments of the present application described in any of the above embodiments of this specification, and specifically the following steps may be implemented:
[0310] Acquire a test question, an answer to the test question, and supplementary learning information of the test question, wherein the supplementary learning information includes at least one of teaching and research guidance logic, supplementary learning records of similar questions, and related knowledge;
[0311] Based on the test question, the answer to the test question, and the auxiliary learning information of the test question, a first task instruction is generated, where the first task instruction is used to instruct the large language model to generate interactive information to guide the user to correctly answer the test question;
[0312] The first task instruction is input into a pre-trained first large language model, so that the first large language model generates interactive information for guiding the user to correctly answer the test question by executing the first task instruction.
[0313] For the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, some 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 preferred embodiments, and the actions and modules involved are not necessarily required by the present application.
[0314] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0315] The steps in the methods of each embodiment of the present application can be adjusted in order, combined and deleted according to actual needs, and the technical features recorded in each embodiment can be replaced or combined.
[0316] The modules and sub-modules in the devices and terminals of the various embodiments of the present application can be combined, divided and deleted according to actual needs.
[0317] In the several embodiments provided in the present application, it should be understood that the disclosed terminals, devices and methods can be implemented in other ways. For example, the terminal embodiments described above are only schematic, for example, the division of modules or submodules is only a logical function division, and there may be other division methods in actual implementation, for example, multiple submodules or modules can be combined or integrated into another module, 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 interfaces, devices or modules, which can be electrical, mechanical or other forms.
[0318] The modules or submodules described as separate components may or may not be physically separated, and the components of the modules or submodules may or may not be physical modules or submodules, that is, they may be located in one place, or they may be distributed on multiple network modules or submodules. Some or all of the modules or submodules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0319] In addition, each functional module or submodule in each embodiment of the present application may be integrated into one processing module, or each module or submodule may exist physically separately, or two or more modules or submodules may be integrated into one module. The above-mentioned integrated modules or submodules may be implemented in the form of hardware or in the form of software functional modules or submodules.
[0320] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0321] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly by hardware, software units executed by a processor, or a combination of the two. The software units may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
[0322] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a..." does not exclude the presence of other identical elements in the process, method, article or device including the element.
[0323] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.< / material> < / mark>
Claims
1. An auxiliary learning method, characterized in that: include: Acquire a test question, an answer to the test question, and supplementary learning information of the test question, wherein the supplementary learning information includes at least one of teaching and research guidance logic, supplementary learning records of similar questions, and related knowledge; Based on the test question, the answer to the test question, and the auxiliary learning information of the test question, a first task instruction is generated, where the first task instruction is used to instruct the large language model to generate interactive information to guide the user to correctly answer the test question; The first task instruction is input into a pre-trained first large language model, so that the first large language model generates interactive information for guiding the user to correctly answer the test question by executing the first task instruction.
2. The method according to claim 1, characterized in that Get the answers to the test questions, including: Solve the test question by using the pre-trained second largest language model to obtain the answer to the test question; The second largest language model is obtained by training test questions and answer data.
3. The method according to claim 2, characterized in that The step of answering the test question by using the pre-trained second largest language model to obtain the answer to the test question includes: generating a second task instruction based on the test question, wherein the second task instruction is used to instruct the large language model to answer the test question and output the answer result to the test question according to the set answer output requirement; The second task instruction is input into a pre-trained second largest language model, so that the second largest language model obtains the answer to the test question by executing the second task instruction.
4. The method according to claim 1, characterized in that: In the case where the auxiliary learning information includes teaching and research guidance logic, obtaining the auxiliary learning information of the test question includes: Determine the test point category corresponding to the test question; Based on the test point category corresponding to the test question, the teaching and research guidance logic corresponding to the test point category is determined from a preset guidance logic knowledge base; wherein the guidance logic knowledge base stores the teaching and research guidance logic corresponding to various test point categories.
5. The method according to claim 1, characterized in that In the case where the supplementary learning information includes supplementary learning records of similar questions, obtaining the supplementary learning information of the test question includes: Based on the test question, the supplementary learning records of the test questions similar to the test question are retrieved from a preset supplementary learning record database; wherein the supplementary learning record database stores supplementary learning record data of a plurality of test questions.
6. The method according to claim 1, characterized in that In the case where the supplementary learning information includes relevant knowledge, obtaining the supplementary learning information of the test question includes: The knowledge related to the test question is retrieved from a preset knowledge base.
7. The method according to any one of claims 1 to 6, characterized in that: The first task instruction includes: The test questions, answers to the test questions, supplementary learning information of the test questions, introduction information of supplementary learning tasks, and introduction information of interactive information output requirements; Among them, the auxiliary learning task is used to interact with the user according to the set interaction process and interaction requirements, and guide the user to correctly answer the test questions; the interaction information output requirements include requirements for the interaction information generated by the large language model output.
8. The method according to claim 7, characterized in that The introduction information of the auxiliary learning task includes at least one of: a description of the auxiliary learning task and an introduction to the auxiliary learning link; and / or, The interactive information output requirements include at least one of the following: interactive information highlighting requirements, material marking requirements, interactive dialogue bottom line requirements, interactive information semantic unit block definitions, and interactive information semantic unit block output requirements.
9. The method according to claim 7, characterized in that: The first task instruction also includes tag introduction information and / or interaction record information; The label introduction information includes introduction information of the set identification label; The interaction record information includes historical interaction content with the user and the user's current input content.
10. A learning assistance device, characterized in that: include: An acquisition unit, used to acquire a test question, an answer to the test question, and supplementary learning information of the test question, wherein the supplementary learning information includes at least one of teaching and research guidance logic, supplementary learning records of similar questions, and related knowledge; A processing unit, configured to generate a first task instruction based on the test question, the answer to the test question, and the auxiliary learning information of the test question, wherein the first task instruction is used to instruct the large language model to generate interactive information for guiding the user to correctly answer the test question; The learning unit is used to input the first task instruction into a pre-trained first large language model, so that the first large language model generates interactive information to guide the user to correctly answer the test question by executing the first task instruction.
11. A learning assistance device, characterized in that: include: An input-output component, and a processor connected to the input-output component; Wherein, the input-output component is used to receive data input by the user and output data sent by the processor; The processor is used to obtain the test questions input by the user through the input-output component, and generate interactive information to guide the user to correctly answer the test questions by executing the auxiliary learning method as described in any one of claims 1 to 9, and output the interactive information through the input-output component.
12. An auxiliary learning system, characterized in that: include: A human-computer interaction client, and a server connected to the human-computer interaction client for communication; The human-computer interaction client is used to receive data input by the user and output data sent by the processor; The server is used to obtain the test questions input by the user through the input-output component, and generate interactive information to guide the user to correctly answer the test questions by executing the auxiliary learning method as described in any one of claims 1 to 9, and output the interactive information through the input-output component.
13. An electronic device, characterized in that: include: Memory and processor; The memory is connected to the processor and is used to store programs; The processor is used to implement the auxiliary learning method as described in any one of claims 1 to 9 by running the program in the memory.
14. A computer program product, characterized in that The method comprises computer program instructions, which, when executed by a processor, enable the processor to perform the assisted learning method according to any one of claims 1 to 9.
Citation Information
Patent Citations
Auxiliary learning method and device
CN107492275A
Natural language problem processing method, electronic equipment and computer readable storage medium
CN117764164A
Knowledge question and answer method, device and equipment and readable storage medium
CN117786068A
Auxiliary learning method and device, electronic equipment and storage medium
CN117877336A
Calculation question answering and model training method and device, equipment and medium
CN118133009A
Cited By
Test question auxiliary learning explanation method and device, related equipment and computer program product
CN119990146A
Method and device for training large language model, equipment and medium
CN120725068A
Information marking method and device based on large model and information marking method and device
CN121033881A
