A knowledge quiz game content generation method, device and electronic equipment

Abstract

The application provides a knowledge quiz game content generation method and device and electronic equipment, and relates to the field of data processing. In the method, target knowledge data corresponding to a knowledge quiz game is obtained; a semantic network is determined according to the target knowledge data; a first question is generated based on the semantic network by using multiple agents; a first answer corresponding to the first question is obtained by a user; the first answer is evolved and a second question is output; and the knowledge quiz game content is obtained through the first question, the first answer and the second question. The technical solution provided by the application facilitates improving the intelligence of knowledge quiz game content generation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to a knowledge quiz game content generation method and device and electronic equipment. BACKGROUND

[0002] As an interactive form combining entertainment and education, knowledge quiz games have gained widespread attention and application worldwide in recent years. It not only occupies an important position in the entertainment industry, but also shows great potential in education, training, scientific research and other fields.

[0003] At present, most existing knowledge quiz games rely on pre-set fixed question banks and logical processes for question generation and problem answering. In this traditional model, the generation of questions and answers is usually static, based on fixed rules or database content. During the game process, the player's choice only determines the order of answering, and the system has almost no real-time feedback mechanism for question generation and adjustment. Therefore, the intelligence of generating knowledge quiz game content in the prior art is low.

[0004] Therefore, there is an urgent need for a knowledge quiz game content generation method, device and electronic equipment. SUMMARY

[0005] The present application provides a knowledge quiz game content generation method, device and electronic equipment, which facilitates to improve the intelligence of generating knowledge quiz game content.

[0006] In a first aspect of the present application, a knowledge quiz game content generation method is provided, which comprises: acquiring target knowledge data corresponding to a knowledge quiz game; determining a semantic network according to the target knowledge data; generating a first question based on the semantic network using multiple agents; acquiring a first answer corresponding to the first question from a user; performing feedback evolution on the first answer to output a second question; and obtaining knowledge quiz game content through the first question, the first answer and the second question.

[0007] By using the above technical solution, first, the semantic network is constructed using the target knowledge data, which can ensure that the generated questions are highly relevant to the knowledge content, enhancing the educational nature and effectiveness of the game. By generating questions through multiple agents, the intelligence and dynamic adjustment of the questions can be realized, avoiding the limitations of traditional question banks, thereby improving the personalization and adaptability of the questions. Further, through the feedback evolution mechanism, the subsequent questions are adjusted in real time according to the user's answers, ensuring that the difficulty of the questions matches the user's knowledge level, and improving the user experience. In summary, this method can provide accurate learning content while maintaining the interactivity, challenge and entertainment of the game, thereby improving the player's participation and learning effect. Therefore, it is convenient to improve the intelligence of generating knowledge quiz game content.

[0008] Optionally, the determining the semantic network according to the target knowledge data specifically comprises: performing text splitting on the target knowledge data to obtain target knowledge cells; determining relationships between the target knowledge cells to obtain semantic relationships; vectorizing the target knowledge cells to obtain a first target knowledge cell and a second target knowledge cell, the first target knowledge cell and the second target knowledge cell being any two target knowledge cells in the target knowledge cells; calculating a cell similarity between the first target knowledge cell and the second target knowledge cell; and constructing the semantic network according to the cell similarity and the semantic relationships.

[0009] By adopting the above technical solution, firstly, by performing text splitting on the target knowledge data, the complex knowledge is converted into more fine-grained target knowledge cells, which can make the knowledge representation more fine and facilitate subsequent processing and analysis. Secondly, by determining the relationships between the target knowledge cells and vectorizing, the semantic relevance between different knowledge points can be accurately captured, thereby providing a solid foundation for subsequent reasoning and question generation. In particular, by calculating the cell similarity, the similarity between different knowledge cells can be revealed, helping the model better understand the relationships between knowledge points, and constructing a high-quality semantic network according to these similarities and semantic relationships, improving the relevance and intelligence of question generation. Finally, this method can ensure that the generated questions and answers are more logical, targeted and personalized, effectively improving the educational nature and user experience of the game, and also providing a solid data support for subsequent intelligent feedback evolution.

[0010] Optionally, the generating the first question based on the semantic network by using multiple agents specifically comprises: extracting a knowledge core according to the semantic network; controlling an expert agent to generate a first draft based on a large language model according to the knowledge core; controlling a student agent to generate a second draft based on a prompt word algorithm according to the knowledge core; performing semantic fusion on the first draft and the second draft to obtain a third draft; and generating the first question by adjusting the third draft through a checking agent, the multiple agents including the expert agent, the student agent and the checking agent.

[0011] By adopting the technical scheme, the generated question can be ensured to closely match the core knowledge points by combining the semantic network, making the question more accurate and relevant. The expert agent and the student agent generate drafts based on the large language model and the prompt word algorithm, respectively, which can realize knowledge expression from multiple angles and different levels, thereby ensuring the diversity and richness of the question. The semantic fusion can combine the advantages of the two drafts to optimize the question content and improve the quality of the question. Finally, the verification agent adjusts the draft to ensure that the generated question is logically strong and accurate, thereby avoiding the bias or deficiency of a single agent. This multi-agent collaboration method not only enhances the flexibility and intelligence of question generation, but also greatly improves the quality and adaptability of the question, making the game content more suitable for players' needs and improving the user experience and learning effect.

[0012] Optionally, the first answer is fed back and evolved to output a second question, specifically including: obtaining a preset answer corresponding to the first question; using a BERT model to determine a preset answer vector corresponding to the preset answer; using the BERT model to determine a first answer vector corresponding to the first answer; calculating the answer similarity between the preset answer vector and the first answer vector; if it is determined that the answer similarity is greater than a preset threshold, obtaining a first keyword from the first answer; comparing the first keyword with a preset keyword library to generate a comparison result, the preset keyword library being constructed according to the preset answer; determining the second question according to the comparison result.

[0013] By adopting the technical scheme, the BERT model is used to vectorize the preset answer and the first answer, and the answer similarity between them is calculated, which can accurately evaluate the matching degree between the player's answer and the correct answer. This method avoids the limitations of traditional systems relying on fuzzy matching or simple comparison through quantitative similarity calculation, improving the accuracy and reliability of feedback. If the player's answer is close to the preset answer, the system will further extract the keyword and compare it with the preset keyword library, which not only helps to confirm the key points of the answer, but also refines the depth and range of the question. Finally, the second question generated based on these feedback and analysis results can better adapt to the player's answering state and knowledge mastery, making the difficulty and content of the question more personalized, thereby enhancing the player's sense of participation and learning effect. This intelligent feedback mechanism not only improves the interactivity of the game, but also effectively promotes the realization of personalized learning.

[0014] Optionally, the determining the second question according to the comparison result specifically comprises: if it is determined that the comparison result indicates that the number of identical keywords between the first keyword and the preset keyword library does not meet the preset number, determining the second question as a complementary question based on the first question; if it is determined that the comparison result indicates that the number of identical keywords between the first keyword and the preset keyword library meets the preset number, determining the second question as an extension question based on the semantic network.

[0015] By adopting the above technical solution, the generation of the second question is more intelligent and personalized through accurate keyword matching and similarity analysis. By comparing the matching degree of the first keyword and the preset keyword library, the system can dynamically adjust the type of the question according to the player's answer performance. If the player's answer fails to cover enough core concepts or knowledge points, the system will generate a complementary question to help the player further master the omitted knowledge; when the answer is more accurate, the system will construct an extension question to guide the player to think deeply and expand the knowledge. This dynamic adjustment based on keyword matching and semantic network not only improves the relevance and pertinence of the question, but also ensures that the player can obtain timely feedback and challenges in the learning process, thereby increasing the depth and breadth of learning and effectively improving the educational and interactive nature of the game.

[0016] Optionally, the method further comprises: obtaining a second answer corresponding to the second question from the user; determining a first accuracy rate corresponding to the first answer and a second accuracy rate corresponding to the second answer; and if it is determined that both the first accuracy rate and the second accuracy rate are higher than a preset accuracy rate, generating a third question, the difficulty of the third question being greater than the difficulty of each of the first question and the second question.

[0017] By adopting the above technical solution, the intelligent and personalized experience of the knowledge quiz game is effectively improved through accuracy evaluation and dynamic difficulty adjustment mechanism. By comparing the accuracy of the first answer and the second answer, the system can objectively evaluate the player's performance in the two rounds of answering. If the player's answer accuracy is high, the system will automatically generate a third question with higher difficulty, which not only meets the player's demand for challenge, but also promotes the player to continuously improve the knowledge level and avoids falling into repetitive and low difficulty questions. Through this performance-based difficulty progression design, the game can dynamically adjust the difficulty according to the player's mastery, ensuring that each player can challenge within the appropriate difficulty range, thereby maintaining the interest and educational effect of the game, and stimulating the player's learning interest and sense of achievement. This mechanism not only improves the interactivity, but also ensures the realization of personalized learning path.

[0018] Optionally, the method further comprises: obtaining training information, the training information comprising training answers and training answer vectors; inputting the training information into the adaptive word vector fusion network for training to obtain a first training result; obtaining a second training result after superimposing and standardizing the first training result and the training information; inputting the second training result into the adaptive word vector fusion network for processing to obtain a third training result; superimposing and standardizing the third training result and the second training result until a training information similarity matrix is output, to obtain the BERT model, the training information similarity matrix satisfying a preset logistic regression condition.

[0019] By adopting the above technical solution, the adaptive word vector fusion network training method has significant advantages and can effectively improve the answer processing accuracy and model performance in the knowledge question and answer game. By repeatedly inputting the training information into the adaptive word vector fusion network for iterative training, the system can gradually optimize the training result, ensuring that the representation of the word vector after multiple rounds of fusion and standardization is more accurate and consistent. This step-by-step superposition and standardization process helps to eliminate noise and improve the stability of vector representation, so that the generated similarity matrix can accurately reflect the semantic relationship between answers. Finally, by satisfying the preset logistic regression condition, the BERT model can better capture the subtle differences and correlations of answers, thereby improving the intelligence and accuracy of question generation and answer matching in the knowledge question and answer game. This method not only enhances the adaptive ability of the model, but also ensures that the game content can accurately adapt to the knowledge level and answer performance of the players in complex semantic reasoning and multiple rounds of interaction.

[0020] In a second aspect of the present application, a knowledge question and answer game content generation device is provided, which comprises an acquisition module and a processing module. The acquisition module is configured to acquire target knowledge data corresponding to a knowledge question and answer game. The processing module is configured to determine a semantic network based on the target knowledge data. The processing module is further configured to generate a first question based on the semantic network using multiple agents. The acquisition module is further configured to acquire a first answer corresponding to the first question from a user. The processing module is configured to perform feedback evolution on the first answer and output a second question. The processing module is configured to obtain knowledge question and answer game content based on the first question, the first answer, and the second question.

[0021] In a third aspect of the present application, an electronic device is provided, which comprises a processor, a memory, a user interface, and a network interface. The memory is configured to store instructions. The user interface and the network interface are both configured to communicate with other devices. The processor is configured to execute the instructions stored in the memory to enable the electronic device to perform the method described above.

[0022] In a fourth aspect of the present application, a computer-readable storage medium is provided, which stores instructions that, when executed, perform the method described above.

[0023] In summary, the one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0024] Firstly, the semantic network is constructed using target knowledge data, which can ensure that the generated questions are highly relevant to the knowledge content, enhancing the educational nature and effectiveness of the game. Through the collaborative generation of questions by multiple agents, the intelligentization and dynamic adjustment of questions can be achieved, avoiding the limitations of traditional question banks, thereby improving the personalization and adaptability of questions. Further, through the feedback evolution mechanism, the subsequent questions are adjusted in real time according to the user's answers, ensuring that the question difficulty matches the user's knowledge mastery level, improving the user experience. In summary, this method can provide accurate learning content while maintaining the interactivity, challenge and entertainment of the game, thereby improving the player's participation and learning effect. Therefore, it is convenient to improve the intelligence of generating knowledge question and answer game content. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 A flowchart of a knowledge answer game content generation method provided by an embodiment of the present application is shown;

[0026] Figure 2 Another flowchart of a knowledge answer game content generation method provided by an embodiment of the present application is shown;

[0027] Figure 3 A module diagram of a knowledge answer game content generation device provided by an embodiment of the present application is shown;

[0028] Figure 4 A structural diagram of an electronic device provided by an embodiment of the present application is shown.

[0029] Explanation of reference signs: 31, acquisition module; 32, processing module; 41, processor; 42, communication bus; 43, user interface; 44, network interface; 45, memory. DETAILED DESCRIPTION

[0030] In order for those skilled in the art to better understand the technical solutions in the present specification, the technical solutions in the present specification will be described clearly and completely in the following with reference to the drawings in the embodiments of the present specification. Obviously, the described embodiments are only a part of the embodiments of the present application, not all embodiments.

[0031] In the description of the embodiments of the present application, the words such as "for example" or "for instance" are used to mean that an example, illustration or description is given. Any embodiment or designations described as "for example" or "for instance" in the embodiments of the present application should not be construed as more preferred or advantageous than other embodiments or designations. Rather, the use of the words such as "for example" or "for instance" is intended to present the relevant concept in a specific manner.

[0032] In the description of the embodiments of the present application, the term "a plurality of" means two or more. For example, a plurality of systems means two or more systems, and a plurality of screen terminals means two or more screen terminals. In addition, the terms "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly indicating the technical features indicated. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more features. The terms "include", "contain", "have" and their variants mean "include but are not limited to", unless otherwise specifically emphasized.

[0033] As an interactive form combining entertainment and education, knowledge quiz games have gained widespread attention and application worldwide in recent years. It not only occupies an important position in the entertainment industry, but also shows great development potential in education, training and scientific research and other fields.

[0034] However, most of the current knowledge quiz games still rely on pre-set fixed question banks and logical processes to generate questions and answers. The questions and answers in these traditional models are usually static, based on fixed rules or database content for processing. During the game, the player's choice only determines the order of answering, and the system lacks real-time feedback mechanism for question generation and adjustment. Therefore, in the prior art, the knowledge quiz game performs weakly in intelligence and personalization, and cannot make flexible adjustments according to the real-time performance and needs of the players.

[0035] To solve the above technical problems, the present application provides a knowledge quiz game content generation method, referring to Figure 1 , Figure 1 A flowchart of a knowledge quiz game content generation method provided by an embodiment of the present application. The method is applied to a server and includes steps S110 to S160, as follows:

[0036] S110, acquiring target knowledge data corresponding to the knowledge quiz game.

[0037] Specifically, the server, as the core part of the system, is responsible for communicating with external resources and obtaining the required data. Here, the server is the subject responsible for extracting data from a certain knowledge source. This data can come from local databases, online open APIs, or specific educational repositories, etc. In a knowledge quiz game, the target knowledge data is usually the basis of the game content. For example, each question in the game may involve a specific knowledge point (such as historical events, mathematical theorems, geographical knowledge, etc.), which constitutes the core of the game. Target knowledge data refers to specific information required in a knowledge quiz game, which can be facts, definitions, formulas, historical events, scientific theories, etc. These data are the basis for generating relevant questions and answers.

[0038] S120, determining a semantic network according to the target knowledge data.

[0039] Specifically, in a knowledge quiz game, the server constructs a semantic network by processing the target knowledge data obtained. The semantic network is a structure that represents knowledge points and their relationships, which helps the system understand the connections between different knowledge points and supports more intelligent question generation, answer matching, etc. The server obtains specific knowledge data from a certain data source (such as a database or an API). In this stage, the server has obtained knowledge information about a specific field or topic. For example, historical data about World War II, biological data about photosynthesis, etc. The semantic network is a graphical representation of knowledge structure, where nodes represent knowledge units (such as concepts, events, terms, etc.), and edges represent the relationships between these knowledge units (such as cause-and-effect relationship, belonging relationship, association relationship, etc.). The semantic network constructs the connections between knowledge points through nodes and edges, helping the system better understand and reason. Based on the target knowledge data obtained, the server analyzes and determines the relationships between these knowledge points. Then, the system connects the knowledge points through semantic relationships (such as "definition", "contain", "happen before and after", "influence", etc.) to form a semantic network. This semantic network is the basis for subsequent question generation, answer recommendation, etc.

[0040] In one possible implementation, the semantic network is determined according to the target knowledge data, specifically including: performing text splitting on the target knowledge data to obtain target knowledge cells; determining the relationships between the target knowledge cells to obtain semantic relationships; vectorizing the target knowledge cells to obtain a first target knowledge cell and a second target knowledge cell, the first target knowledge cell and the second target knowledge cell being any two target knowledge cells in the plurality of target knowledge cells; calculating the cell similarity between the first target knowledge cell and the second target knowledge cell; and constructing the semantic network according to the cell similarity and the semantic relationships.

[0041] Specifically, text splitting refers to breaking down the target knowledge data (usually raw text or structured information) into smaller, more basic units called knowledge cells. These knowledge cells can be terms, concepts, definitions, events, formulas, etc., which are the smallest blocks that make up the knowledge system. For example, assuming the target knowledge data comes from "The Second World War began in 1939, and the Normandy Landings were a turning point in the war," after text splitting, the possible target knowledge cells include "Second World War," "1939," "Normandy Landings," and "turning point in the war." Semantic relationships refer to the interconnections between these knowledge cells. For example, the relationship between "Second World War" and "1939" is "begins in," while the relationship between "Normandy Landings" and "Second World War" is "key event" or "battle." In this step, the system analyzes and labels the relationships between these cells, such as: The Second World War began in 1939, and the Normandy Landings were a key event in the Second World War.

[0042] Vectorization is the process of converting text or knowledge cells into vectors for subsequent computation and comparison. Typically, natural language processing (NLP) methods such as BERT, Word2Vec, or other embedding models are used to generate vector representations for each target knowledge cell. Vectorization converts text information into numerical form, enabling computers to process and analyze it. In this step, the first target knowledge cell (e.g., "Second World War") and the second target knowledge cell (e.g., "1939") are converted into vector representations, respectively. Cell similarity is measured by calculating the vector representations between target knowledge cells to measure their similarity. Common methods for calculating similarity include cosine similarity, Euclidean distance, etc. For example, after generating vector representations for "Second World War" and "1939" using the BERT model, the similarity between these two vectors can be calculated. If their similarity is high, it means that their relationship is close and they may have a strong semantic connection. Semantic networks are graphs composed of nodes (target knowledge cells) and edges (relationships between them). Through cell similarity and semantic relationships, the system connects knowledge cells into a network, where each node represents a knowledge cell and the edge represents the semantic relationship between knowledge cells. For example, "Second World War" and "Normandy Landings" are connected through the battle relationship, while "Normandy Landings" and "1939" are connected through the time point relationship. Semantic networks help the system understand how these knowledge points are related to each other, making it easier to generate questions and answers later.

[0043] S130, based on the semantic network, generating a first question using multiple agents.

[0044] Specifically, the server is the core part of the entire system, which is responsible for processing and calculating data. The "semantic network" here refers to a network of knowledge points with clear relationships built through the processing of target knowledge data. The semantic network can help the server understand the connections between different knowledge points and provide the necessary background and support for generating questions. Multi-agent refers to multiple agents working together to solve tasks. In the knowledge quiz game, multi-agent includes: expert agent: focuses on the professionalism and accuracy of the question, and can generate deep questions. Student agent: focuses on the easy understanding and popularity of the question, ensuring that the question is acceptable to the player. Verification agent: responsible for checking the logic, grammar and accuracy of the question, ensuring that the question has no ambiguity and conforms to the structure of the knowledge point. These agents are responsible for different tasks and work together to generate the final question. Through multi-agent collaboration, the system can generate questions that are both deep and easy to understand.

[0045] For example: Suppose we have a biology knowledge quiz game where players need to answer questions about photosynthesis. Suppose the server builds the following semantic network based on biological knowledge: knowledge cells (nodes): photosynthesis, chloroplast, carbon dioxide, water, chlorophyll, reaction formula. Semantic relationships (edges): photosynthesis occurs in chloroplast, photosynthesis requires water and carbon dioxide, chlorophyll helps capture light energy, photosynthesis reaction formula is XXX. Using this semantic network, the server generates the first question "What is the core step of the electron transport chain in photosynthesis" through multi-agent.

[0046] In one possible implementation, based on the semantic network, the first question is generated by using multi-agent, specifically including: extracting knowledge core according to the semantic network; controlling the expert agent to generate a first draft based on a large language model according to the knowledge core; controlling the student agent to generate a second draft based on a prompt word algorithm according to the knowledge core; performing semantic fusion on the first draft and the second draft to obtain a third draft; and generating the first question by adjusting the third draft through the verification agent, the multi-agent including the expert agent, the student agent and the verification agent.

[0047] Specifically, the semantic network helps the system understand the structure of knowledge by representing the relationships between different knowledge points. In this step, the system extracts the knowledge core from the semantic network, which is the most critical and central knowledge point. For example, if the target knowledge is "photosynthesis", the system may extract keywords such as "photosynthesis reaction formula", "chlorophyll", "light energy" from the semantic network. The task of the expert agent is to generate a more professional draft, ensuring that the question has sufficient academic depth. Based on the extracted knowledge core, the expert agent generates the first question draft through a large language model, which contains detailed and high-level knowledge content. For example: assuming the knowledge core is "photosynthesis reaction formula", the expert agent may generate a draft similar to the following: "Please describe the electron transport chain and its key steps in photosynthesis in detail."

[0048] The student agent focuses on generating questions that are more suitable for ordinary players or students and easy to understand. The student agent uses the prompt word algorithm based on the knowledge core to generate the second draft, which will be more concise and easy to understand, and help players understand the core knowledge. For example: based on the core "photosynthesis reaction formula", the student agent may generate a more simplified question such as: "What is the reaction formula of photosynthesis?"

[0049] Next, the server will perform semantic fusion on the drafts generated by the expert agent and the student agent, combining the advantages of both to produce a balanced question suitable for a wide range of players. Semantic fusion refers to integrating the content of the two drafts, removing redundant or unnecessary parts, and ensuring that the question has both depth and is understandable to most players. For example: after merging the first draft and the second draft, a question similar to the following may be obtained: "What is the reaction formula of photosynthesis? Please explain the key reaction steps."

[0050] The task of the verification agent is to perform logical and language verification on the third draft generated, ensuring the accuracy, clarity, and unambiguousness of the question, while checking whether it meets the design goals of the game (such as difficulty, learning effect, etc.). The verification agent may adjust the wording or content of the question to make it more accurate and standardized. For example: if the wording in the draft is not clear enough or the question expression has ambiguity, the verification agent may rewrite the question, for example, changing "Please explain the key reaction steps" to "Please briefly describe the key reaction steps in the reaction formula of photosynthesis". After the collaborative work of multiple agents, the first question generated will combine academic depth, ease of understanding, and accuracy, becoming a challenging question that meets the learning goals. For example: the final first question may be: "What is the reaction formula of photosynthesis? Please briefly describe the key reaction steps in the reaction formula."

[0051] For example: Suppose the goal of the knowledge quiz game is to test the player's understanding of photosynthesis. The following is the detailed generation process: Knowledge core: "Photosynthesis reaction equation", "The role of chlorophyll in photosynthesis", "The process of photosynthesis". The expert agent generates a draft question: "Please describe in detail the electron transport chain and its key steps in photosynthesis. How is light energy converted into chemical energy? Please give examples." The student agent generates a second draft: "What is the reaction equation for photosynthesis? How does photosynthesis use light energy?" After merging the expert and student drafts, the third draft is obtained: "What is the reaction equation for photosynthesis? Please briefly describe the key steps of photosynthesis." The verification agent may find that the "brief description" is not clear enough and may modify it to: "Please briefly describe the key reaction steps in the photosynthesis reaction equation." After multi-agent adjustment, the first question is finally generated: "What is the reaction equation for photosynthesis? Please briefly describe the key reaction steps in the reaction equation."

[0052] This passage describes the process of generating questions in a quiz game through multi-agent collaboration. The expert agent generates academic questions, the student agent generates easy-to-understand questions, and the verifier agent ensures the accuracy and logic of the questions. Through semantic fusion, the final questions meet the depth requirements and are suitable for the player's comprehension level. This multi-agent collaborative approach ensures intelligent, flexible, and personalized question generation, greatly enhancing the game's educational and entertainment value.

[0053] S140: Obtain a first answer from the user corresponding to the first question.

[0054] Specifically, during the quiz game, the server receives and obtains the user's response to the first question in real time. This is the user's answer to the system-generated first question, which can be in various forms, such as text, voice, or multiple-choice options. Once the server receives this first answer, it can further analyze, judge, and process it, such as determining whether the answer is correct, analyzing the key elements of the answer, or using it for intelligent generation and feedback evolution of subsequent questions.

[0055] For example, assume that in a knowledge quiz game, the first question generated by the server is: "What are the main processes of water cycle?" The user inputs the answer in the game as: "Water cycle mainly includes evaporation, condensation and precipitation processes." At this time, the server will automatically obtain and record the content of the user's input answer, that is, this paragraph is the "first answer" of the user for the first question. The subsequent system can be based on the first answer to perform answer matching, keyword extraction, answer accuracy calculation, and even be used to automatically generate supplementary questions or more challenging questions. The advantage of this process is that it can make the knowledge quiz game more intelligent and dynamic, not simply judging right or wrong, but deeply understanding the content of the user's answer and providing personalized feedback and next content design based on the real answer, greatly improving the interactivity and learning effect of the game.

[0056] S150, feedback evolution is performed on the first answer, and a second question is output.

[0057] Specifically, after obtaining the answer (first answer) of the user for the first question, the server does not simply judge right or wrong or directly enter the next question, but deeply analyzes and processes the answer of the user. The server will analyze and evolve the answer of the user according to the content of the user's answer, in combination with existing knowledge data and semantic networks, through certain intelligent algorithms (such as answer similarity calculation, keyword extraction, semantic analysis, etc.), to dynamically generate a more targeted and personalized second question. The second question may be to help the user further understand the knowledge, find the missing part, or increase the difficulty of answering, so as to realize a more in-depth and logical interactive question and answer experience.

[0058] It makes knowledge quiz no longer a mechanical rote learning, but has a "teaching students in accordance with their aptitude" type feedback ability. The server can discover knowledge blind spots or understanding biases according to the specific answer content of the user, and intelligently generate supplementary or expanded questions to help the user perfect the knowledge structure and deepen the understanding, while improving the interest and personalized experience of the game. The dynamic evolution of the question and answer mechanism can effectively enhance the learning effect and participation of the user.

[0059] In one possible implementation, the feedback evolution is performed on the first answer, and a second question is output, specifically comprising: obtaining a preset answer corresponding to the first question; determining a preset answer vector corresponding to the preset answer by using a BERT model; determining a first answer vector corresponding to the first answer by using the BERT model; calculating an answer similarity between the preset answer vector and the first answer vector; if it is determined that the answer similarity is greater than a preset threshold, obtaining a first keyword from the first answer; comparing the first keyword with a preset keyword library to generate a comparison result, the preset keyword library being constructed according to the preset answer; determining the second question according to the comparison result.

[0060] Specifically, the standard answer (preset answer) is obtained: the server internally prepares the correct or standard answer to the question in advance. BERT is a deep learning model for natural language understanding, and the server uses it to convert the preset answer into a vector (preset answer vector) and also converts the user's first answer into a vector (first answer vector). By comparing the similarity between the two vectors, the closeness of the user's answer and the standard answer is determined. If the answer similarity is greater than a certain set threshold (indicating that the answer is not bad, but it may not be comprehensive or detailed enough), the server extracts the core first keyword from the user's first answer. Compare the keywords in the user's answer with the preset keyword library (which is a set of knowledge points based on the standard answer), analyze which keywords have been mentioned and which keywords have not been mentioned, and form a comparison result. The server designs the second question according to the missing or weak parts of the user's answer to supplement the knowledge blind spot or expand related knowledge.

[0061] This "feedback evolution" mechanism breaks the limitations of traditional question and answer games that simply judge right or wrong or fixed question setting, and realizes deep understanding and personalized diagnosis of user answers through BERT semantic model and keyword comparison. This not only stimulates users' thinking and exploration, but also automatically identifies blind spots or weak points in users' knowledge structure, and generates follow-up questions that are more targeted, inspiring and teaching in real time. The entire knowledge question and answer process is more intelligent, flexible and adaptive to different users' learning paths, greatly improving the educational value and user experience of the game.

[0062] In one possible implementation, the second question is determined according to the comparison result, specifically including: if it is determined that the comparison result indicates that the number of same keywords between the first keyword and the preset keyword library does not meet the preset number, the second question is determined to be a complementary question based on the first question; if it is determined that the comparison result indicates that the number of same keywords between the first keyword and the preset keyword library meets the preset number, the second question is determined to be an expansion question based on the semantic network.

[0063] Specifically, after the server compares the keywords in the user's first answer, it will determine the type of the second question according to the comparison result. If the number of keywords in the user's answer and the number of keywords in the standard answer is quite different (indicating that the user's answer is incomplete or missing key information), the server will generate a "complementary question" to continue around the original question and guide the user to improve the answer. On the contrary, if the user's keywords are quite comprehensive (indicating that the user's grasp is good), the server will not be limited to the original question, but will generate an "expansion question" based on the semantic network to guide the user to explore new content related to the original knowledge, which is more in-depth or more extensive.

[0064] For example, suppose the first question is "Briefly explain why the earth has four seasons", and the user answers "Because the earth revolves around the sun". After keyword comparison, it is found that the important keyword "earth's axis tilt" is missing, which does not meet the preset number. The server will generate a supplementary question: "Do you know whether the earth's rotation axis affects the change of seasons?"; if the user answers "Because the earth revolves around the sun, and the earth's axis is tilted, which causes the sun's direct point to change", the number of keywords has been met, and the server will generate an extended question: "Do you know that in addition to the earth, other planets will also have similar four seasons?"

[0065] S160: Obtain the content of the quiz game through the first question, the first answer, and the second question.

[0066] Specifically, the server integrates and processes three key elements—the "first question," the "user's first answer," and the "second question generated based on the first answer"—to generate a complete and personalized quiz game. In other words, the server doesn't simply stack questions; instead, it dynamically adjusts question design, content presentation, and the answering process based on the user's actual responses, making the entire game more targeted, interactive, and engaging.

[0067] For example, let's assume the theme of a quiz game is "Astronomy." The server generates the first question: "Please briefly explain the composition of the solar system." The user's first answer is: "The Sun, Earth, and Mars." After analysis, the server discovers that the user only mentioned some of the planets, omitting others. It then generates a second question: "Besides the Sun, Earth, and Mars, what other planets or celestial bodies are in the solar system?" Finally, the server combines these two questions with the user's answer to generate the complete quiz game content. This approach allows quiz games to move beyond simply answering questions from a database and become more like a "thoughtful, responsive, and interactive" teaching assistant. It can automatically adjust game content based on user responses, truly tailoring teaching to individual students and enhancing the game's fun, educational value, and immersive experience.

[0068] In one possible implementation, a second answer corresponding to a second question from the user is obtained; a first accuracy rate corresponding to the first answer and a second accuracy rate corresponding to the second answer are determined; if it is determined that both the first accuracy rate and the second accuracy rate are higher than a preset accuracy rate, a third question is generated, and the difficulty corresponding to the third question is greater than the difficulty corresponding to each of the first question and the second question.

[0069] Specifically, the server will continue to obtain the user's answer to the "second question" (i.e., the second answer), and then evaluate the accuracy of the user's answers to the "first question" and the "second question" respectively. If the evaluation result shows that the user's answer accuracy on the two questions exceeds the system set standard (such as 70%, 80% of the preset accuracy), it means that the user's knowledge level is relatively high, and the answer performance is excellent. At this time, the server will automatically generate a new, more difficult third question to upgrade the challenge of the game and further test and improve the user's knowledge level.

[0070] In one possible implementation, with reference to Figure 2 , Figure 2 Another flowchart of a knowledge Q&A game content generation method provided by an embodiment of the present application is shown in FIG. 10. The method includes steps S210 to S250, which are as follows: S210, obtaining training information, the training information including training answers and training answer vectors; S220, inputting the training information into the adaptive word vector fusion network for training to obtain a first training result; S230, after superimposing and standardizing the first training result and the training information, obtaining a second training result; S240, inputting the second training result into the adaptive word vector fusion network for processing to obtain a third training result; and S250, superimposing and standardizing the third training result and the second training result until a training information similarity matrix is output, obtaining a BERT model, the training information similarity matrix satisfying a preset logistic regression condition.

[0071] Specifically, the training information is the core data for model training, including two parts: training answers: correct answers or preset standard answers. Training answer vectors: vector representation of training answers, usually converted into numerical vector form through some word embedding (such as Word2Vec or BERT, etc.) method, so that the computer can be processed. The adaptive word vector fusion network is a network model for optimizing word vector learning. Through training, the network will gradually adjust the weight and relationship of each word vector, so that the calculation of answers and similarity is more accurate. In this step, the training information is input into the network for the first training, and the first training result is obtained, that is, the word vector adjustment obtained after network processing. The first training result will be "superimposed and standardized" with the original training information. This means that after the first iteration of the network, the result is adjusted through some mathematical operations (such as weighted summation, normalization, etc.) to make it more stable and accurate. The second training result is obtained, that is, the optimized and adjusted word vector. The second training result is input into the adaptive word vector fusion network for the second round of processing to obtain the third training result. This process is to further fine-tune the accuracy of the model. Through repeated training and adjustment, a training information similarity matrix is finally output, which is used to measure the similarity between different training information. The similarity matrix is a matrix calculated by the model, where each element represents the similarity between two training information (usually calculated by cosine similarity algorithm, etc.). This matrix will eventually meet certain preset conditions, such as logistic regression conditions, meaning that the prediction and calculation accuracy of the model has reached a certain standard. Finally, through this adaptive training process, the similarity matrix and training result obtained will be used to optimize the BERT model, so that BERT can more accurately understand and process natural language in subsequent tasks.

[0072] The application also provides a knowledge Q&A game content generation device, referring to Figure 3 , Figure 3 A module schematic diagram of a knowledge Q&A game content generation device provided by an embodiment of the application. The device is a server, including an acquisition module 31 and a processing module 32, wherein the acquisition module 31 acquires target knowledge data corresponding to a knowledge Q&A game; the processing module 32 determines a semantic network according to the target knowledge data; the processing module 32 generates a first question based on the semantic network using multiple agents; the acquisition module 31 acquires a first answer corresponding to the first question from a user; the processing module 32 performs feedback evolution on the first answer and outputs a second question; and the processing module 32 obtains knowledge Q&A game content through the first question, the first answer, and the second question.

[0073] In a possible implementation, the processing module 32 determines the semantic network according to the target knowledge data, specifically including: the processing module 32 performs text splitting on the target knowledge data to obtain target knowledge cells; the processing module 32 determines the relationship between each target knowledge cell to obtain semantic relationships; the processing module 32 vectorizes the target knowledge cells to obtain a first target knowledge cell and a second target knowledge cell, the first target knowledge cell and the second target knowledge cell being any two target knowledge cells in the plurality of target knowledge cells; the processing module 32 calculates the cell similarity between the first target knowledge cell and the second target knowledge cell; and the processing module 32 constructs the semantic network according to the cell similarity and the semantic relationships.

[0074] In a possible implementation, the processing module 32 generates the first question by using multiple agents based on the semantic network, specifically including: the processing module 32 extracts a knowledge core according to the semantic network; the processing module 32 controls an expert agent to generate a first draft based on a large language model according to the knowledge core; the processing module 32 controls a student agent to generate a second draft based on a prompt word algorithm according to the knowledge core; the processing module 32 performs semantic fusion on the first draft and the second draft to obtain a third draft; and the processing module 32 generates the first question by adjusting the third draft through a verification agent, the multiple agents including the expert agent, the student agent, and the verification agent.

[0075] In a possible implementation, the processing module 32 performs feedback evolution on the first answer and outputs a second question, specifically including: the acquisition module 31 acquires a preset answer corresponding to the first question; the processing module 32 determines a preset answer vector corresponding to the preset answer by using a BERT model; the processing module 32 determines a first answer vector corresponding to the first answer by using the BERT model; the processing module 32 calculates an answer similarity between the preset answer vector and the first answer vector; the processing module 32 acquires a first keyword from the first answer if it is determined that the answer similarity is greater than a preset threshold; the processing module 32 compares the first keyword with a preset keyword library to generate a comparison result, the preset keyword library being constructed according to the preset answer; and the processing module 32 determines the second question according to the comparison result.

[0076] In a possible implementation, the processing module 32 determines the second question according to the comparison result, specifically including: the processing module 32 determines that the second question is a complementary question based on the first question if it is determined that the comparison result indicates that the number of same keywords between the first keyword and the preset keyword library does not meet a preset number; and the processing module 32 determines that the second question is an expansion question based on the semantic network if it is determined that the comparison result indicates that the number of same keywords between the first keyword and the preset keyword library meets the preset number.

[0077] In one possible implementation, the acquisition module 31 acquires the user's second answer corresponding to the second question; the processing module 32 determines a first accuracy rate corresponding to the first answer, and a second accuracy rate corresponding to the second answer; if the processing module 32 determines that the first accuracy rate and the second accuracy rate are both higher than the preset accuracy rate, a third question is generated, and the difficulty corresponding to the third question is greater than the difficulty corresponding to the first question and the second question.

[0078] In one possible implementation, the acquisition module 31 acquires training information, which includes training answers and training answer vectors; the processing module 32 inputs the training information into the adaptive word vector fusion network for training to obtain a first training result; the processing module 32 superimposes and standardizes the first training result and the training information to obtain a second training result; the processing module 32 inputs the second training result into the adaptive word vector fusion network for processing to obtain a third training result; the processing module 32 superimposes and standardizes the third training result and the second training result until a training information similarity matrix is ​​output to obtain a BERT model, and the training information similarity matrix meets the preset logistic regression conditions.

[0079] It should be noted that the above embodiments provide devices that implement their functions using only the division of the above functional modules as examples. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiments are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0080] This application also provides an electronic device, referring to Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include: at least one processor 41, at least one network interface 44, a user interface 43, a memory 45, and at least one communication bus 42.

[0081] The communication bus 42 is used to realize the connection and communication between these components.

[0082] The user interface 43 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 43 may also include a standard wired interface and a wireless interface.

[0083] The network interface 44 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0084] The processor 41 can include one or more processing cores. The processor 41 connects various parts within the server through various interfaces and lines, performs various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 45, and calling data stored in the memory 45. Alternatively, the processor 41 can be implemented in at least one of a hardware form of a digital signal processing (DSP), a field-programmable gate array (FPGA), and a programmable logic array (PLA). The processor 41 can be integrated with a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. Among them, the CPU mainly processes operating systems, user interfaces, and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor 41, but can be realized by a separate chip.

[0085] The memory 45 can include a random access memory (RAM) and a read-only memory (ROM). Alternatively, the memory 45 includes a non-transitory computer-readable storage medium. The memory 45 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 45 can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area can store data involved in the above-mentioned various method embodiments, etc. The memory 45 can also be at least one storage device located away from the aforementioned processor 41. As shown in the figure, the memory 45 as a computer storage medium can include an operating system, a network communication module, a user interface module, and an application program of a knowledge Q&A game content generation method. Figure 4

[0086] In Figure 4 ​In the electronic device shown, the user interface 43 is mainly used to provide an interface for the user to input, and obtain data input by the user; and the processor 41 can be used to call an application program stored in the memory 45 and storing a knowledge quiz game content generation method, which, when executed by one or more processors, causes the electronic device to perform the method of one or more of the above embodiments.

[0087] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all described as a series of action combinations, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other order or at the same time. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0088] The present application also provides a computer-readable storage medium, which stores instructions. When executed by one or more processors, the instructions cause the electronic device to perform the method of one or more of the above embodiments.

[0089] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0090] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of the units is only a logical function division. There can be another division manner in actual implementation, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some services interfaces, devices or units, and can be electrical or other forms.

[0091] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0092] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware, or in the form of a software functional unit.

[0093] If the integrated unit is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned memory includes: a U disk, a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0094] The above-described are only exemplary embodiments of the present disclosure, and cannot limit the scope of the present disclosure. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure are still within the scope of the present disclosure. Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon considering the specification and practicing the true principles of the present disclosure. The present application is intended to cover any variations, uses or adaptive changes of the present disclosure that follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not disclosed by the present disclosure. The specification and examples are only considered as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for generating content for a trivia game, the method comprising: The method comprises: acquiring target knowledge data corresponding to a knowledge question and answer game; determining a semantic network according to the target knowledge data; generating a first question based on the semantic network using multiple agents, specifically including: extracting a knowledge core according to the semantic network; controlling an expert agent to generate a first draft based on a large language model according to the knowledge core; controlling a student agent to generate a second draft based on a prompt word algorithm according to the knowledge core; performing semantic fusion on the first draft and the second draft to obtain a third draft; adjusting the third draft through a checking agent to generate the first question, wherein the multiple agents include the expert agent, the student agent, and the checking agent; acquiring a first answer corresponding to the first question from a user; feedback evolution on the first answer to output a second question, specifically including: acquiring a preset answer corresponding to the first question; determining a preset answer vector corresponding to the preset answer using a BERT model; determining a first answer vector corresponding to the first answer using the BERT model; calculating an answer similarity between the preset answer vector and the first answer vector; if it is determined that the answer similarity is greater than a preset threshold, obtaining a first keyword from the first answer; comparing the first keyword with a preset keyword library to generate a comparison result, wherein the preset keyword library is constructed according to the preset answer; if it is determined that the comparison result indicates that the number of same keywords between the first keyword and the preset keyword library does not meet a preset number, determining that the second question is a supplementary question based on the first question; if it is determined that the comparison result indicates that the number of same keywords between the first keyword and the preset keyword library meets the preset number, determining that the second question is an expansion question based on the semantic network; obtaining knowledge question and answer game content through the first question, the first answer, and the second question.

2. The knowledge Q&A game content generation method of claim 1, wherein, The determining of the semantic network according to the target knowledge data specifically comprises: text splitting of the target knowledge data to obtain target knowledge cells; determining relationships between the target knowledge cells to obtain semantic relationships; vectorizing the target knowledge cells to obtain a first target knowledge cell and a second target knowledge cell, wherein the first target knowledge cell and the second target knowledge cell are any two target knowledge cells in the multiple target knowledge cells; calculating a cell similarity between the first target knowledge cell and the second target knowledge cell; constructing the semantic network according to the cell similarity and the semantic relationship.

3. The knowledge Q&A game content generation method of claim 1, wherein, The method further comprises: acquiring a second answer corresponding to the second question from the user; determining a first accuracy corresponding to the first answer and a second accuracy corresponding to the second answer; if it is determined that the first accuracy and the second accuracy are both higher than a preset accuracy, generating a third question, wherein a difficulty corresponding to the third question is greater than difficulties corresponding to the first question and the second question respectively.

4. The knowledge Q&A game content generation method of claim 1, wherein, The method further comprises: Obtain training information, the training information including training answers and training answer vectors; Input the training information into an adaptive word vector fusion network for training to obtain a first training result; After superimposing and standardizing the first training result and the training information, a second training result is obtained; Input the second training result into the adaptive word vector fusion network for processing to obtain a third training result; Superimpose and standardize the third training result and the second training result until a training information similarity matrix is output, to obtain the BERT model, the training information similarity matrix satisfying a preset logistic regression condition.

5. A knowledge quiz game content generating apparatus characterized by comprising: The device for implementing the knowledge question and answer game content generation method according to claim 1 comprises an acquisition module (31) and a processing module (32), wherein The acquisition module (31) is configured to obtain target knowledge data corresponding to a knowledge question and answer game; and the processing module (32) is configured to determine a semantic network according to the target knowledge data. The processing module (32) is further configured to generate a first question based on the semantic network using multiple intelligent agents. The acquisition module (31) is further configured to obtain a first answer corresponding to the first question from a user; the processing module (32) is configured to perform feedback evolution on the first answer to output a second question; and the processing module (32) is configured to obtain knowledge question and answer game content through the first question, the first answer, and the second question.

6. An electronic device, comprising: The electronic device comprises a processor (41), a memory (45), a user interface (43), and a network interface (44), the memory (45) is configured to store instructions, the user interface (43) and the network interface (44) are configured to communicate with other devices, and the processor (41) is configured to execute the instructions stored in the memory (45) to enable the electronic device to perform the method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method according to any one of claims 1 to 4.

Images