An education system and method based on artificial intelligence model

Abstract

An education system and method based on an artificial intelligence model, including an information collection unit, an information processing unit, an information retrieval unit, and a result output unit; the information collection unit collects education information; the information processing unit processes the collected education information to obtain a corresponding query vector; the information retrieval unit performs a similarity search for resource vectors in a vector database according to the query vector to obtain a search result; the result output unit queries the corresponding resources in the resource database according to the search result and outputs the corresponding education resources. The present invention combines enhanced search with multimodal technology to innovatively construct a generative artificial intelligence teaching system, which not only provides students with an efficient and accurate personalized learning path, but also assists teachers in formulating lesson plans and accurately grasping students' learning situations, thereby showing great practicality and economic value in the field of education, especially engineering education, and has broad application prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of network teaching, and more specifically, to an education system and method based on an artificial intelligence model. Background Art

[0002] Well-known domestic universities and research institutions have achieved important research results in the fields of deep learning, natural language processing, image generation, etc. of generative AI.

[0003] Although traditional general generative artificial intelligence technology has penetrated into many aspects of daily life and demonstrated its great potential, it is also faced with several significant limitations. The main ones include: first, the answers are often too broad and difficult to accurately match the actual needs of users, affecting the user experience; second, the knowledge base is limited by timeliness, and it is often unable to cope with emerging issues or inquiries beyond the scope of established knowledge; third, when it comes to highly professional or complex knowledge points, the quality of its answers is often unsatisfactory; finally, due to the singleness of the text form, it is difficult to directly provide rich multimedia resources or deep links, which limits the comprehensiveness and interactivity of the information.

[0004] Therefore, the problems existing in the prior art need to be further improved and developed. Summary of the invention

[0005] (I) Purpose of the invention: In order to solve the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide an education system and method based on an artificial intelligence model.

[0006] (II) Technical solution: In order to solve the above technical problems, the present technical solution provides an education system based on an artificial intelligence model, including an information collection unit, an information processing unit, an information retrieval unit, and a result output unit; the information collection unit collects education information; The information processing unit processes the collected educational information to obtain a corresponding query vector; The information retrieval unit performs a similarity search for resource vectors in a vector database according to the query vector to obtain a retrieval result; The result output unit searches for corresponding resources in the resource database according to the search results, and outputs corresponding educational resources.

[0007] Preferably, the educational information includes teaching resources, educational conditions and teaching objectives; The teaching resources include all resources used by teachers in the teaching process or by students in the learning process; the educational situation refers to the degree to which students have mastered knowledge and the speed at which they understand knowledge; the teaching objectives refer to the specific goals that need to be achieved after the completion of this teaching.

[0008] Preferably, the information processing unit performs data processing on the educational information to obtain a corresponding educational information text block, and performs natural language processing on the educational information text block to obtain a corresponding query vector; the query vector is stored in a vector database, the educational information corresponding to the query vector is stored in a resource database, and the vector database and the resource database are mapped according to the educational information.

[0009] Preferably, the data processing includes data cleaning, format conversion, block processing, and data standardization; The data cleaning refers to the process of removing duplicate, missing or abnormal data records; the format conversion refers to converting the collected data into a unified format; the block processing refers to document segmentation of educational information in a unified format to obtain multiple text blocks of educational information; the data standardization refers to data standardization processing of multiple text blocks of segmented educational information to obtain text blocks of educational information in a standardized format.

[0010] Preferably, natural language processing is performed on the educational information text block, and the educational information text block is converted into a high-dimensional query vector according to semantics.

[0011] Preferably, the information retrieval unit obtains the corresponding query vector through information processing based on the newly collected educational information and performs similarity search in the vector database; performs a first screening by calculating the Euclidean distance between the high-dimensional query vector and the resource vector to obtain the first resource vector set, and then calculates the cosine similarity of the resource vectors in the first resource vector set with the query vector respectively, sorts the calculation results, and obtains the resource vector with the highest similarity.

[0012] Preferably, the Euclidean distance measures the similarity between two vectors by calculating the straight-line distance between the two vectors, and the Euclidean distance is a positive number greater than or equal to 0; , in, and Represents the i-th component of vectors A and B respectively; n represents the dimension of the vector, that is, the number of component elements contained in vectors A and B; By calculating the Euclidean distance between the query vector and the resource vector, all resource vectors whose distance is less than or equal to a first distance threshold are grouped into a first resource vector set.

[0013] Preferably, the cosine similarity measures the similarity between two vectors by calculating the inner product of the two vectors divided by the product of their moduli, and the value range of the cosine similarity is [-1, 1]. , Where (A·B) represents the inner product of vectors A and B; and Represent the magnitude of vectors A and B respectively; The cosine similarity calculation is performed based on the input query vector and the resource vectors in the first resource data set, and the calculation results are sorted in descending order according to the similarity order to obtain the resource vector with the highest similarity.

[0014] Preferably, the result output unit searches for relevant information corresponding to the resource vector in the corresponding resource database according to the search result and outputs the information to the result display module, thereby obtaining a corresponding result report and a specific education strategy.

[0015] An educational method based on an artificial intelligence model, applicable to an educational system based on an artificial intelligence model, specifically comprising: Step 1, the information collection unit collects educational information; Step 2: The information processing unit processes the collected educational information to obtain a corresponding query vector; Step 3: The information retrieval unit performs a similarity search for resource vectors in the vector database according to the query vector to obtain a retrieval result; Step 4: The result output unit searches for corresponding resources in the resource database according to the search results and outputs the corresponding educational resources.

[0016] (III) Beneficial effects: The educational system and method based on the artificial intelligence model of the present invention firstly combines enhanced search and multimodal technology to innovatively construct a generative artificial intelligence teaching system, which can not only provide students with efficient and accurate personalized learning paths, but also assist teachers in formulating teaching plans and accurately grasping students' learning situations, thereby showing great practicality and economic value in the field of education, especially engineering education, and has broad application prospects; secondly, the efficient knowledge slicing technology is used to perform refined analysis and reorganization of uploaded resources, effectively dealing with the possible misleading output, insufficient knowledge timeliness and lack of professional knowledge in specific fields of generative artificial intelligence. Finally, through intelligent processing, not only the applicability and accuracy of resources are improved, but also the timeliness and professionalism of teaching content are ensured, providing learners with richer and more accurate knowledge resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a structural schematic diagram of an education system based on an artificial intelligence model of the present invention; Figure 2 It is a process step diagram of an educational method based on an artificial intelligence model of the present invention. DETAILED DESCRIPTION

[0018] The present invention is further described in detail below in conjunction with preferred embodiments. More details are elaborated in the following description to facilitate a full understanding of the present invention. However, the present invention can obviously be implemented in a variety of other ways different from the description. Those skilled in the art can make similar generalizations and deductions based on actual application situations without violating the connotation of the present invention. Therefore, the protection scope of the present invention should not be limited by the content of this specific embodiment.

[0019] The accompanying drawings are schematic diagrams of embodiments of the present invention. It should be noted that the drawings are only examples and are not drawn to scale, and should not be used to limit the actual protection scope of the present invention.

[0020] An educational system based on artificial intelligence models, such as Figure 1 As shown, it includes an information collection unit, an information processing unit, an information retrieval unit, a result output unit and a control unit. The information collection unit collects educational information; the information processing unit processes the collected educational information to obtain a corresponding query vector; the information retrieval unit performs a similarity search for resource vectors in the vector database according to the query vector to obtain a search result; the result output unit searches for corresponding resources in the resource database according to the search result and outputs the corresponding educational resources; the control unit is connected to the information collection unit, the information processing unit, the information retrieval unit and the result output unit through wired and / or wireless connections to control the operation of each unit, thereby ensuring the normal operation of the entire system.

[0021] The information collection unit includes a display input module, through which the educational information is input and displayed. The display input module can be a liquid crystal display or a display with a touch screen function, etc., as long as it can ensure that the educational information can be input and displayed, and the specific structure is not limited.

[0022] The educational information includes teaching resources, educational learning conditions, and teaching objectives. The teaching resources include all resources that may be used by teachers during teaching or students during learning, such as WORD documents, PPT presentations, PDF files, pictures, and videos. The educational learning conditions refer to the degree of knowledge that students have mastered and the speed of understanding knowledge. The teaching objectives refer to the specific goals that need to be achieved after the completion of this teaching, including clear expected learning outcomes, knowledge and skills that students should master, etc.

[0023] The information processing unit performs data processing on the educational information to obtain a corresponding educational information text block, and performs natural language processing on the educational information text block to obtain a corresponding query vector; the query vector is stored in a vector database, and the educational information corresponding to the query vector is stored in a resource database, and the vector database and the resource database are mapped according to the educational information. More specifically, a resource database and a vector database are constructed to store the corresponding educational information and query vectors respectively, and the stored educational information and query vectors can be used as educational resource information and resource vectors for the next educational query. By repeating this process, the resource database and the vector database can be continuously enriched.

[0024] First, the collected educational information is read. The present invention adopts distributed network storage to ensure efficient access and scalability of the educational information. Automated tools are used to process the uploaded educational information to provide content documents in a standardized format for subsequent knowledge extraction and intelligent applications.

[0025] The automation tool refers to tools such as image-to-text and audio-to-text based on the big model. The education information is processed into text by the image-to-text tool and / or the audio-to-text tool, and the triple knowledge modeling tool is used in conjunction with the big model to translate the content of the relevant text and convert the terms to form a professional and formatted knowledge document. The education information is processed by the automation tool, which can accurately adapt to the dialogue query of various spoken languages ​​and standard terms, and realizes accurate query based on semantics, and does not need to rely on whether the dialogue contains keywords.

[0026] The triple knowledge modeling tools include pyhanlp, stanfordnlp, pyltp, OpenNRE, BERT, ChatGPT, etc. These modeling tools have their own advantages and disadvantages, and the choice should be based on specific needs and the collected educational information.

[0027] The distributed network storage is to store the educational information in multiple independent devices. The distributed network storage adopts an expandable network structure, uses multiple storage servers to share the storage load, and uses a location server to locate the storage information, which not only improves the reliability, availability and access efficiency of the system, but also makes it easy to expand.

[0028] The data processing includes operations such as data cleaning, format conversion, block processing, and data standardization. Data cleaning refers to the process of removing duplicate, missing, or abnormal data records, and is a key step in improving the quality of the educational information text block. In view of the large amount of noise information such as blank lines and irrelevant characters in the document, a large model is used to accurately match the document pattern, and then the natural language understanding and generation capabilities of the Qianfan large model are used to dynamically generate regular expressions, and each piece of data is refined. Regular expressions are used to process spaces for "(r'\s+',"")"; the replace method is used to filter irrelevant characters, replace symbols with "null", ignore short data, and uniformly process continuous blank characters and blank lines into a single blank line. The data is analyzed through the large model, the knowledge category to which the document belongs is archived, and a unique identification code UUID corresponding to each piece of data is generated. At the same time, the key symbols in the text are retained to maintain the integrity and readability of the semantics. This process significantly improves the purity of the educational information text block.

[0029] The regular expression is a string that uses a single string to describe and match a syntactic rule, and is usually used to retrieve and replace text that meets a certain rule. The regular expression is a powerful text processing tool that does not require specific expert experience and can adapt to documents of various types and writing styles.

[0030] The format conversion refers to converting the collected data into a unified format and converting the text data into an analyzable format. In the present invention, the data is converted into a custom JSON format {"doc":*, "zsdimg":*, "zsd":[], "UUID":*} and the data in JSON format is saved through the json.dump function.

[0031] More specifically, the present invention constructs a variety of file loaders to read the educational information in different formats, including TextLoader for plain text files, PyPDFLoader for PDF files, and UnstructuredMarkdownLoader for processing Markdown format documents and converting video content into text format.

[0032] The chunking process refers to document segmentation of educational information in a unified format to obtain multiple text blocks of educational information. Since the content of a single document cannot be determined and may exceed the context length limit of the artificial intelligence model processing, document segmentation technology is used to divide long documents into multiple smaller text blocks (chunks). Langchain provides a variety of flexible segmentation methods, such as Recursive Character Text Splitter (recursive segmentation by string), Character Text Splitter (segmentation by character), Markdown Header Text Splitter (segmentation based on Markdown title), etc. According to the specific type and requirements of the document, Recursive CharacterText Splitter is selected as the segmentation tool in the present invention, and the appropriate block size and overlap parameters are set to ensure that the segmentation results meet the requirements of the artificial intelligence model while retaining sufficient context information.

[0033] The block size refers to the maximum length of each text block obtained after segmenting the text, and the text blocks obtained after segmentation can be processed by the context. The overlap parameter refers to the maximum semantic coherence and structural integrity when segmenting the text. The block size and the overlap parameter are not specifically limited in the present invention and can be customized according to the actual context information, as long as it is guaranteed that the information processing unit can perfectly process them after segmentation.

[0034] The data standardization refers to performing data standardization processing on multiple text blocks of segmented educational information to obtain educational information text blocks in a standardized format. The data standardization refers to performing data standardization processing to ensure that data from different sources or types can be compared with each other.

[0035] After the data processing is completed, natural language processing is performed on the educational information text block, and the educational information text block is converted into a query vector that can be understood and / or processed by the customized generative artificial intelligence model according to the semantics. Specifically, the standardized educational information text block is converted into a corresponding high-dimensional vector, and the semantic information of the word, sentence or document is represented by the high-dimensional vector.

[0036] The natural language processing refers to capturing the semantic and syntactic relationships of the educational information text block in a high-dimensional vector space. Words used in similar contexts will have similar vectors, allowing mathematical operations to be performed on the words to understand their meanings. In the present invention, the FAISS vector library is used to extract features from the educational information text block and then converted into a high-dimensional query vector.

[0037] The FAISS vector library (Facebook AI Similarity Search) is an open source library developed by Facebook AI Research (FAIR) that is mainly used to efficiently perform similarity search and dense vector clustering. Similar vectors refer to words that are close in the vector space, for example, the word vectors of "king" and "queen" are closer than the word vectors of "king" and "apple".

[0038] Vectors are usually high-dimensional, with each dimension representing a certain feature of an object. In machine learning and deep learning, a set of vectors is often used to represent the features of a training set, with each dimension capturing different semantic features. These vectors accurately capture the semantic features of the text content, convert the educational information text block into a vector form and store it in a vector database as a resource vector. At the same time, ensure that the vector database can be saved locally to support fast access, and can also be uploaded to the cloud to enable sharing and remote retrieval of educational information text blocks. Converting the educational information text block into a query vector facilitates mathematical operations and similarity calculations, and the relationship between words can be calculated through simple vector addition and subtraction operations.

[0039] The method for performing natural language processing based on the educational information text block in the present invention can be the Embedding-V1 model of Wenxinyiyan, etc., as long as it can ensure that the educational information text block can be converted into a query vector that can be understood and processed by the artificial intelligence model, and the specific method is not limited.

[0040] The vector database is a database specially used for storing, indexing and retrieving high-dimensional vectors. The query vector representing the educational information text block is stored in the vector database. The vector database is linked to a resource database, and the teaching resources corresponding to the resource vectors in the vector database are stored in the resource database.

[0041] The information retrieval unit obtains a corresponding query vector through information processing according to the newly collected educational information, and performs similarity retrieval in the vector database, thereby obtaining a resource vector with the highest similarity to the query vector. By calculating the similarity between the query vector and the resource vectors contained in the vector database, the educational resource most similar to the educational information input by the user can be quickly found.

[0042] Since the document has been processed, mainly segmented, a document may correspond to multiple query vectors. When performing query vector retrieval, multiple similar resource vectors will be retrieved from the vector database, and the greater the vector similarity between the query vector and the resource vector, the more similar the educational resources linked to the resource vector are to the queried educational information.

[0043] In the present invention, the similarity between the query vector and the resource vectors contained in the vector database is measured by calculating the distance between the vectors, and the distance calculation methods between the vectors include cosine similarity, Euclidean distance, etc. The cosine similarity measures the angle between two vectors and is used for text similarity calculation; while the Euclidean distance is the straight-line distance between two vectors in space, which is used to quickly screen out resource vectors within a first distance threshold in the vector database: the first resource vector data set. First, the first resource vector set is obtained by calculating the Euclidean distance between the high-dimensional query vector and the resource vector for the first screening, and then the resource vectors in the first resource vector set are respectively subjected to cosine similarity calculation with the query vector, and the calculation results are sorted to obtain the resource vector with the highest similarity.

[0044] The Euclidean distance measures the similarity between two vectors by calculating the straight-line distance between the two vectors. As shown in formula (1), the Euclidean distance is a positive number greater than or equal to 0. The smaller the Euclidean distance is and the closer it is to 0, the shorter the straight-line distance between the query vector and the resource vector is, and the higher the similarity between the two vectors is; when the Euclidean distance is equal to 0, it means that the distance between the query vector and the resource vector is 0, and the two vectors coincide; conversely, the larger the Euclidean distance is, the larger the straight-line distance between the query vector and the resource vector is, and the lower the similarity between the two vectors is.

[0045] (1), in, and Represents the i-th component of vectors A and B respectively; n represents the dimension of the vector, that is, the number of component elements contained in vectors A and B.

[0046] When the value of the Euclidean distance is greater than or equal to the first distance threshold, it indicates that the two are similar, and the larger the value, the higher the similarity; conversely, when the similarity value is less than the first threshold, it indicates that the two are not similar, and the smaller the value, the lower the similarity. In the present invention, the first distance threshold is 0-1, and the preferred value is 0.8, that is, when the Euclidean distance between the query vector and the resource vector is less than or equal to 0.8, it indicates that the two are similar. All resource vectors that are less than or equal to the first distance threshold form a first resource vector set.

[0047] The cosine similarity measures the similarity between two vectors by calculating the inner product of the two vectors divided by the product of their moduli. As shown in formula (2), the value range of the cosine similarity is [-1,1]. When the value is 1, it means that the query vector and the resource vector are exactly the same, that is, the similarity between the two vectors is the largest; when the value is 0, it means that the query vector has nothing to do with the resource vector; when the value is -1, it means that the query vector and the resource vector are completely opposite, that is, the similarity between the two vectors is the smallest.

[0048] (2), Where (A·B) represents the inner product of vectors A and B; and Represent the magnitude of vectors A and B respectively.

[0049] The cosine similarity between the input query vector and the resource vectors in the first resource data set is calculated, and the calculation results are sorted in descending order according to the similarity order by the sorted function to obtain the resource vector with the highest similarity. Ensure that the most relevant resource vector is at the front, present the sorted results to the user in the form of a list or card, and provide detailed information corresponding to the resource vector and a preview function for the detailed information of the resource vector.

[0050] The result output unit searches for the relevant information corresponding to the resource vector in the corresponding resource database according to the search results and outputs it to the result display module, thereby obtaining the corresponding result report and specific educational strategy. Specifically, the teaching resources matching the resource vector are searched in the resource database according to the search results and output as learning resources.

[0051] The result report includes a variety of different grades, such as failed, passed, average, good, excellent, etc. The educational strategy is a learning plan, learning progress, and corresponding matching learning resources for different objects. The learning resources include video courses or corresponding test exercises corresponding to weak links, etc. The learning resources can be viewed online or downloaded by clicking, and there is no specific limitation in the present invention.

[0052] The result display module may also be a separate LCD display, or the same display as the information input module, or other display terminals. The present invention does not impose any specific restrictions, as long as it can display the final result.

[0053] An educational method based on an artificial intelligence model, applicable to an educational system based on an artificial intelligence model, such as Figure 2 As shown, the specific steps include: Step 1: The information collection unit collects educational information.

[0054] Step 2: The information processing unit processes the collected educational information to obtain a corresponding query vector.

[0055] Step 3: The information retrieval unit performs a similarity search for resource vectors in the vector database according to the query vector to obtain a retrieval result.

[0056] Step 4: The result output unit searches for corresponding resources in the resource database according to the search results and outputs the corresponding educational resources.

[0057] The following is a detailed description with reference to specific embodiments: Embodiment 1: Use an education system and method based on an artificial intelligence model of the present invention to detect students' learning situation, formulate personalized learning paths and plans, and output corresponding result reports and educational resources for weak links.

[0058] First, collect the basic information of all selected students and record all their activities on the learning platform, including video viewing time, lesson plan reading time, courseware browsing time, AI intelligent question and answer times and content, number of notes, etc.; count the types of resources issued (such as videos, lesson plans, courseware, etc.) and their total duration; record the participation in each teaching activity, submitted grades, etc.; collect the completion of post-class tests, including submitted knowledge, actual scores, etc. In this way, we can collect students' learning history, interests and ability levels, and evaluate students' learning progress and knowledge weaknesses and other educational information.

[0059] After that, the collected educational information is processed to obtain educational information text blocks, and duplicate, missing or abnormal data records are removed through data cleaning; data formatting converts the collected data into a unified format, such as converting time data into a unified unit (minutes or seconds), and converting text data into an analyzable format (such as word segmentation, removal of stop words). Data standardization standardizes the data to ensure that data from different sources or types can be compared with each other. For example, convert learning time, number of questions and answers, etc. into relative indicators or scores. Data association associates learning behavior data, teaching resource data, teaching activity data, and test data through key fields such as student ID or time to build a complete learning data chain.

[0060] After slicing and other processing of the educational information text block, natural language processing is performed to convert the educational information text block into a corresponding query vector. The query vector performs similarity retrieval in the vector database, and the resource vectors in the vector database are sorted and displayed in descending order according to the similarity. At the same time, the relevant educational resource information corresponding to the resource vector is queried in the resource database according to the retrieval results, and the corresponding conclusion report and educational resources are output.

[0061] According to students' learning needs and knowledge weaknesses, we set clear learning goals, develop personalized learning paths, and select relevant learning resources from the knowledge map or teaching resource library, including courses, videos, documents, exercises, etc. Based on students' interests and learning styles, we sort and recommend the selected learning resources in a personalized way to ensure that the recommended content not only meets students' learning needs but also stimulates their learning interest.

[0062] According to the students' answers, their scores are divided into different levels (fail, pass, medium, good, and excellent). Different teaching strategies are obtained for different levels: the corresponding learning resources, including videos, simulation operations, etc., are retrieved from the resource database. When the level is failed, all data resources of the current chapter are pushed, including all videos and simulation operations. When the level is passed, only the video resources of the current chapter are pushed. When the level is medium / good, two other chapters OTS (operation training system) resources and 5 video resources of the current and / or other chapters are randomly pushed. When the level is excellent, two to three other chapters OTS resources are pushed to challenge higher difficulty.

[0063] If not enough learning resources are found, three videos (not limited to the current chapter) will be randomly pushed for video resources, ensuring that at least one simulation operation resource is pushed. If this type of learning resources is insufficient, a random selection will be made.

[0064] Embodiment 2: Use an education system and method based on an artificial intelligence model of the present invention to detect the teaching situation of teachers, provide convenient lesson plan design templates and material libraries, provide matching templates and teaching materials according to teaching objectives, and support teachers to quickly create personalized teaching plans.

[0065] First, the lesson plan is uploaded to the information collection unit, which then starts the intelligent processing process, extracts the information data of the lesson plan, and performs fine segmentation on the uploaded document, with a default of 250 characters and 50 characters overlapping, thereby dividing the document into smaller slices, and performing a secondary inspection on the text slices through a recursive algorithm. After filtering out invalid text slices, the uploaded lesson plan information data is vectorized to ensure that the cut content can be vectorized. Structured output is performed according to the attributes such as the name of the lesson plan, number of class hours, course information, and teaching objectives, ensuring that the information is presented in a structured form. It is not only convenient for local storage management, but also realizes instant upload to the resource database, which promotes cloud sharing and efficient use of teaching resources.

[0066] The result of text slicing here may not meet expectations, so it is necessary to use recursion (call itself) to verify in order to achieve the purpose of data availability.

[0067] At the front end of lesson plan design, the information processing unit keenly captures every instruction input by the teacher, and accurately analyzes the content intent and specific needs through advanced natural language processing technology. This process is not limited to extracting key knowledge points and expected class time planning, but also to deeply understand the teacher's personalized teaching vision. The information processing unit also identifies and confirms whether it contains direct instructions to create a lesson plan. Once it is confirmed that it contains instructions to create a lesson plan, it immediately triggers the lesson plan design automation process.

[0068] While parsing the user input, the input teaching plan is processed, and then the information retrieval function is performed, including intelligent screening and matching of resource vectors in the vector database. Based on the extracted key knowledge points, the information retrieval unit compares the query vector obtained by vectorizing the instruction text with the resource vector stored in the vector database. By setting a scientific and reasonable teaching plan benchmark value, the information retrieval unit can quickly screen out teaching tasks and resources that meet or exceed the benchmark value in terms of teaching quality, applicability, etc., and finally intuitively display these preferred contents on the user interface, support the rapid distribution and update of course content, and provide teachers with rich and accurate teaching design support.

[0069] It can also monitor students' completion of learning tasks under the output teaching design in real time, provide visual analysis reports of learning data, and provide early warning and intervention mechanisms to help teachers understand students' learning status and needs and adjust teaching strategies in a timely manner.

[0070] The monitoring of students' learning task completion can be achieved through classroom interaction data collection and analysis, quantifying students' classroom interactions and recording their responses, generating classroom participation scores, quantitatively evaluating teachers' teaching effectiveness, and providing feedback. In the present invention, the evaluation of teachers' teaching effectiveness uses a large model to track students' learning process, and matches the learning content with the teacher's teaching knowledge points to grade students' mastery of knowledge points. The evaluation of the large model is combined with classroom-based professional teaching experience to quantify the evaluation results.

[0071] Specifically, conduct a comprehensive evaluation of the scores before, during, and after class: Pre-class analysis: students' actual preview time / preset time of the distributed resources (i.e., the total time of the video + teaching plan + courseware) * 60 + the number of questions and answers asked by students / 3 * 40 (i.e., asking more than 3 questions indicates a certain understanding, and the number of questions and answers is greater than 3, which is counted as 3, that is, the single score does not exceed 100%). In-class analysis: students' in-class scores = (scores of teaching activity 1 / total score of activity 1 + scores of teaching activity 2 / total score of activity 2 + ... + scores of teaching activity N / total score of activity N) * 100 / N; N ≥ 2. Post-class analysis: completion rate of after-class homework = number of people who submit after-class homework / total number of people; take the average of multiple tests.

[0072] The final grade classification: when the score is less than 60 points, it is a failure; when the score is greater than or equal to 60 points and less than 70 points, it is a pass; when the score is greater than or equal to 70 points and less than 80 points, it is a medium; when the score is greater than or equal to 80 points and less than 90 points, it is a good; when the score is greater than or equal to 90 points, it is an excellent. Therefore, for students whose personal after-class scores for a single knowledge point are less than 60 points, the corresponding knowledge point-related resources are recommended.

[0073] The present invention is an educational system and method based on an artificial intelligence model. First, it combines enhanced search and multimodal technology to innovatively construct a generative artificial intelligence teaching system, which can not only provide students with efficient and accurate personalized learning paths, but also assist teachers in formulating lesson plans and accurately grasping students' learning situations, thereby showing great practicality and economic value in the field of education, especially engineering education, and has broad application prospects. Secondly, efficient knowledge slicing technology is used to perform refined analysis and reorganization of uploaded resources, effectively dealing with the problems of misleading output, insufficient knowledge timeliness, and lack of professional knowledge in specific fields that may exist in generative artificial intelligence. Finally, through intelligent processing, not only the applicability and accuracy of resources are improved, but also the timeliness and professionalism of teaching content are ensured, providing learners with richer and more accurate knowledge resources.

[0074] The above content is an explanation of the preferred embodiments of the present invention, which can help those skilled in the art to more fully understand the technical solution of the present invention. However, these embodiments are merely illustrative, and it cannot be determined that the specific implementation methods of the present invention are limited to the description of these embodiments. For ordinary technicians in the technical field to which the present invention belongs, without departing from the concept of the present invention, several simple deductions and transformations can be made, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. An education system based on an artificial intelligence model, characterized in that: It includes an information collection unit, an information processing unit, an information retrieval unit, and a result output unit; the information collection unit collects educational information; The information processing unit processes the collected educational information to obtain a corresponding query vector; The information retrieval unit performs a similarity search for resource vectors in a vector database according to the query vector to obtain a retrieval result; The result output unit searches for corresponding resources in the resource database according to the search results, and outputs corresponding educational resources.

2. An education system based on an artificial intelligence model according to claim 1, characterized in that: The educational information includes teaching resources, educational conditions and teaching objectives; The teaching resources include all resources used by teachers in the teaching process or by students in the learning process; the educational situation refers to the degree to which students have mastered knowledge and the speed at which they understand knowledge; the teaching objectives refer to the specific goals that need to be achieved after the completion of this teaching.

3. The educational system based on artificial intelligence model according to claim 1, characterized in that: The information processing unit performs data processing on the educational information to obtain a corresponding educational information text block, and performs natural language processing on the educational information text block to obtain a corresponding query vector; the query vector is stored in a vector database, the educational information corresponding to the query vector is stored in a resource database, and the vector database and the resource database are mapped according to the educational information.

4. An education system based on an artificial intelligence model according to claim 3, characterized in that: The data processing includes data cleaning, format conversion, block processing, and data standardization; The data cleaning refers to the process of removing duplicate, missing or abnormal data records; the format conversion refers to converting the collected data into a unified format; the block processing refers to document segmentation of educational information in a unified format to obtain multiple text blocks of educational information; the data standardization refers to data standardization processing of multiple text blocks of segmented educational information to obtain text blocks of educational information in a standardized format.

5. The educational system based on artificial intelligence model according to claim 4, characterized in that: Natural language processing is performed on the educational information text block, and the educational information text block is converted into a high-dimensional query vector according to semantics.

6. The educational system based on artificial intelligence model according to claim 1, characterized in that: The information retrieval unit obtains a corresponding query vector through information processing based on the newly collected educational information, and performs similarity retrieval in the vector database; performs a first screening by calculating the Euclidean distance between the high-dimensional query vector and the resource vector to obtain a first resource vector set, and then calculates the cosine similarity between the resource vectors in the first resource vector set and the query vector respectively, sorts the calculation results, and obtains the resource vector with the highest similarity.

7. An education system based on an artificial intelligence model according to claim 6, characterized in that: The Euclidean distance measures the similarity between two vectors by calculating the straight-line distance between the two vectors. The Euclidean distance is a positive number greater than or equal to 0. ， in, and Represents the i-th component of vectors A and B respectively; n represents the dimension of the vector, that is, the number of component elements contained in vectors A and B; By calculating the Euclidean distance between the query vector and the resource vector, all resource vectors whose distance is less than or equal to a first distance threshold are grouped into a first resource vector set.

8. The educational system based on artificial intelligence model according to claim 6, characterized in that: Cosine similarity measures the similarity between two vectors by dividing the inner product of the two vectors by the product of their moduli. The value range of the cosine similarity is [-1, 1]. ， Where (A·B) represents the inner product of vectors A and B; and Represent the magnitude of vectors A and B respectively; The cosine similarity calculation is performed based on the input query vector and the resource vectors in the first resource data set, and the calculation results are sorted in descending order according to the similarity order to obtain the resource vector with the highest similarity.

9. The educational system based on artificial intelligence model according to claim 1, characterized in that: The result output unit searches for the relevant information corresponding to the resource vector in the corresponding resource database according to the search result and outputs it to the result display module, thereby obtaining the corresponding result report and specific education strategy.

10. An educational method based on an artificial intelligence model, applicable to an educational system based on an artificial intelligence model, characterized in that: Specifically include: Step 1, the information collection unit collects educational information; Step 2: The information processing unit processes the collected educational information to obtain a corresponding query vector; Step 3: The information retrieval unit performs a similarity search for resource vectors in the vector database according to the query vector to obtain a retrieval result; Step 4: The result output unit searches for corresponding resources in the resource database according to the search results and outputs the corresponding educational resources.

Images