Course method based on knowledge graph

Abstract

A course method based on a knowledge graph comprises the following steps: step 1, constructing a course system; 2, teaching content optimization; 3, teaching method innovation; 4, personalized learning path recommendation; and 5, evaluating and feeding back the teaching effect. By constructing the knowledge graph, effective integration of a course knowledge system is realized, and the knowledge graph is applied to course teaching reform of'three-dimensional modeling and additive manufacturing of metal parts', so that a scientific, systematic and efficient talent training system is constructed, excellent talents with innovative thinking and practical ability are trained, and the training efficiency is improved. The method has important teaching practical significance and application value.

Description

Technical Field

[0001] This invention relates to the field of educational technology, specifically to a knowledge graph-based curriculum method that optimizes the curriculum system, integrates teaching content, innovates teaching methods, and provides personalized recommendations for learning paths and feedback on teaching effectiveness by constructing a knowledge graph. Background Technology

[0002] Knowledge graphs, as the foundation and core of the transformation of artificial intelligence from perceptual intelligence to cognitive intelligence, can intuitively display the core knowledge structure, development history, cutting-edge dynamics, and overall system architecture of a specific field using visual graphs.

[0003] Additive manufacturing technology, as a cutting-edge and pioneering intelligent manufacturing technology, is driving profound changes in traditional production methods and processes. It is considered a key driver of the new industrial revolution and has attracted widespread attention from countries around the world. "3D Modeling and Additive Manufacturing of Metal Parts" is a foundational degree course for postgraduate students in mechanical engineering and intelligent manufacturing.

[0004] The existing teaching of the course "3D Modeling and Additive Manufacturing of Metal Parts" suffers from several problems, including a fragmented knowledge system, insufficient cutting-edge content, poor interactivity and effectiveness of teaching methods, difficulty in meeting the personalized learning needs of students from different professional backgrounds, and untimely feedback on teaching effectiveness. Traditional teaching models fail to effectively integrate knowledge from related courses such as mechanical design, theoretical mechanics, and fundamental engineering materials, and also struggle to incorporate ideological and political education and cutting-edge academic research into teaching, thus hindering the cultivation of interdisciplinary additive manufacturing talents with cross-level expertise, versatility, and strong integration. Summary of the Invention

[0005] This invention aims to provide a knowledge graph-based curriculum method. By constructing a knowledge graph, integrating the curriculum knowledge system, optimizing teaching content, and innovating teaching methods, it enables personalized learning path recommendations and precise teaching effect evaluation and feedback, thereby improving teaching quality and cultivating additive manufacturing interdisciplinary talents with innovative abilities and comprehensive qualities.

[0006] The technical solution adopted by this invention to solve its technical problem is:

[0007] A knowledge graph-based course methodology includes the following steps:

[0008] Step 1: Curriculum Construction: A knowledge graph for the course "3D Modeling and Additive Manufacturing of Metal Parts" is constructed. Python and DrissionPage are used to obtain course chapter data from the course slides. Then, pandas and py2neo are used to import the chapter data and construct a neo4j graph. The large model OpenAIGPT-4O is called to expand the graph data based on rule semantics and predefined entity relationships. Finally, Python and py2neo are used to read the JSON file, forming a complete knowledge graph.

[0009] Step 2: Optimize teaching content. Integrate the constructed knowledge graph with classroom teaching, starting from the knowledge domain, linking knowledge units, deconstructing knowledge points at all levels, and presenting visualized knowledge points including knowledge point profiles, knowledge point introductions, and knowledge point structures, thereby refining the "granularity" of knowledge.

[0010] Step 3: Innovative Teaching Methods: Complete the classroom teaching practice linked to the knowledge graph. Based on different logical structural relationships such as symbiosis and progression, connect various knowledge points of the major to construct a complete knowledge graph encompassing dimensions such as training objectives, competency graphs, problem graphs, knowledge graphs, and academic graphs, forming a structured knowledge system. Organize accumulated teaching materials such as videos, application cases, cutting-edge research, and student hot topics, and grid-link knowledge points from different fields and the teaching resources involved in each knowledge point to each knowledge point, forming a three-dimensional professional knowledge network.

[0011] Step 4: Personalized Learning Path Recommendation. Based on the differences in graduate students' professional backgrounds, learner profiles are constructed using knowledge graphs, and differentiated learning paths are recommended in conjunction with prior knowledge assessment. Reinforcement learning is used to dynamically optimize the recommendation strategy and adapt learning content to meet the personalized learning needs of different students.

[0012] Step 5: Teaching Effectiveness Evaluation and Feedback: By analyzing learning behavior data, such as the frequency of knowledge point access and test error patterns, the knowledge graph structure is optimized, the impact of the knowledge graph on learning effectiveness is quantified, and timely and effective teaching effectiveness evaluation and feedback are achieved to continuously improve teaching.

[0013] The present invention also has the following additional technical features:

[0014] As a further specific optimization of the technical solution of this invention: based on step 1, a connection is established between courses such as mechanical design, theoretical mechanics, and engineering materials fundamentals and this course, while introducing ideological and political education and cutting-edge academic research into the course, and constructing a cross-level, diversified, and highly integrated additive manufacturing compound talent training system.

[0015] As a further specific optimization of the technical solution of this invention: Based on step 1, the course chapter data in the course PPT is obtained using Python and DrissionPage. The method for this step is as follows: Step 1: Install DrissionPage and the python-pptx library for processing PPT; Step 2: Write a Python script to download the PPT through DrissionPage and then parse the chapter data therein; Step 3: Adjust the code according to the actual situation.

[0016] As a further specific optimization of the technical solution of the present invention: Based on step 1, the large model openaigpt-4o is called to expand the graph data according to the rule semantics and predefined entity relationships. The core principle of this step is to combine natural language understanding ability with structured rule constraints, and generate entity relationship data that conforms to the graph architecture through the model's capture of semantic patterns and logical reasoning.

[0017] As a further specific optimization of the technical solution of this invention: Based on step 1, the JSON file is finally read using Python + py2neo to form a complete knowledge graph. This step is based on a three-layer linkage of "data structuring transformation + graph model mapping + database operation", wherein: the data layer is Python parsing JSON text and extracting semantic units of entities, relations and attributes; the model layer is py2neo converting semantic units into graph model objects, following the rules of graph theory structure and transaction consistency; the storage layer interacts with Neo4j through the Bolt protocol, using constraints, indexes and other mechanisms to ensure the semantic integrity and query efficiency of the graph.

[0018] As a further specific optimization of the technical solution of the present invention: based on step 2, combined with relevant cutting-edge academic research, students are encouraged to establish a knowledge system for courses and majors, enhance the breadth and depth of course knowledge, and establish an innovative training model that integrates multiple fields and crosses multiple disciplines, and connects undergraduate and master's degrees.

[0019] As a further specific optimization of the technical solution of the present invention: on the basis of step 3, an additive manufacturing experimental comprehensive teaching platform for first-year professional master's students is established, forming a deep learning model based on knowledge graphs, realizing innovative teaching methods such as case teaching and problem-oriented learning, and constructing a highly interactive and effective teaching method and talent training paradigm.

[0020] Compared with the prior art, the advantages of this invention are:

[0021] This invention achieves effective integration of course knowledge systems by constructing a knowledge graph. Applying this knowledge graph to the teaching reform of the course "3D Modeling and Additive Manufacturing of Metal Parts," it integrates course teaching resources and content, utilizing dataset construction, knowledge extraction, and visualization technologies to construct a course knowledge graph based on the cross-disciplinary and multi-domain knowledge integration under the "New Engineering" perspective. This promotes graduate students' deeper understanding and practical application of mechanical engineering and intelligent manufacturing knowledge, significantly expanding their knowledge horizons, improving their self-learning ability, enhancing teaching effectiveness, and raising the quality of talent cultivation. Ultimately, it constructs a scientific, systematic, and efficient talent cultivation system, nurturing outstanding talents with innovative thinking and practical abilities, and possesses significant teaching practice significance and application value. Detailed Implementation

[0022] A knowledge graph-based course methodology includes the following steps:

[0023] Step 1: Curriculum System Construction: A knowledge graph for the course "3D Modeling and Additive Manufacturing of Metal Parts" is constructed. Python and DrissionPage are used to obtain course chapter data from the course slides. Then, pandas and py2neo are used to import the chapter data and construct a neo4j graph. The large model OpenAIGPT-4O is called to expand the graph data based on rule semantics and predefined entity relationships. Finally, Python + py2neo reads the JSON file to form a complete knowledge graph. Based on this, connections are established between this course and courses such as Mechanical Design, Theoretical Mechanics, and Fundamentals of Engineering Materials. Simultaneously, ideological and political education and cutting-edge academic research are introduced to construct a cross-level, diversified, and highly integrated additive manufacturing talent training system.

[0024] Building upon step 1, the course chapter data from the course PowerPoint presentation is retrieved using Python and DrissionPage. The method for this step is as follows:

[0025] Step 1: Install DrissionPage and the python-pptx library for processing PPT presentations;

[0026] Step 2: Write a Python script to download the PPT using DrissionPage and then parse the chapter data within it;

[0027] from DrissionPage import ChromiumPagefrom pptx importPresentationimport osimport refrom urllib.parse import urlparse

[0028] def download_ppt(url,save_path='. / downloads'):

[0029] "Download PPT files using DrissionPage"

[0030] #Create a save directory

[0031] os.makedirs(save_path,exist_ok=True)

[0032] #Initialize browser page

[0033] page = ChromiumPage()

[0034] page.get(url)

[0035] #Get PPT filenames (extract from URL or page)

[0036] parsed_url = urlparse(url)

[0037] filename=os.path.basename(parsed_url.path)

[0038] # If the filename does not contain the extension, try to extract it from the page. if not filename.endswith(('.ppt','.pptx')):

[0039] #The selector needs to be adjusted according to the actual page structure.

[0040] download_link=page.ele('a:has-text("Download"):has-text("PPT")')filename=download_link.attr('download')or'course_ppt.pptx'

[0041] download_link.click()

[0042] except:

[0043] #If you cannot locate the download link, try using the filename from the URL: filename='course_ppt.pptx'

[0044] #Wait for download to complete (the specific implementation may need to be adjusted according to browser behavior) page.wait.load_start()

[0045] page.wait.load_complete()

[0046] # Move the downloaded file to the specified path

[0047] # Note: In actual use, you may need to handle the browser's download path and file name download_path = os.path.join(save_path, filename)

[0048] print(f"PPT has been downloaded to: {download_path}")

[0049] return download_path

[0050] def extract_chapters_from_ppt(ppt_path):

[0051] "Extract chapter data from the PPT file"

[0052] prs = Presentation(ppt_path)

[0053] chapters = []

[0054] # Define the regular expression pattern for chapter titles (adjust according to the PPT format) chapter_pattern = re.compile(r'^第[一二三四五六七八九十0-9]+章\s*(.*)') for slide in prs.slides:

[0055] for shape in slide.shapes:

[0056] if not shape.has_text_frame:

[0057] continue

[0058] # Get all paragraphs in the text frame

[0059] for paragraph in shape.text_frame.paragraphs:

[0060] text = paragraph.text.strip()

[0061] if not text:

[0062] continue

[0063] #Check if it is a chapter title

[0064] match=chapter_pattern.match(text)

[0065] if match:

[0066] chapter_title=match.group(1)or text

[0067] chapters.append(chapter_title)

[0068] print(f"Chapter found: {chapter_title}")

[0069] return chapters

[0070] if __name__ == "__main__":

[0071] #Replace with the actual course PPT URL

[0072] course_ppt_url="https: / / example.com / course.pptx"

[0073] #Download PPT

[0074] ppt_file=download_ppt(course_ppt_url)

[0075] #Extract chapter data

[0076] chapters=extract_chapters_from_ppt(ppt_file)

[0077] # Output results

[0078] print("\nExtracted course chapters:")

[0079] for i,chapter in enumerate(chapters,1):

[0080] print(f"{i}.{chapter}")

[0081] Step 3: Adjust the code according to the actual situation;

[0082] PPT download logic: If the PPT requires login to access, login logic needs to be added to the download_ppt function.

[0083] Chapter extraction rules: The regular expression `chapter_pattern` needs to be adjusted according to the actual format of the chapter titles in the PPT. If the chapter titles have other characteristics (such as specific fonts, colors, or positions), corresponding judgment logic can be added in the code.

[0084] Error handling: More error handling code should be added in actual use, such as for network errors, missing files, etc.

[0085] Building upon step 1, we further import the chapter file data using pandas and py2neo and construct it into a neo4j graph. The code for this step is:

[0086] import pandas as pd

[0087] from py2neo import Graph,Node,Relationship

[0088] import argparse

[0089] import os

[0090] import logging

[0091] from typing import List,Dict,Any,Tuple

[0092] #Configuration Log

[0093] logging.basicConfig(

[0094] level = logging.INFO,

[0095] format='%(asctime)s-%(name)s-%(levelname)s-%(message)s' )

[0097] logger=logging.getLogger(__name__)

[0098] class Neo4jImporter:

[0099] """Utility class for importing chapter data into Neo4j""

[0100] def__init__(self,uri:str,user:str,password:str):

[0101] Initialize Neo4j connection.

[0102] try:

[0103] self.graph=Graph(uri,auth=(user,password))

[0104] logger.info("Successfully connected to the Neo4j database")

[0105] except Exception as e:

[0106] logger.error(f"Failed to connect to Neo4j:{e}")

[0107] raise

[0108] def import_chapters_from_csv(self,csv_path:str,clear_db:bool=False)->None:

[0109] Importing chapter data from a CSV file.

[0110] if clear_db:

[0111] self._clear_database()

[0112] #Read CSV file

[0113] try:

[0114] df = pd.read_csv(csv_path)

[0115] logger.info(f"Successfully read CSV file, containing {len(df)} records")

[0116] except Exception as e:

[0117] logger.error(f"Failed to read CSV file: {e}")

[0118] return

[0119] #Process the data and import it into Neo4j

[0120] self._process_and_import_data(df)

[0121] def_clear_database(self)->None:

[0122] "Clear all nodes and relationships in the Neo4j database."

[0123] logger.warning("Clearing all data in the Neo4j database...")

[0124] self.graph.delete_all()

[0125] logger.info("Database has been cleared")

[0126] def_process_and_import_data(self,df:pd.DataFrame)->None:

[0127] Process the data and import it into Neo4j.

[0128] #Assume the CSV file contains the following: id, title, parent_id, content

[0129] # Create a node cache to avoid creating the same node repeatedly.

[0130] node_cache = {}

[0131] #First create all chapter nodes

[0132] for _, row in df.iterrows():

[0133] chapter_id = row['id']

[0134] title = row['title']

[0135] content=row.get('content',")

[0136] #Creating chapter nodes

[0137] chapter_node=Node("Chapter",id=chapter_id,title=title,content=content)

[0138] self.graph.create(chapter_node)

[0139] node_cache[chapter_id]=chapter_node

[0140] logger.debug(f"Creating a chapter node:{title}")

[0141] #Then create the relationships between chapters

[0142] for _, row in df.iterrows():

[0143] chapter_id = row['id']

[0144] parent_id=row.get('parent_id')

[0145] if pd.notna(parent_id)and parent_id in node_cache:

[0146] #Get the current chapter and parent chapter nodes

[0147] chapter_node=node_cache[chapter_id]

[0148] parent_node=node_cache[parent_id]

[0149] #Create Relationship

[0150] rel=Relationship(chapter_node,"SUBCHAPTER_OF",parent_node)

[0151] self.graph.create(rel)

[0152] logger.debug(f"Create relationship: {chapter_node['title']}->{parent_node['title']}")

[0153] logger.info("Data import complete")

[0154] def main():

[0155] ""Main Function""

[0156] parser = argparse.ArgumentParser(description = 'Import chapter data from CSV to Neo4j')

[0157] parser.add_argument('--csv_path', required=True, help='CSV file path')

[0158] parser.add_argument('--uri',default='bolt: / / localhost:7687',help='Neo4j connection URI')

[0159] parser.add_argument('--user',default='neo4j',help='Neo4jusername')

[0160] parser.add_argument('--password',required=True,help='Neo4j password')

[0161] parser.add_argument('--clear', action='store_true', help='Whether to clear existing data')

[0162] args = parser.parse_args()

[0163] #Check if the file exists

[0164] if not os.path.exists(args.csv_path):

[0165] logger.error(f"File does not exist: {args.csv_path}")

[0166] return

[0167] try:

[0168] # Create an importer and execute the import

[0169] importer=Neo4jImporter(args.uri,args.user,args.password)

[0170] importer.import_chapters_from_csv(args.csv_path,args.clear)

[0171] logger.info("Import operation completed")

[0172] except Exception as e:

[0173] logger.error(f"An error occurred during import: {e}")

[0174] if __name__ == "__main__":

[0175] main()

[0176] Building upon step 1, the large model OpenAigpt-4O is invoked to expand the graph data based on rule semantics and predefined entity relationships. The core principle of this step is to combine natural language understanding capabilities with structured rule constraints, and generate entity relationship data that conforms to the graph architecture through the model's capture of semantic patterns and logical reasoning.

[0177] It mainly includes the following four layers of logic:

[0178] Logic 1: The Structured Mapping Principle of Rule Semantics

[0179] Semantic encoding of rules: Predefined rules need to be transformed into a semantic cue framework that the model can understand. Through cue engineering, rules are abstracted into "condition-action" logic (e.g., "If entity A is of type X, and entity B is of type Y, then the relationship between A and B is Z"). Essentially, this allows the model to learn the mapping between "rule patterns" and "relationship generation." The structured representation of rules guides the model to follow specific logic during generation, avoiding free association.

[0180] Semantic anchoring of the entity relationship system: Predefined entity types and relationship types constitute the semantic skeleton of the graph. The model already possesses a semantic understanding of these concepts through pre-training, and rules further strengthen the determinism of this association.

[0181] Logic II. Semantic Reasoning and Generation Mechanisms of Large Models

[0182] Logical derivation of contextual association: GPT-4o's Transformer architecture possesses multi-layer semantic representation capabilities. The input graph fragment serves as context, and the model extracts entity types and attribute keywords, matching them with conditions in the rules. This matching essentially calculates the semantic similarity between the input information and the rule conditions through an attention mechanism; when the matching degree reaches a threshold, the corresponding relationship is generated.

[0183] Generative control under constraints: By adjusting the rule descriptions and output format requirements in the prompts, the model's generation process is confined within a predefined logical framework. If the prompts explicitly state that "the relation type must be selected from {*, *, *}", the model will choose the relation word that best fits the rule based on probability calculations, rather than freely generating other words. This is similar to a "constrained fill-in-the-blank task," where the model fills in the rule-compliant relation structure by predicting the token sequence.

[0184] Logic 3: Principles of Semantic Consistency Verification of Entity Relationships

[0185] Reverse validation by the rule engine: The generated graph data needs to be validated a second time by the rule engine. The principle is to substitute the relationships output by the model back into the predefined rules and check whether the conditions are met. If not, the data is determined to be invalid.

[0186] Semantic matching for entity disambiguation: Utilizing entity linking techniques (such as matching knowledge bases by name and attributes) to ensure semantic consistency between generated entities and existing entities in the graph.

[0187] Logic 4: Structured Semantic Integration of Graph Fusion

[0188] Schema consistency maintenance: The schema of a knowledge graph defines the legal groups of entity types and relationships. Newly generated data must conform to the schema constraints of the graph. This is achieved by verifying the compatibility of relationship types and entity types through structured queries (such as SPARQL), ensuring that new data does not disrupt the semantic structure of the graph.

[0189] Enhanced Semantic Network Relationships: The relationships generated by the model are essentially new edges added to the semantic network of the existing graph, expanding the knowledge density of the graph through the relationships between entities.

[0190] Building upon step 1, the final step involves using Python and py2neo to read the JSON file and form a complete knowledge graph. This step is based on a three-layer linkage of "data structuring transformation + graph model mapping + database operations." Specifically: the data layer uses Python to parse the JSON text and extract semantic units of entities, relationships, and attributes; the model layer uses py2neo to transform these semantic units into graph model objects, adhering to graph theory structure and transaction consistency rules; and the storage layer interacts with Neo4j through the Bolt protocol, utilizing constraints, indexes, and other mechanisms to ensure the semantic integrity and query efficiency of the knowledge graph.

[0191] Step 2: Optimize teaching content. Integrate the constructed knowledge graph with classroom teaching, starting from the knowledge domain, linking knowledge units, deconstructing knowledge points at all levels, and presenting visualized knowledge points including knowledge point profiles, knowledge point introductions, and knowledge point structures, thereby refining the "granularity" of knowledge.

[0192] Its core logic lies in: using graph models to define the boundaries (nodes) and connections (edges) of knowledge particles, making abstract knowledge quantifiable and operable; using educational attributes to label the cognitive value (profiles) of particles, making knowledge deconstruction conform to teaching principles; and using visualization and interactive technologies to lower the threshold for understanding granularity, transforming complex knowledge systems into explorable networks that conform to human cognitive habits.

[0193] For example, by combining relevant cutting-edge academic research, we can help students build a knowledge system for their courses and majors, enhance the breadth and depth of their course knowledge, and establish an innovative undergraduate-master's integrated training model that integrates multiple fields and crosses multiple disciplines.

[0194] Step 3: Innovative Teaching Methods: Complete the classroom teaching practice linked to the knowledge graph. Based on different logical structural relationships such as symbiosis and progression, connect various knowledge points of the major to construct a complete knowledge graph encompassing dimensions such as training objectives, competency graphs, problem graphs, knowledge graphs, and academic graphs, forming a structured knowledge system. Organize accumulated teaching materials such as videos, application cases, cutting-edge research, and student hot topics, and grid-link knowledge points from different fields and the teaching resources involved in each knowledge point to each knowledge point, forming a three-dimensional professional knowledge network.

[0195] Step 3, Teaching Method Innovation, constructs a six-in-one knowledge ecosystem encompassing "goals-ability-knowledge-problems-resources-academics" through multi-dimensional graph semantic modeling, formal mapping of educational goals and knowledge structures, and networked association of resources and knowledge. Its core logic lies in: using graph models to express the symbiotic and progressive relationships between educational elements, transforming abstract teaching goals into an operational knowledge network; using semantic associations to break down the boundaries between knowledge, resources, and problems, forming a three-dimensional cognitive support system; and using a dynamic iteration mechanism to ensure the knowledge system evolves synchronously with academic frontiers, achieving a unity of breadth and depth of knowledge in integrated undergraduate and master's education. This synergy leverages the technological advantages of knowledge graphs in knowledge organization and reasoning while aligning with the student-centered and competency-oriented teaching philosophy in the education field.

[0196] For example, an additive manufacturing experimental comprehensive teaching platform will be established for first-year professional master's students, forming a deep learning model based on knowledge graphs, realizing innovative teaching methods such as case teaching and problem-oriented learning, and constructing a highly interactive and effective teaching method and talent training paradigm.

[0197] Step 4: Personalized Learning Path Recommendation. Based on the differences in graduate students' professional backgrounds, learner profiles are constructed using knowledge graphs, and differentiated learning paths are recommended in conjunction with prior knowledge assessment. Reinforcement learning is used to dynamically optimize the recommendation strategy and adapt learning content to meet the personalized learning needs of different students.

[0198] Its core logic lies in: using graph models to uniformly represent educational elements, making professional background, knowledge structure, and learning behavior computable graph nodes and edges; using semantic reasoning to bridge the gap between goals and the current situation, and using topological analysis of the knowledge graph to locate learning gaps; and using reinforcement learning to build a closed-loop feedback system, enabling recommendation strategies to evolve from "preset rules" to "data-driven adaptive decision-making." This integration leverages the advantages of knowledge graphs in knowledge organization and reasoning, while also endowing the system with dynamic evolutionary capabilities through reinforcement learning, ultimately achieving a leap from "standardized teaching" to "personalized knowledge navigation," which is particularly suitable for postgraduate education scenarios with complex professional backgrounds and high knowledge depth requirements.

[0199] Step 5: Teaching Effectiveness Evaluation and Feedback: By analyzing learning behavior data, such as the frequency of knowledge point access and test error patterns, the knowledge graph structure is optimized, the impact of the knowledge graph on learning effectiveness is quantified, and timely and effective teaching effectiveness evaluation and feedback are achieved to continuously improve teaching.

[0200] Its core logic lies in this process: it uses behavioral data to infer cognitive patterns, transforming the knowledge graph from a "static knowledge architecture" into a "dynamic cognitive mapping"—reflecting both the logical structure of subject knowledge and adapting to learners' actual cognitive paths. Ultimately, through a data-driven iterative mechanism, it achieves bidirectional optimization of both the knowledge graph and teaching effectiveness. This principle integrates multidisciplinary methods such as educational cognitive science (learning behavior analysis), graph theory and network science (graph structure optimization), and statistical learning (effectiveness quantification), forming a closed-loop teaching system.

[0201] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention have been clearly and completely described above in conjunction with the implementation methods of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the embodiments herein can be arranged and designed in various different configurations.

[0202] Therefore, the above detailed description of the embodiments of the present invention provided in the implementation is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

Claims

1. A knowledge graph-based curriculum method, characterized in that: It includes the following steps: Step 1: Curriculum System Construction: By constructing a knowledge graph for the course "3D Modeling and Additive Manufacturing of Metal Parts", we use Python and DrissionPage to obtain course chapter data from the course PPT, and then use pandas and py2neo to import the chapter file data and construct it into a neo4j graph; we call the large model openaigpt-4o, expand the graph data according to rule semantics and predefined entity relationships, and finally use Python + py2neo to read the JSON file to form a complete knowledge graph. Step 2: Optimize teaching content: Integrate the constructed knowledge graph with classroom teaching, starting from the knowledge domain, linking knowledge units, deconstructing knowledge points at all levels, and presenting visualized knowledge points including knowledge point profiles, knowledge point introductions, and knowledge point structures, thereby refining the "granularity" of knowledge. Step 3: Innovative Teaching Methods: Complete the classroom teaching practice linked to the knowledge graph. Based on different logical structural relationships such as symbiosis and progression, connect various knowledge points of the major to build a complete knowledge graph that includes dimensions such as training objectives, ability graph, problem graph, knowledge graph, and academic graph, forming a structured knowledge system. Organize accumulated teaching materials such as content videos, application cases, cutting-edge research, and student hot topics, and grid-link knowledge points from different fields and the teaching resources involved in each knowledge point to each knowledge point, forming a three-dimensional professional knowledge network. Step 4: Personalized learning path recommendation: Based on the differences in graduate students' professional backgrounds, learner profiles are constructed using knowledge graphs, and differentiated learning paths are recommended in combination with prior knowledge assessment. By using reinforcement learning to dynamically optimize recommendation strategies and dynamically adapt learning content, we can meet the personalized learning needs of different students. Step 5: Teaching effectiveness evaluation and feedback: By analyzing learning behavior data, optimizing the knowledge graph structure, quantifying the impact of the knowledge graph on learning effectiveness, and achieving timely and effective teaching effectiveness evaluation and feedback, so as to continuously improve teaching.

2. The knowledge graph-based curriculum method according to claim 1, characterized in that: Building upon step 1, this course establishes connections between courses such as mechanical design, theoretical mechanics, and fundamental engineering materials, while also incorporating ideological and political education and cutting-edge academic research to construct a cross-level, diversified, and highly integrated additive manufacturing talent training system.

3. The knowledge graph-based curriculum method according to claim 1, characterized in that: Building upon step 1, this step utilizes Python and DrissionPage to retrieve course chapter data from the course presentation slides. The steps are as follows: Step 1: Install DrissionPage and the python-pptx library for processing PPT slides; Step 2: Write a Python script to download the PPT slides via DrissionPage and then parse the chapter data; Step 3: Adjust the code according to the actual situation.

4. The knowledge graph-based curriculum method according to claim 1, characterized in that: Building upon step 1, the large model OpenAigpt-4O is invoked to expand the graph data based on rule semantics and predefined entity relationships. The core principle of this step is to combine natural language understanding capabilities with structured rule constraints, and generate entity relationship data that conforms to the graph architecture through the model's capture of semantic patterns and logical reasoning.

5. The knowledge graph-based curriculum method according to claim 1, characterized in that: Building upon step 1, the final step involves using Python and py2neo to read the JSON file and form a complete knowledge graph. This step is based on a three-layer linkage of "data structuring transformation + graph model mapping + database operations." Specifically: the data layer uses Python to parse the JSON text and extract semantic units of entities, relationships, and attributes; the model layer uses py2neo to transform these semantic units into graph model objects, adhering to graph theory structure and transaction consistency rules; and the storage layer interacts with Neo4j through the Bolt protocol, utilizing constraints, indexes, and other mechanisms to ensure the semantic integrity and query efficiency of the graph.

6. The knowledge graph-based curriculum method according to claim 1, characterized in that: Building upon step 2, and incorporating relevant cutting-edge academic research, we aim to help students establish a knowledge system for their courses and majors, enhance the breadth and depth of their course knowledge, and establish an innovative undergraduate-master's integrated training model that combines multiple fields and interdisciplinary approaches.

7. The knowledge graph-based curriculum method according to claim 1, characterized in that: Building upon step 3, an additive manufacturing experimental comprehensive teaching platform will be established for first-year professional master's students. This platform will form a deep learning model based on knowledge graphs, enabling innovative teaching methods such as case-based teaching and problem-oriented learning, and constructing a highly interactive and effective teaching method and talent training paradigm.