Online course teaching decision-making auxiliary method and system based on multi-modal artificial intelligence

Abstract

The invention discloses an online course teaching decision-making auxiliary method and system based on multi-modal artificial intelligence, and belongs to the crossing field of education technology science and artificial intelligence technology, and the method comprises the steps: obtaining student comment data from an online course platform, carrying out the preprocessing, and constructing a labeling data set; based on a pre-trained text classification model, comments are classified according to preset dimensions, and multi-dimensional visual analysis of teaching feedback is realized in combination with sentiment analysis and keyword extraction technologies; aiming at the negative emotion comments, guiding a large language model through prompt language engineering to carry out deep semantic analysis, and identifying specific problems and roots thereof; based on the learning analysis technology framework, a targeted teaching optimization scheme is generated; the system is composed of a data layer module, an analysis layer module and a decision layer module and executes corresponding steps in the method. The system provided by the invention can be embedded into an existing online learning platform to serve as a teacher end auxiliary tool, supports normalized and periodic teaching decision and optimization, and has the characteristics of high practicability and easiness in popularization.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of educational technology and artificial intelligence, and specifically relates to a method and system for assisting online course teaching decisions based on multimodal artificial intelligence. Background Technology

[0002] In today's large-scale online learning environment, student-generated course review data is a crucial reflection of their learning experiences and needs. Uncovering the value behind this data is key to helping teachers make data-driven instructional decisions. Existing technologies have already attempted to utilize intelligent methods to analyze course review data.

[0003] Existing techniques (Xu Zhenguo et al., "Research on Mining User Needs of Online Courses by Integrating BERTopic and KANO Models" [J]. Information Science, 2024, 42(8): 126-135) utilize deep learning models such as BERTopic to mine topics and perform sentiment analysis on course reviews in order to identify user needs. However, such methods mostly remain at the level of data analysis and visualization. Their analysis results are often indicative or descriptive, lacking in-depth semantic interpretation of the review content and targeted mathematical decision-making suggestions. Teachers still need to rely on their own experience to transform the analysis results into specific teaching improvement plans, making it difficult to guarantee the efficiency and quality of decision-making.

[0004] Existing technique two (Liu Qingtang et al., "A Study on the Influencing Factors of MOOC Course Quality Based on Learner Review Data Mining" [J]. Journal of Distance Education, 2023, (1): 80-90) uses LTP part-of-speech tagging and dependency parsing techniques to extract influencing factors of course quality. Such methods focus more on constructing a course evaluation system rather than directly serving teachers' teaching decision-making cycle. The analysis results do not provide sufficient support for teachers to make immediate and accompanying teaching decisions, and it is difficult to form a closed loop of "data-analysis-decision-intervention".

[0005] In summary, the main shortcomings and deficiencies of existing technologies are as follows: on the one hand, the decision support is not deep enough, and it often fails to deeply understand and explore the specific problems reflected in student comments, and fails to generate feasible teaching optimization solutions in a timely manner; on the other hand, the technology is not directly implemented, and teachers need to spend a lot of time interpreting and analyzing the model's operating structure, and need to conceive practical solutions on their own in combination with the relevant system and the actual situation of the course, which puts a certain pressure on teachers' routine teaching decision-making improvements. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention proposes an online course teaching decision support method and system based on multimodal artificial intelligence. This method automatically and accurately identifies key issues affecting the learning experience from massive amounts of unstructured student comment texts. By integrating a learning analytics framework with a large language model, the identified student needs are automatically transformed into specific teaching decision solutions that are theoretically sound, highly operable, and targeted, thereby significantly improving the efficiency and accuracy of teachers' teaching decisions.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0008] A decision support method for online course teaching based on multimodal artificial intelligence includes the following steps:

[0009] Step 1: Data Collection and Preprocessing: Crawl student course review data from online course platforms, perform preprocessing, and construct a labeled dataset;

[0010] Step 2: Perform sentiment analysis and visualization on the course review data, including: classifying the course review data according to preset dimensions using a pre-trained text classification model based on the labeled dataset; calling natural language processing tools to perform sentiment analysis on the course review data under each dimension and extracting keywords; and visualizing the classification results, sentiment distribution, and keywords.

[0011] Step 3: Intelligent Teaching Decision Generation: Based on the obtained negative sentiment comment data, a large language model is used to guide the large language model to conduct in-depth semantic analysis of negative comments under specific dimensions by building a prompting project, and to summarize the specific problems and root causes of student feedback; based on the learning analytics technology framework, the large language model is guided by the prompting project to generate specific teaching optimization solutions for the summarized problems.

[0012] Furthermore, the text classification model is an optimized ERNIE model, with preset dimensions including content experience, learning assistance experience, teacher evaluation, and overall evaluation.

[0013] Furthermore, the natural language processing tool uses Baidu's sentiment analysis API.

[0014] Furthermore, the large language model is the DeepSeek-R1 model.

[0015] Furthermore, the prompting engineering includes necessary elements and optional elements. The necessary elements include the roles, objectives, and contexts set for the large language model, while the optional elements include processing steps, analysis requirements, analysis methods, and output formats.

[0016] Furthermore, the prompting project assigns the role of the large language model as an education quality optimization consultant, with the goal of generating course optimization suggestions based on negative sentiment comments and analyzing issues in student comments based on specific online course guidance.

[0017] Furthermore, the processing steps include: summarizing all negative comment data and clarifying the dimension to which each comment belongs; attributing the problems to their causes and exploring the issues and root causes reflected in the number of negative comments in each dimension; and designing improvement plans for each problem, based on cutting-edge teaching theories or teaching practice strategies and driven by students' learning needs or problems, designing course improvement or optimization plans one by one for the specific problems reflected in the negative comment data of different dimensions.

[0018] Furthermore, the analysis requires that, based on cutting-edge teaching theories, and from four perspectives—student learning perspective, teacher instructional design perspective, teacher teaching practice process perspective, and teaching management perspective—actionable suggestions be provided for the specific issues reflected in the negative comments data across each dimension.

[0019] Furthermore, the analytical methods include problem assessment, targeted intervention, personalized recommendations, and deep reflection. These analytical methods together constitute a learning analytics framework, guiding the large language model to focus on the student's learning process and providing targeted services for optimizing teaching decisions from the student's perspective.

[0020] Furthermore, the generated suggestions are in JSON format, and each suggestion must reference specific comment content, with a maximum of two referenced comment IDs.

[0021] This invention also protects an online course teaching decision support system based on multimodal artificial intelligence, comprising: a data layer module for collecting and storing student comment data from the online course platform, and performing preprocessing and annotation; an analysis layer module for configuring a text classification model, calling sentiment analysis APIs and keyword extraction APIs, and performing visual analysis and mining of student course comment data; and a decision layer module for combining learning analytics technology to build a prompting project, guiding a large language model to attribute the problems reflected in student comments, and generating teaching decision schemes to assist teachers in making accurate teaching decisions.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] This invention employs multi-module artificial intelligence technology, from ERNIE model categorization to Baidu sentiment analysis API and keyword extraction, to DeepSeek large model-assisted generation of teaching decisions, forming an integrated intelligent analysis process that deeply mines the potential value in student course review data, thereby generating effective teaching decision-making solutions and resulting in a deeper level of educational value mining.

[0024] This invention can quickly generate practical and effective course improvement plans based on student course review data, and can efficiently assist teachers in making routine teaching decisions to accurately grasp the learning needs of students, continuously iterate and optimize teaching plans, and greatly improve the efficiency of teaching decisions.

[0025] The system provided by this invention can be directly embedded into existing Massive Open Online Courses (MOOC) platforms or other online learning management systems as an auxiliary tool for teachers, supporting routine and periodic teaching optimization. It is highly practical and easy to promote. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the overall process of the method of the present invention.

[0027] Figure 2 This is a visual result of student comments on the course "Psychological Science Popularization - College Students' Mental Health" based on the present invention;

[0028] Figure 3 This is a keyword cloud map of negative comments in the "learning assistance experience" dimension of student comments on the course "Psychological Science Popularization - College Students' Mental Health" generated by this invention. Detailed Implementation

[0029] To make the technical solution of the present invention clearer, the technical solution of the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

[0030] like Figure 1 As shown, the online course teaching decision support method based on multimodal artificial intelligence provided by the present invention includes the following steps:

[0031] Step 1: Data Collection and Preprocessing: Student comments for the course "Psychological Science Popularization - College Students' Mental Health" from September 11, 2023 to August 1, 2025 were collected from the MOOC platform using web crawling technology. After data cleaning and deduplication, 1,898 valid comments were obtained and stored in a MySQL database as a labeled dataset.

[0032] Step 2: Perform sentiment analysis and visualization on the course review data, including: Based on the labeled dataset, use the pre-trained ERNIE-3.0-mini-zh model to classify the course review data according to four dimensions: content experience, learning assistance experience, teacher evaluation, and overall evaluation; call the Baidu sentiment analysis API and keyword extraction API to perform sentiment analysis and keyword extraction on the course review data for each dimension; visualize the classification results, sentiment distribution, and keywords, such as... Figure 2As shown, negative sentiment comments accounted for a relatively high proportion in the "learning assistance experience" dimension, and a word cloud was generated from the keywords of negative comments in the "learning assistance experience" dimension (e.g.) Figure 3 (As shown), to visually present the focus of student feedback;

[0033] Step 3: Intelligent Teaching Decision Generation: Based on the obtained negative sentiment comment dataset in the "learning assistance experience" dimension, a prompting project is built and input into the DeepSeek-R1 large language model. This guides the DeepSeek-R1 large language model to conduct in-depth semantic analysis of the negative comments in the "learning assistance experience" dimension, summarizing the specific problems and root causes of student feedback. Based on the learning analytics framework and combined with teaching theories including active learning strategies, specific teaching optimization solutions are generated for the summarized specific problems.

[0034] In one specific embodiment, in response to the problem of "monotonous teaching methods" identified in the DeepSeek-R1 large language model, the following suggestions are generated: From the perspective of teacher instructional design, based on active learning strategies, some PPT reading sessions should be changed to problem-based learning (PBL), and pre-class thinking questions should be designed; From the perspective of teacher teaching practice, interactive Q&A sessions should be added when explaining key and difficult points to promote students' active thinking.

[0035] Finally, the effectiveness of the solution was verified through interviews with instructors and analysis based on the Technology Acceptance Model (TAM). Instructor feedback indicated that the generated solution accurately identified needs and assisted in targeted interventions; the system's interactive logic was clear, facilitating routine improvements. This demonstrated the technical usefulness and ease of use of the invention, thus fully proving the feasibility and effectiveness of the technical solution.

[0036] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made by those skilled in the art within the scope of the technology disclosed in this invention, based on the technical solution and concept of the present invention, should be included within the protection scope of this invention. Therefore, the protection scope of this invention should be determined by the scope of the claims.

Claims

1. A decision support method for online course teaching based on multimodal artificial intelligence, characterized in that, Includes the following steps: Step 1: Data Collection and Preprocessing: Crawl student course review data from online course platforms, perform preprocessing, and construct a labeled dataset; Step 2: Perform sentiment analysis and visualization on the course review data, including: classifying the course review data according to preset dimensions using a pre-trained text classification model based on the labeled dataset; calling natural language processing tools to perform sentiment analysis on the course review data under each dimension and extracting keywords; and visualizing the classification results, sentiment distribution, and keywords. Step 3: Intelligent Teaching Decision Generation: Based on the obtained negative sentiment comment data, a large language model is used to guide the large language model to conduct in-depth semantic analysis of negative comments under specific dimensions by building a prompting project, and to summarize the specific problems and root causes of student feedback; based on the learning analytics technology framework, the large language model is guided by the prompting project to generate specific teaching optimization solutions for the summarized problems.

2. The online course teaching decision support method based on multimodal artificial intelligence according to claim 1, characterized in that, The text classification model is an optimized ERNIE model, with preset dimensions including content experience, learning assistance experience, teacher evaluation, and overall evaluation.

3. The online course teaching decision support method based on multimodal artificial intelligence according to claim 1, characterized in that, The natural language processing tool uses Baidu's sentiment analysis API.

4. The online course teaching decision support method based on multimodal artificial intelligence according to claim 1, characterized in that, The large language model is the DeepSeek-R1 model.

5. The online course teaching decision support method based on multimodal artificial intelligence according to claim 1, characterized in that, The prompting engineering includes necessary elements and optional elements. The necessary elements include the roles, objectives, and contexts set for the large language model, while the optional elements include processing steps, analysis requirements, analysis methods, and output formats.

6. The online course teaching decision support method based on multimodal artificial intelligence according to claim 5, characterized in that, The prompting project assigns the role of the large language model as an educational quality optimization consultant. Its goal is to generate course optimization suggestions based on negative sentiment comments and to analyze issues in student comments based on specific online course guidance.

7. The online course teaching decision support method based on multimodal artificial intelligence according to claim 5, characterized in that, The processing steps include: summarizing all negative review data and clarifying the dimension to which each review belongs; attributing the problems to their causes and exploring the issues and root causes reflected in the negative review data of each dimension; and designing improvement plans for each problem, based on cutting-edge teaching theories or teaching practice strategies and driven by students' learning needs or problems, and designing course improvement or optimization plans one by one for the specific problems reflected in the negative review data of different dimensions.

8. The online course teaching decision support method based on multimodal artificial intelligence according to claim 5, characterized in that, The analysis requires that it be based on cutting-edge teaching theories and should take into account four perspectives: student learning, teacher instructional design, teacher teaching practice, and teaching management. It should provide actionable suggestions for the specific problems reflected in the negative comments data of each dimension.

9. The online course teaching decision support method based on multimodal artificial intelligence according to claim 5, characterized in that, The analytical methods include problem assessment, targeted intervention, personalized recommendations, and deep reflection. These methods together constitute a learning analytics framework that guides the large language model to focus on the student's learning process and provides targeted services for optimizing teaching decisions from the student's perspective.

10. An online course teaching decision support system based on multimodal artificial intelligence, characterized in that, The method for implementing the online course teaching decision support method based on multimodal artificial intelligence as described in any one of claims 1 to 9 includes: a data layer module for collecting and storing student comment data from an online course platform, and performing preprocessing and annotation; an analysis layer module for configuring a text classification model, calling sentiment analysis APIs and keyword extraction APIs, and performing visual analysis and mining of student course comment data; and a decision layer module for combining learning analytics techniques to build a prompting project, guiding a large language model to attribute the problems reflected in student comments, and generating teaching decision schemes to assist teachers in making accurate teaching decisions.

Images