Key Topics

NLP-Based User Understanding

• Applying natural language processing to analyze learners’ story interactions and quiz responses
• Extracting keywords, concepts, and comprehension indicators from user-generated content
• Mapping extracted features to learning objectives and cognitive development levels

Dataset Construction & Topic Modeling

• Building a labeled dataset of stories, topics, and associated learner feedback
• Using unsupervised techniques (e.g., LDA, embeddings clustering) to identify topic relationships
• Generating topic transition graphs to recommend contextually relevant next topics

Recommendation System Integration

• Implementing a rule-based and model-based hybrid recommendation engine within the Flask backend
• Dynamically updating user profiles with learning history, feedback, and inferred interests
• Delivering next-topic suggestions through an interactive front-end UI post story interaction

Summary

NLP-Based User Understanding

Our team designed a system to implicitly analyze how some young learners engaged with an AI generated storybook and quizzes, through the automatic generation of content using natural language processing. Based on their responses, feedback, and keywords provided, we derived in-depth insights for understanding and the topic focus. We were able to derive something of a learner profile to understand their knowledge state and topic interests.

Dataset Construction & Topic Modeling

To inform personalized topic recommendations, we built a dataset of educational story content and tags for theme, the difficulty, and user responses. Then we applied topic modeling approaches using Latent Dirichlet Allocation (LDA) techniques and semantic embedding to cluster related theme clusters. We were able to generate a topic transition graph about how we modeled topic clusters to simulate a prediction of what topic the learner might benefit from.