DysLexLens: A Low-Resource LLM Framework for Analysing Dyslexic Learners Insights from Online Forums

Short summary

DysLexLens is a low-resource LLM framework for understanding how dyslexic learners use AI tools by analyzing online forum discussions. It combines dictionary-driven noise filtering, knowledge-graph-based reasoning, and RAGAS evaluation metrics with human validation to ensure reliability and reduce hallucinations. The approach demonstrates generalizability to other low-resource forum contexts with code and datasets available for reproducibility.

• •Framework extracts insights from noisy forum data on how dyslexic learners actually use AI for reading, writing, and study tasks
• •Architecture includes four components: semantic filtering, KG-based reasoning, quantitative evaluation (RAGAS), and qualitative validation
• •Methodology tested on 30 questions over dyslexia-focused Reddit corpus; designed for reproducibility and adaptation to other low-resource contexts

Generated with AI, which can make mistakes.