DysLexLens framework mines Reddit to analyze dyslexic learners' AI tool usage

DysLexLens is a novel framework developed by Dana Rezazadegan and eight co-authors to systematically analyze how dyslexic learners experience AI tools (like reading and writing assistants) by extracting insights from online forums—specifically Reddit. The framework is designed as a low-resource, end-to-end pipeline that addresses the challenge of noisy, unstructured social media data. It first applies dictionary-driven filtering to build a focused corpus on dyslexia and AI, removing unrelated posts. Then, it integrates LLM-assisted semantic analysis with a knowledge graph (KG) to provide evidence-traceable query reasoning, meaning each answer can be traced back to specific forum posts. DysLexLens also includes quantitative evaluation via RAGAS and Query Robustness metrics, plus structured qualitative guidelines to assess hallucination and evidence alignment. The researchers tested the framework using dyslexia-related Reddit data and 30 questions, demonstrating its effectiveness and potential generalizability to other low-resource, forum-based contexts. The full system, sample data, and evaluation results are available on GitHub for reproducibility.

By surfacing the real-world, organic feedback of dyslexic learners on AI tools, DysLexLens offers a scalable, transparent method for researchers and product teams to understand what works—and what doesn't—for this underserved population. Its low-resource design means it can be applied to other niche communities that rely on forum discussions, from ADHD support groups to language learners. The combination of automated filtering, knowledge-graph reasoning, and rigorous evaluation makes it a practical tool for extracting actionable, evidence-backed insights from the messiness of online conversations.