AI-assisted Context-Aware Document Reader

Problem Statement

Long, complex documents (e.g., research articles, reports, textbooks) are full of cross-references to figures, tables, sections, hypotheses, and related concepts. Navigating these references often forces readers to interrupt their flow, scroll around, and lose context. This thesis explores an AI-augmented reading interface that detects references and entities in a document and presents inline, context-aware previews (e.g., a tooltip or side pane with the referenced table, image, section summary, or definition). The goal is to reduce navigation overhead, support comprehension, and keep readers “in the flow.” By combining document parsing with large language models (LLMs) and thoughtful interaction design, the project will investigate how context-sensitive assistance can make reading and sensemaking more fluid and transparent.

Tasks

• Survey relevant work on AI-assisted reading, document parsing, and interaction techniques for sensemaking to ground the design.
• Design and prototype an interactive reader that surfaces context-aware previews for referenced content.
• Implement LLM- or rule-based methods for reference detection, semantic linking, and concise, audience-appropriate summaries.
• Evaluate the prototype with target users to assess usability, comprehension support, and reduction of navigation effort.
• Document the system and findings in a BA/MA thesis and present the results in the Disputationsseminar.
• (Optional) Prepare a manuscript for submission to an HCI venue.

Skills Required

• Solid programming skills (e.g., Python, TypeScript/JavaScript) and experience building web applications.
• Frontend development with React; backend with Node.js or Python (FastAPI/Flask) for service integration.
• Familiarity with LLMs/NLP and document processing (PDF/HTML parsing) is a plus.
• Interest in UX design and prototyping; ability to translate user needs into interaction patterns.