An AI-Enhanced System for Context-Aware Information Retrieval and Summarization in AI-Assisted Learning

Abstract

In AI-assisted learning environments, the capacity to effectively extract and summarize educational content is critical for increasing student comprehension and engagement. This work describes an AI-enhanced approach for context-aware information retrieval and summarization that is specifically designed to extract pertinent text and associated tables and graphics from academic PDFs. The system comprises two phases. The first step employs a hierarchical retrieval technique that uses topic-, heading-, and text-level similarity checks to extract contextually relevant content. In terms of retrieval speed, the framework outperforms existing solutions such as ChatGPT, DeepSeek, and Grok.ai, and it is the only one that supports the extraction of multimodal content, which is vital for technical subjects. In the second phase, the collected content is run through a fine-tuned T5-small language model that is designed for abstractive summarization. The fine-tuned model produces more accurate and cohesive summaries than the base model, while also retaining important visual and structural elements like figures and tables. Comparative evaluations show that the system outperforms in both retrieval performance and summarization quality. The suggested system bridges the gap between text-only and multimodal understanding, providing a scalable and domain-adaptive solution for intelligent content delivery in digital education platforms.