Knowledge Extraction from Video Training Material using Retrieval-Augmented Generation
Abstract
The increasing availability of video-based training material has created new challenges related to knowledge extraction, accessibility and information retrieval. While Automatic Speech Recognition (ASR) systems can transform speech into text, the resulting transcripts are often difficult to search and reuse efficiently. Retrieval-Augmented Generation (RAG) systems provide an opportunity to combine semantic retrieval and Large Language Models (LLMs) in order to support grounded question answering over unstructured transcript data. In this work, technical training videos were transcribed into text using the Whisper ASR model and processed through a RAG pipeline implemented using Langflow. The pipeline performed text splitting, embeddings generation, vector storage using Chroma DB, semantic retrieval and grounded answer generation, where multiple embedding and language model configurations were evaluated. Additionally, an evaluation pipeline based on an LLM-as-a-judge approach was implemented in order to assess the system performance quality. The evaluation framework analyzed metrics including faithfulness, context relevance, answer relevance, completeness, accuracy, hallucinations and retrieval failures across fifty evaluation questions. According to the experimental results, all configurations achieved high performance, particularly with respect to faithfulness and answer accuracy. Among the evaluated configurations, the Nomic-GPT implementation achieved the strongest…
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.