Knowledge Scout: From Talks and Slides to Smart Summaries
summary
A working pipeline converts scraped conference video, slides and metadata into well-structured markdown: a Python project called secretary service runs OCR and Whisper, normalizes formats, and emits talk pages and a four-minute human summary. A retrieval step ranks ~5,000 text fragments down to ~100 for answer generation, then composes context-aware responses with a dynamic table of contents and perspective switching (policy maker, student, child). Libraries sync via file-system library or Nextcloud, code is open source and usable but acknowledged as unfinished.
AI-written summary · about
chapters (6)
- 0:00Introduction and motivation
- 4:01Building the knowledge pipeline
- 7:32Smart summaries and insights
- 10:09Personalized and contextual answers
- 13:22Collective knowledge vision
- 17:15Future applications and Q&A
related talks
35mThe Podcast Pipeline Nobody Sees: Behind the Scenes with AI Agents and AI gen
22mConversations With AI: Transforming Conference Learning & Data Engagement
21mChalkTalk: Globalizing Data Science Education with AI-generated Videos
06mHierarchical Summarization for Longform Spoken Dialog
01h 25mLearn to Unlock Document Intelligence with Open-Source AI
23mHidden in plain sight: Using law to summarize the law
40mTurn 10,994 Notes Into Memory
27mPractical Sphinx