Hierarchical Summarization for Longform Spoken Dialog
summary
A system combines a two-stage ASR+summarization pipeline with semantic segmentation and merging algorithms to make 20–30 minute spoken recordings skimmable; it outputs hierarchical summaries (short→detailed), links back to ASR transcripts and audio, and exposes editing for error recovery. Evaluation on 10 participants scores 556 summaries with 71 immediately understandable summaries and 98 understandable after checking, and users spend about 27% of the listening time to find points of interest.
AI-written summary · about
related talks
58mAutomatic text summarization
36mSAMENVATTR, or, How to Make a Text Summarization Tool for your Language of Choice
30mChallenges & Solutions for AI-Based Message Summarization
22mEmpowering journalism-approaches to automatic text summarization
49mAutomatic Summarization of 10K Daily Meetings: from Requirements to Production
20mKnowledge Scout: From Talks and Slides to Smart Summaries
26mUnsupervised learning for news summarisation
19mPhrazIt : Tool for automatic text summarization