PhrazIt : Tool for automatic text summarization
summary
PhrazIt performs extraction summarization by ranking sentences with context-based indexing instead of just term frequency/position/length, building sentence nodes from contextualized phrasal vectors (LSA/LDA-augmented word2vec) and running a TextRank-style Markov ordering. It integrates AlchemyLanguage for concepts, produces cohesive reordering, supports single- and multi-document summaries with cross-information subsumption to remove paraphrases, and reports accuracy 70.02 and F-score 59.9 versus a baseline 40.3/27.8, with a live demo extracting correct action items.
AI-written summary · about
chapters (9)
- 0:00Introduction to PhrazIt
- 1:19Extraction vs abstraction summarization
- 3:06Context-based indexing in PhrazIt
- 4:51Unsupervised algorithm choice
- 6:37How PhrazIt ranks sentences
- 9:47PhrazIt performance results
- 10:31Live demo of PhrazIt
- 14:23Multi-document summarization
- 15:29Q&A session
related talks
58mAutomatic text summarization
22mEmpowering journalism-approaches to automatic text summarization
36mSAMENVATTR, or, How to Make a Text Summarization Tool for your Language of Choice
28mSemantic SentenceRank to Facilitate Reading....
26mUnsupervised learning for news summarisation
07mScreen2Words: Automatic Mobile UI Summarization with Multimodal Learning
06mHierarchical Summarization for Longform Spoken Dialog
39mMachine Learning for Relevance and Serendipity