Making neural content meaningful and truthful relying on OpenNLP tools
summary
A pipeline that cross-breeds raw and true content uses the OpenNLP similarity component to parse, form queries from noun/verb phrases, mine authoritative sources (Wikipedia, health sites), align semantic and syntactic graphs, and substitute real entities and sentences into neural skeletons. Results show discourse-marker conditioning plus value substitution improves coherence and factuality; a discriminator still separates outputs from humans roughly at ~0.8 accuracy, so quality rises but does not fully match human text. openNLP similarity component
AI-written summary · about
chapters (17)
- 0:00Introduction
- 2:04Raw content
- 4:02Motivations
- 5:38Natural Language Generation
- 6:43Neural Content
- 8:03Example
- 10:00Limitations
- 13:12Discourse
- 13:57Discourse Tree
- 16:27Discourse Generation
- 18:42Experiment
- 20:18Query
- 22:09Alignment
- 25:51Semantic representation
- 28:30Biography returns
- 30:14Personalized drug recommendation
- 34:37Summary
related talks
01h 31mNatural Language Understanding: Foundations and State-of-the-Art
30mSemantic Search and Neural Information Retrieval
54mData doesn't lie — but it can mislead: How to ensure integrity of your Machine Learning applications
01h 30mFrom Synthetic Examples to Production Signals
34mExposing Dark Data in the enterprise with custom NLP
55mDemystifying Natural Language Processing across several industry verticals
17mNatural Language Processing in Automation
44mThe Limits of Imagination : An Open Source journey