Completed
Study on Label Efficiency (TACRED)
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Teaching Machines Through Human Explanations
Automatically move to the next video in the Classroom when playback concludes
- 1 Intro
- 2 A Surprisingly "Simple" Recipe for Modern NLP
- 3 Cost of data labeling: relation extraction
- 4 Cost of data labeling: more complex task
- 5 Workaround for (less) data labeling?
- 6 How "labels" alone could make things wrong
- 7 From "labels" to "explanations of labels" One explanation generalizes to many examples
- 8 Learning from Human Explanation
- 9 Our Focus: Natural Language Explanations
- 10 Learning with Human Explanations
- 11 Explanations to "labeling rules"
- 12 Generalizing explanations Matching labeling rules to create pseudo labeled data
- 13 Challenge: Language Variations
- 14 Neural Rule Grounding for rule generalization
- 15 A Learable, Soft Rule Matching Function
- 16 Neural Execution Tree (NEXT) for Soft Matching
- 17 Study on Label Efficiency (TACRED)
- 18 Results: Hate Speech (Binary) Classification
- 19 Take-aways . "One explanation generalizes to many examples" - better label efficiency vs. conventional supervision
