Semi-Automatic Indexing of Full Text Biomedical Articles

The following talk was presented at the Fall AMIA conference in Washington DC on October 25, 2005 by Clifford W. Gay. By selecting the "Slide 1: Start" link, or by selecting the picture of the first slide below, you can then proceed in a slide show format.


Slides: Select the slide below to begin slide show
Slide 1: Start
Slide 2: Acknowledgments
Slide 3: Outline
Slide 4: Introduction
Slide 5: Why Semi-Automatic Indexing?
Slide 6: Medical Text Indexer (MTI)
Slide 7: DCMS with MTI Suggestions
Slide 8: The Data: Online biomedical journals
Slide 9: Why Full Text?
Slide 10: Test Collection Selection
Slide 11: Test Collection
Slide 12: The Task: Emulate Medline indexing using full text
Slide 13: Indexing Task
Slide 14: Example Article
Slide 15: Evaluation
Slide 16: Results:Observations on PubMed Central articles
Slide 17: Section Header Classes
Slide 18: Section Class Performance
Slide 19: Results: Model selection results
Slide 20: Experiments
Slide 21: Production Baseline
Select here to begin slide show
Slide 22: Naive Mode
Slide 23: MetaMap Indexing Mode
Slide 24: Augmented Mode
Slide 25: Refined Augmented Mode
Slide 26: Full MTI Mode
Slide 27: Refined Full MTI
Slide 28: MTI Performance Summary
Slide 29: Results: Recent work
Slide 30: Improvement Potential
Slide 31: Increase REL Citations
Slide 32: Summarization
Slide 33: NLM Indexing Initiative
Slide 34: NONE Sections
Slide 35: Other Sections
Slide 36: Ranking Function
Slide 37: Tuning Path Weight
Slide 38: Partial Weight for Singleton Headers