The document summarizes research on using support vector machines (SVMs) and syntactic features to identify negated events in biomedical literature. Key points:
- The goal is to classify molecular events from abstracts as negated or not negated. SVMs are used with features including negation cues, part-of-speech tags, and the "command relation" between cues and events.
- Incrementally adding features such as negation cue positions, syntactic relations, and event types improves precision and recall from 14% to 51% while maintaining high specificity.
- The "command relation" captures syntactic scope and improves a rule-based baseline, but an SVM approach leveraging multiple linguistic features achieves better performance.