Information Extraction: Relations, Events, and Time

21.8 Template filling

1. What are scripts?
   Abstract stereotypical situations or stories (Schank & Abelson, 1977): prototypical sequences of sub-events with participants and roles, whose expectations support classification, role assignment and inference of unstated facts.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 21, §21.8, p. 494
2. How is template filling modeled with machine learning?
   Two supervised systems: template recognition (sometimes called event recognition), a text classifier deciding whether a sentence invokes the template; and role-filler extraction, a separate classifier (or sequence model) per slot, e.g. LEAD-AIRLINE, AMOUNT.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 21, §21.8, pp. 494–495
3. How can non-identical fillers for the same slot (United vs United Airlines) be reconciled?
   With coreference resolution.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 21, §21.8, p. 495
4. What was the MUC-5 joint venture task, and how did FASTUS solve it?
   Producing hierarchically linked templates describing joint ventures (a TIE-UP template whose ACTIVITY slot is itself a template). FASTUS used a cascade of finite-state transducers with hand-written rules: tokenization, multiword and name recognition, noun/verb group chunking, complex groups, entity and event patterns filling templates, then merging references.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 21, §21.8, p. 495