Coreference Resolution and Entity Linking

24.2 Coreference tasks and datasets

1. What must an end-to-end coreference system deal with on raw text?
   Detecting mentions, pronominal anaphora (her → Victoria Chen), filtering non-referential pronouns (pleonastic it), definite NPs (the 38-year-old, the company) and name variants (Megabucks = Megabucks Banking).
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 24, §24.2, p. 547
2. Why are scores much higher on corpora that don’t annotate singletons?
   Singletons are the majority of mentions in running text (60–70% of entities) and are hard to tell from non-referential NPs.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 24, §24.2, p. 548
3. What is gold mention detection in coreference evaluation?
   The system is given human-labeled mention boundaries and only has to cluster them.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 24, §24.2, p. 548
4. What are the main characteristics of OntoNotes coreference annotation?
   English and Chinese (about a million words each, many genres) plus Arabic newswire; the CoNLL-2012 shared task is based on it. No singletons; generics and pleonastic pronouns unmarked; appositives included inside the mention; prenominal modifiers are mentions only if proper nouns (UN in UN policy, not wheat in wheat fields).
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 24, §24.2, p. 548
5. Name coreference corpora that mark richer phenomena than OntoNotes.
   ISNotes (information status, bridging), LitBank (literary novels, with singletons), AnCora-CO (Spanish and Catalan, with discourse deixis), ARRAU (all NPs including singletons; dialogue and fiction; bridging, discourse deixis, generics, ambiguity).
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 24, §24.2, p. 548