nObjects:
words in text
nAttributes:
sense, gender, number, pos, …
nLinks:
nGrammatical
relations (subject-object, modifier,…)
nClose
semantic relations (is-a, cause-of, …)
nSame
word in different sentences (one-sense-per-discourse)
nCompatibility
parameters:
nLearned
from tagged data
nBased
on prior knowledge (e.g., WordNet, FrameNet)
Can we infer grammatical structure
and disambiguate word senses
simultaneously rather than sequentially?
Can we integrate inter-word relationships directly into our probabilistic model?