nlp - what are the methods to estimate probabilities of production rules? -
i know n-gram useful finding probability of words,i want know, how estimate probabilities of production rules? how many methods or rules calculate probabilities of production rules?
i not find blog or on topic.now studying on probabilistic context free grammar & cky parsing algorithm.
as understand question, asking how estimate parameters of pcfg model data.
in short, it's easy make empirical production-rule probability estimates when have ground-truth parses in training data. if want estimate probability s -> np vp
, count(s -> np vp) / count(s -> *)
, *
possible subtree.
you can find more formal statement in lots of places on web (search "pcfg estimation" or "pcfg learning"). here's nice 1 michael collins' lecture notes: http://www.cs.columbia.edu/~mcollins/courses/nlp2011/notes/pcfgs.pdf#page=9
Comments
Post a Comment