# Parsing of Stochastic Contex-Free Grammars (SCFGs)

I am interested in parsing of general SCFGs.

I am aware of the Earley parser for the general CFGs. The only general algorithm for parsing SCFGs that I am aware of is the Earley-Stolcke parser : http://arxiv.org/abs/cmp-lg/9411029

1) Is there more recent work on the subject?

2) Is there publicly available software for deriving the Viterbi parse, compute prefix probabilities, and estimating rule probabilities of SCFGs?

Thank you very much!

-

Both Earley parser and CYK parser can be used for parsing SCFG. See http://en.wikipedia.org/wiki/CYK_algorithm#Parsing_weighted_context-free_grammars

There are many software that include a SCFG parser, for example, Berkeley parser. Estimating rule probabilities from training corpus with parse trees annotated is straightforward, i.e., counting and normalization. Estimating rule probabilities from unannotated training corpus can be done using EM (the inside-outside algorithm) or MCMC.

-