资 源 简 介
PLEASE NOTE THAT A NEW VERSION OF JDAGEEM IS HOSTED ON GITHUB.COM.
The URL is: https://github.com/shaybcohen/jdageem
JDageem is an extensible Java package that includes several implementations of parsing and training algorithms for dependency grammar induction. More specifically, JDageem includes:
An implementation of the split head automaton parsing algorithm of Eisner and Satta (1999), also specialized to handle the dependency model with valence of Klein and Manning (2004). Both Viterbi and the mininum Bayes risk decoding (Goodman, 1996) versions are implemented.
An implementation of the expectation-maximization algorithm for the DMV.
An implementation of the variational expectation-maximization algorithm of Cohen and Smith (2010), which makes use of the logistic normal priors (for the DMV).
An implementation of a variant of the harmonic DMV initializer that appears in Klein and Manning (2004).
