graehl/kenlm
KenLM: Faster and Smaller Language Model Queries
Automatically exported from code.google.com/p/cld2
Public repository record indexed from GitHub. Explore verified star velocity metrics, source code repositories, and curated developer tool directories across the GitHubRepo ecosystem.
KenLM: Faster and Smaller Language Model Queries
USC/ISI nplm 0.1 via kpu/nplm fork, updated for C++11
Fork of http://nlg.isi.edu/software/nplm/ with some efficiency tweaks.
GIZA++ is a statistical machine translation toolkit that is used to train IBM Models 1-5 and an HMM word alignment model. This package also contains the source for the mkcls tool which generates the word classes necessary for training some of the alignment models.