GitHub - srush/relax-decode

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 211 Commits
data_management		data_management
examples		examples
experiments		experiments
formats		formats
graph		graph
hypergraph		hypergraph
interfaces		interfaces
lattice		lattice
lp		lp
mrf		mrf
optimization		optimization
parse		parse
phrasebased		phrasebased
scripts		scripts
tagger		tagger
third-party		third-party
trans_decode		trans_decode
transforest		transforest
writing		writing
.gitignore		.gitignore
CommandLine.h		CommandLine.h
CubeLM.cpp		CubeLM.cpp
CubeLM.h		CubeLM.h
DecompParser.cpp		DecompParser.cpp
DecompTagger.cpp		DecompTagger.cpp
Doxyfile		Doxyfile
DualDecompTagger.cpp		DualDecompTagger.cpp
Exact.cpp		Exact.cpp
FullParser.cpp		FullParser.cpp
FullTagger.cpp		FullTagger.cpp
MRFSolver.cpp		MRFSolver.cpp
Main.cpp		Main.cpp
Marginals.cpp		Marginals.cpp
Parse.cpp		Parse.cpp
PhraseLP.cpp		PhraseLP.cpp
PhraseViterbi.cpp		PhraseViterbi.cpp
PottsTagger.cpp		PottsTagger.cpp
README		README
README.isi		README.isi
Rates.h		Rates.h
Run.cpp		Run.cpp
SConstruct		SConstruct
SimpleViterbi.cpp		SimpleViterbi.cpp
Tag.cpp		Tag.cpp
TransLP.cpp		TransLP.cpp
build_config.py.template		build_config.py.template
common.h		common.h
doxygen.py		doxygen.py
protoc.py		protoc.py
util.cpp		util.cpp
util.h		util.h

Repository files navigation

relax-decode: a framework for decoding with relaxation in NLP. 


At its current state it consists of a series of libraries for various problems and methods. 

graph/ - Algorithms and data structures for simple graph problems

hypergraph/  - Algorithms and data structures for hypergraph decoding. Includes viterbi, inside-outside, A* search, cube pruning, fsa intersection.

interfaces/ - External interfaces for the library implemented through Protocol buffers. Interfaces for hypergraph (translation, parse, tagging) and graph(MRF, fsa).  

lp/ - (Integer) Linear programming interface. Includes bindings for general hypergraph, parsing, tagging, MRF.   

optimization/ - Optimization algorithms. Mainly subgradient descent, dual decomposition, MRF (move).  

parse/ - Wrapper for parsing problems (as hypergraph). 

tagger/ - Wrapper for tagging problems (as lattice). 

transforest/ - Wrapper for translation forests (as hypergraph).