These are the code and data for 'Joint Semantic Relevance Learning with Text Data and Graph Knowledge'

before running the code, please download data from "http://pan.baidu.com/s/1bn0LJun" and unzip it into ./data/


==Directory==
./data
	corpus.4word2vec:training corpus for word vector.
	test.animal-143: animal-143 test set/
	test.sim-301: sim-301 test set. Note that each word in sim-353 may match several words in 
                     WordNet due to ambiguity. Here in our experiments, for each 
                     pair in sim-301, we compare the combination of all cases and 
                     take similarity value of the most similar pair as its result.
	train.wordnet-noun.pairs:graph training data of wordnet-noun.
	train.wordnet-noun.wikipage.filter:joint text training data of wordnet-noun.
	train.yago-animal.pairs:graph training data of yago-animal.
	train.yago-animal.wikipage.filter:joint text training data of yago-animal.
	train.wordnet-all.pairs:all the graph of Wordnet, which contains 15 types of relations.
	train.wordnet-all.wikipage:joint text training data of all wordnet.
	word2vec.100:the model we trained using word2vec with dimension 100.
	word2vec.200:the model we trained using word2vec with dimension 200.
	word2vec.50:the model we trained using word2vec with dimension 50.
	word2vec.wordnet-noun.prt.100:the file of using word2vec to initialize entry vector of wordnet-noun, dimension 100.
	word2vec.wordnet-noun.forpv.200:the file of using word2vec to initialize entry vector of wordnet-noun, dimension 200.
	word2vec.wordnet-noun.forpv.50:the file of using word2vec to initialize entry vector of wordnet-noun, dimension 50.
	word2vec.yago-animal.forpv.100:the file of using word2vec to initialize entry vector of yago-animal, dimension 100.
	word2vec.yago-animal.forpv.200:the file of using word2vec to initialize entry vector of yago-animal, dimension 200.
	word2vec.yago-animal.forpv.50:the file of using word2vec to initialize entry vector of yago-animal, dimension 50.
	
./src
	main.cpp

==Compile==
"g++ src/main.cpp -o JTGR -lpthread"

==Implement Experiments==
"./JTGR" for instruction.

==NOTE==
The parameter setup of word2vec in our experiment is as followed,
"./word2vec -train corpus.4word2vec -output word2vec.dim100 -cbow 0 -size 100 -window 8 -negative 25 -hs 1 -sample 1e-4 -threads 10 -binary 0 -iter 10"