ms2vec
This is a python package for MSSG and NP-MSSG by Neelakantan. I already checked the score in SimLex-999 and that is almost same in the article.
Perhaps, I can response issuses until 2019/03/31 when I will get Master's degree.
The original information is like this.
NP-MSSG
the code from research paper "Efficient Non-parametric Estimation of Multiple Embeddings per Word in Vector Space" by Arvind Neelakantan, Jeevan Shankar, Alexandre Passos, Andrew McCallum
- the code provided by the authors of the paper. I took the code from [bitbucket account] (https://bitbucket.org/jeevan_shankar/multi-sense-skipgram/src/74aeafd22528710cd60c72d6d1d0c0edad37f567?at=master.),
- to download the vectors trained by the provided method, follow the link, archive size is 3.3Gb
github page
https://github.com/yauhen-info/NP-MSSG#np-mssg
how to use
This code is available in the same way as gensim because this program is the extenion of gensim 3.7.2.
However, the learning method of this program skip-gram and negative sampling only. Other combinations don't work.
MSSG
MSSG is executed if np-value is -1.
from gensim.models.word2vec import LineSentence
from ms2vec.ms2vec import MultiSense2Vec
corpus = LineSentence("your_corpus")
model = MultiSense2Vec(corpus, sg=1, negative=5, workers=4, iter=5, window=5,
min_count=10, min_sense_count=5000, max_sense_num=3,
size=300, np_value=-1, cv2zero=True, use_all_window=True, seed=777)
model_name = "mssg_model"
model.save(model_name)
# bin format is available in gensim's KeyedVectors
model.wv.save_word2vec_format(model_name+".bin", binary=True)
NP-MSSG
NP-MSSG is executed if np-value is not -1.
from gensim.models.word2vec import LineSentence
from ms2vec.ms2vec import MultiSense2Vec
corpus = LineSentence("your_corpus")
model = MultiSense2Vec(corpus, sg=1, negative=5, workers=4, iter=5, window=5,
min_count=10, min_sense_count=5000, max_sense_num=10,
size=300, np_value=-0.5, cv2zero=True, use_all_window=True, seed=777)
model_name = "npmssg_model"
model.save(model_name)
# bin format is available in gensim's KeyedVectors
model.wv.save_word2vec_format(model_name+".bin", binary=True)
Thank you for reading
I hope that this program works well and word embedding can contribute into NLP.
If this program is helpful for you, I want you to give the star this program for me. Have a nice day.