Text2Vector generated word-embeddings for a given text.
- FastText -
pip3 install fasttext - NLTK -
pip3 install nltk - NumPy -
pip3 install numpy
Word embedding model - Skipgram - trained on raw text, with target size of 100-dim.
- For every word in text
- extract word-embedding from pre-trained model
- Element-wise addition of word-embedding with earlier embedding
- Averaging the word embeddings to obtain text embedding
Clone the repo, save word-embedding model from "Release" under the top-directory.
from text2vector import Text2Vector
path_to_word_embedding_model = 'skipgram-100/skipgram.bin'
text2vec = Text2Vector(path_to_word_embedding_model)
print('Generating embedding ..')
text = "Any text sample here!!!"
print(text2vec.convert(text))
"""
[ 0.44153227 0.12361106 0.00607821 0.53515722 0.28272 -0.11348521
-0.22341401 0.16395177 -0.01178203 0.06159557 0.01883419 -0.12652569
-0.15079305 -0.04731999 0.20319002 -0.05538354 -0.33085034 0.37050098
0.09624304 -0.42489241 -0.12226875 -0.2626278 -0.06713643 -0.31778053
0.34330154 0.27109923 0.15488822 0.17079256 -0.15065777 0.1636747
-0.12178582 0.08251881 -0.21455672 0.47482748 -0.31790714 -0.12989535
0.22780022 -0.08566074 -0.21855019 -0.31932005 -0.13109301 0.00146381
0.26255177 0.33713142 0.19220728 -0.10542174 -0.37171204 0.03750945
0.1633902 0.12948658 -0.19956239 -0.1510046 0.12710889 -0.10108179
-0.07666321 0.42590444 -0.07935622 0.07618748 -0.11696555 -0.12906784
0.07358533 -0.34065487 0.8037034 -0.02860275 0.06394197 0.23639533
-0.2513136 -0.28632619 0.12479368 -0.24286127 0.05824339 -0.17754047
-0.04253732 0.34774542 -0.10136852 0.44451588 0.07106352 0.62934348
0.38469637 0.13533326 0.00748966 0.17116917 0.14566171 -0.01778589
0.22916658 0.22703299 0.17693119 -0.1667371 -0.44423185 0.23692855
-0.0115481 0.081972 0.04752844 0.45813912 -0.13476718 0.23231126
0.06251153 0.07475965 0.25681703 0.02036907]
"""
