Simple python implementation of two popular word embedding algorithm: Word2vec and GloVe.

• The project is only for educational purposes.
• The word2vec code is build under the instructions of cs224n assigment #1.
• The glove implementation is followed along with Hans blog.

• The existing dataset in this project is SST(Stanford Sentiment Treebank)
• SST contain sentiment analysis labels which can be used to evaluating the pros & cons of each embedding model.

• Install the dependencies (Python2.7)

pip install -r requirement.txt

• Download dataset

sh get_datasets.sh

• train word2vec

python train.py -m word2vec --save-every=True --vector-path=./model/word2vec -s 10 --learning-rate=0.3 -w 5 --iterations=40000

• train glove

python train.py -m glove -s 50 --learning-rate=0.05 --iterations=200 --save-every=True --vector-path=./model/glove