Creating the sentence embedding using the auto-encoders in pytorch
The dataset used here are hotel reviews obtained online. The autoencoder here as bi-LSTMs as the encoder and decoder with no dropout. The word embeddings are generated using gensim's word2vec.
The accuracy metric is BLEU score using the smoothing function from nltk.
- Python 3.6.4
- pytorch 1.2.0+cu92
- matplotlib 3.1.2
- pandas 0.22.0
- re 2.2.1
- gensim 3.8.1
- numpy 1.17.0
- sklearn 0.19.1
- nltk 3.2.5
python run.py