Deep learning implementation for NLP with NNabla

Tiny implementation of deep learning models for NLP with Sony's NNabla.

Tested environment

• Python 3.7.2
• NNabla v1.0.11

Models

Language models

• A vanilla recurrent neural network language model (language-models/rnnlm/)
• LSTM language model (language-models/lstmlm/)
• Character-level convolutional LSTM language model (language-models/char-cnn-lstmlm/)
• Continuous Bag-of-Words (CBOW) model (language-models/cbow/)

Word embeddings

• GloVe model (word-embeddings/glove/)
• Poincaré embeddngs (word-embeddings/poincare-embeddings/)

Seq2Seq models

• Encoder-decoder (seq2seq/encoder-decoder/)
• Encoder-decoder + global attention (seq2seq/encoder-decoder-with-attention/)

Text classifiers

• fastText (text-classifications/fasttext/)
• Self attention (text-classifications/self_attention/)
• LSTM classifier (text-classifications/lstm-classifier/)

Future work

• Skip-gram model
• Peephole LSTM
• GRU
• Transformer
• ELMo
• etc.