Neural Tree Indexers
Chainer implementation of Neural Tree Indexers for Text Understanding.
This implementation contains:
- Full tree matching NTI-SLSTM-LSTM
- Combines both recurrent sequential and recursive tree models
- Performs tree matching with standart LSTM units
- Fast global and tree attentions described in the paper
- Python 2.7
- chainer (tested on chainer 1.7.1 and 1.12.0)
- Other data utils: sklearn, pandas, numpy etc.
To train a model with SNLI dataset:
$ python train_snli.py --snli path/to/snli_1.0 --glove path/to/glove.840B.300d.txt
Full tree matching NTI-SLSTM-LSTM model with global attention achieves around 87.3% accuracy on Stanford NLI dataset.
Tsendsuren Munkhdalai / @tsendeemts