Various BERT-CRF and BiLSTM-CRF based model for NER

TO-DO

Datset

- conll2003
- atis

Neural NER

- CharLSTM+WordLSTM+CRF: Lample .etc, NAACL16
- - Make a CoNLL-2003 batcher
- - Implement trainer
- - Implement WordLSTM + softmax
- - Implement CharLSTM + WordLSTM + softmax
- - Implement CharLSTM + WordLSTM + CRF
- Tranformer encoder + CRF
- BERT encoder + CRF
- pytorch JIT compilable Viterbi Decoder https://github.com/atulkum/sequence_prediction/blob/master/NER_BERT/decoder.py#L9

Slot Filling + intent prediciton

- Attention-Based Recurrent Neural Network Models for Joint Intent Detection and Slot Filling
- - Make a ATIS batcher
- - Implement trainer
- - Implement slot filler
- - Implement intent

Tree VAE

- STRUCTVAE: Tree-structured Latent Variable Models for Semi-supervised Semantic Parsing

Requiremet (python 3)

conda install pytorch  -c pytorch

CoNLL-2003 can be downloaded from https://www.clips.uantwerpen.be/conll2003/ner/

ATIS dataset can be downloaded from split 0 split 1 split 2 split 3 split 4