Chunking Shared Task in Conll2000

Dataset: https://www.clips.uantwerpen.be/conll2000/chunking/
connlleval.py: https://github.com/sighsmile/conlleval/blob/master/conlleval.py

1. Download and unzip dataset.
2. Set the dataset path in preprocessing.py.
3. Set the hyperparameter in train.py.
4. Change other setting you want to change.
5. Do python train.py if you want to train with BERT, do python bert_train.py.
6. Set model path and do python pred.py or python bert_train.py.
7. Do python connlleval.py.
8. You can monitor or confirm　loss curve by tensorboard.

trained models: https://drive.google.com/drive/folders/11uyVskbp9oLQVsj7A5lfIsKhhPJOoFKr?usp=sharing

current score  
LSTM-crf: 92.6,  
LSTM-w2v-crf: 92.49,  
BERT-crf: 96.38: current output.txt,