NERToolkit

A simple NER tagger.

Running

You can run the whole process very easily. Take the conll2003 corpus for example,

Step 1: Clean the pretraind embeddings.

./scripts/conll2003.sh clean

Step 2: Convert to bieos tagging scheme.

./scripts/conll2003.sh bieos

Step 3: Prepare data.

./scripts/conll2003.sh preprocess

Step 4: Use elmo.

./scripts/conll2003.sh elmo

Step 5: Train (with elmo).

./scripts/conll2003.sh train(train_elmo)

Performance Comparision

-	Dataset	F1
Original	conll2003	91.62%
This Implementation (BiLSTM-CRF)	conll2003	91.19%