Multi Task Learning for Intent Classification and Named Entity Recognition in HuggingFace

In this repo, we explore the approach to create a multi-task-learning system to do NER and Intent classification jointly.

Training

To train, use the following command:

python train.py --task ner

Where task can have three values: ner, rel and mtl

• rel -> relation only model
• ner -> slot tagging only model
• mtl -> Multi task learning

Testing

To test, use the following comman:

python test.py --file_path

• Default file path is 'data/hw1_test.xlsx'