JointBERT

Paper: BERT for Joint Intent Classification and Slot Filling

JointBERT Architecture

JointBERT with POS tagging Architecture

Dataset

	Train	Dev	Test	Intent Labels	Slot Labels
ATIS	4,478	500	893	21	120
Snips	13,084	700	700	7	72

Training & Evaluation

$ python3 main.py --task {task_name} \
                  --epoch {epoch} \
                  --batch {batch_size} \

# For ATIS
$ python3 main.py --task atis \
                  --epoch 30 \
                  --batch 128 \
                  
# For Snips
$ python3 main.py --task snips \
                  --epoch 30 \
                  --batch 128 \
                  
# For JointBERT with POS tagging
$ python3 main_POS.py --task {task_name} \
                      --epoch {epoch} \
                      --batch {batch_size} \

Results

• 30 epoch
• 128 batch size

	Snips			ATIS
	Intent acc (%)	Slot F1 (%)	Sentence acc (%)	Intent acc (%)	Slot F1 (%)	Sentence acc (%)
JointBERT	98.4	95.7	90.1	97.5	94.4	85.1
JointBERT + POS	98.3	94.0	86.7	97.1	95.2	86.3

References

• Huggingface Transformers
• monologg / JointBERT