VinAIResearch / JointIDSF

BERT-based joint intent detection and slot filling with intent-slot attention mechanism (INTERSPEECH 2021)

JointIDSF: Joint intent detection and slot filling

• We propose a joint model (namely, JointIDSF) for intent detection and slot filling, that extends the recent state-of-the-art JointBERT+CRF model with an intent-slot attention layer to explicitly incorporate intent context information into slot filling via "soft" intent label embedding.
• We also introduce the first public intent detection and slot filling dataset for Vietnamese.
• Experimental results on our Vietnamese dataset show that our proposed model significantly outperforms JointBERT+CRF.

Details of our JointIDSF model architecture, dataset construction and experimental results can be found in our following paper:

@inproceedings{JointIDSF,
    title     = {{Intent Detection and Slot Filling for Vietnamese}},
    author    = {Mai Hoang Dao and Thinh Hung Truong and Dat Quoc Nguyen},
    booktitle = {Proceedings of the 22nd Annual Conference of the International Speech Communication Association (INTERSPEECH)},
    year      = {2021}
}

Please CITE our paper whenever our dataset or model implementation is used to help produce published results or incorporated into other software.

Dataset

By downloading our dataset, USER agrees:

• to use the dataset for research or educational purposes only.
• to not distribute the dataset or part of the dataset in any original or modified form.
• and to cite our paper above whenever the dataset is employed to help produce published results.

Model installation, training and evaluation

Installation

• Python version >= 3.6
• PyTorch version >= 1.4.0

    git clone https://github.com/VinAIResearch/JointIDSF.git
    cd JointIDSF/
    pip3 install -r requirements.txt

Training and Evaluation

Run the following two bash files to reproduce results presented in our paper:

    ./run_jointIDSF_PhoBERTencoder.sh
    ./run_jointIDSF_XLM-Rencoder.sh

• Here, in these bash files, we include running scripts to train both our JointIDSF and the baseline JointBERT+CRF.
• Although we conduct experiments using our Vietnamese dataset, the running scripts in run_jointIDSF_XLM-Rencoder.sh can adapt for other languages that have gold annotated corpora available for intent detection and slot filling. Please prepare your data with the same format as in the data directory.

Inference

We also provide model checkpoints of JointBERT+CRF and JointIDSF. Please download these checkpoints if you want to make inference on a new text file without training the models from scratch.

• JointIDSF

http://public.vinai.io/JointIDSF_PhoBERTencoder.tar.gz

http://public.vinai.io/JointIDSF_XLM-Rencoder.tar.gz

• JointBERT+CRF

http://public.vinai.io/JointBERT-CRF_PhoBERTencoder.tar.gz

http://public.vinai.io/JointBERT-CRF_XLM-Rencoder.tar.gz

Example of tagging a new text file using JointIDSF model:

python3 predict.py  --input_file <path_to_input_file> \
                    --output_file <output_file_name> \
                    --model_dir JointIDSF_XLM-Rencoder

where the input file is a raw text file (one utterance per line).

Acknowledgement

Our code is based on the unofficial implementation of the JointBERT+CRF paper from https://github.com/monologg/JointBERT