Task-oriented LLM Bot

Code accompaniment for the paper Are Large Language Models All You Need for Task-Oriented Dialogue? (Hudeček & Dušek, SIGDIAL 2023).

Installation

Use your Conda/Virtualenv and simply run:

pip install -r requirements.txt

Usage

1. You need to create the FAISS vector store DB:

python create_faiss_db.py --output_faiss_db mwoz_db.pkl --total 10 --dataset multiwoz

2. To use GPT-3.5, add your OpenAI API key into the file openai_key.txt.

3. Edit the run_interact.sh file in case you want to include your own CUDA/Python etc. setup.

4. You can now run the interactive bot with default settings like this:

./run_interact.sh

Alternatively, you can run interact.py directly and change the parameters.

• num_examples changes the number of few-shot examples in the prompt
• context_size changes the length of the dialogue context passed to the bot
• goals_trial.json is just a trial file to try out user goal messages, you can use MultiWOZ 2.1's data.json directly (this is just a small sample of that file).
• Other required files (database, ontology) are directly taken from MultiWOZ.

Cite us

If you use this code in your work, please cite this paper:

@inproceedings{hudecek_are_2023,
	address = {Prague, Czechia},
	title = {Are {Large} {Language} {Models} {All} {You} {Need} for {Task}-{Oriented} {Dialogue}?},
	url = {https://aclanthology.org/2023.sigdial-1.21},
	doi = {10.18653/v1/2023.sigdial-1.21},
	booktitle = {Proceedings of the 24th {Annual} {Meeting} of the {Special} {Interest} {Group} on {Discourse} and {Dialogue}},
	publisher = {Association for Computational Linguistics},
	author = {Hudeček, Vojtěch and Dusek, Ondrej},
	month = sep,
	year = {2023},
	pages = {216--228},
}

Acknowledgements

This work was funded by the European Union (ERC, NG-NLG, 101039303), project TL05000236 AI asistent pro žáky a učitele co-financed by the Technological Agency of the Czech Republic within the ÉTA 5 Programme, and Charles University project SVV 260 698. It used resources of the LINDAT/CLARIAH-CZ Research Infrastructure (Czech Ministry of Education, Youth, and Sports project No. LM2018101).