Description
- MultiWOZ2.1 dataset parser for dialogue state tracking
- MultiWOZ2.1 dataset
- The dataset consists of 10438 multi-domain dialogues
- Train: 8438, Validate: 1000, Test: 1000
- 16 domains: taxi, restaurant, police, hotel, hospital, attraction, bus, train
- File: data.json, ontology.json, dialogue_acts.json, taxi_db.json etc.
How to use
python main.py --load_data_path . --save_data_path .
Datasets
- Details of MultiWOZ
https://arxiv.org/pdf/1810.00278.pdf
https://arxiv.org/pdf/1907.01669.pdf
- MultiWOZ download
http://dialogue.mi.eng.cam.ac.uk/index.php/corpus/