This repository contains the official code for the ACL 2022 paper "Towards Fair Evaluation of Dialogue State Tracking by Flexible Incorporation of Turn-level Performances".
Dependencies: Download MultiWOZ 2.1 dataset from https://github.com/budzianowski/multiwoz/blob/master/data/MultiWOZ_2.1.zip. Copy the data.json file to the home directory of the code repo.
-
Trade and SOM-DST: Run compute_accuracy_trade_somdst.py
-
Hi-DST: Run compute_accuracy_hi-dst.py
-
Trippy: Run compute_accuracy_trippy.py
Note: We have provided the inference result of all the four model in their respective directories. One can also generate the same by running the original codes.