Improving Dialogue State Tracking by Joint Slot Modeling

TL;DR

This paper introduces joint slot modeling techniques, TripPy-MRF and TripPy-LSTM, to improve dialogue state tracking by reducing slot type confusion, achieving state-of-the-art results on MultiWoZ 2.1 dataset.

Contribution

It proposes novel joint modeling methods for dialogue state tracking that address slot type confusion, advancing the state-of-the-art performance.

Findings

TripPy-MRF and TripPy-LSTM alleviate slot type confusion.

They achieve a new state-of-the-art accuracy of 61.3 on MultiWoZ 2.1.

Models outperform previous approaches in dialogue state tracking.

Abstract

Dialogue state tracking models play an important role in a task-oriented dialogue system. However, most of them model the slot types conditionally independently given the input. We discover that it may cause the model to be confused by slot types that share the same data type. To mitigate this issue, we propose TripPy-MRF and TripPy-LSTM that models the slots jointly. Our results show that they are able to alleviate the confusion mentioned above, and they push the state-of-the-art on dataset MultiWoZ 2.1 from 58.7 to 61.3. Our implementation is available at https://github.com/CTinRay/Trippy-Joint.