Divide, Conquer, and Combine: Mixture of Semantic-Independent Experts for Zero-Shot Dialogue State Tracking

TL;DR

This paper introduces a novel zero-shot dialogue state tracking method that disentangles sample semantics and uses a mixture-of-experts approach, significantly improving performance without external data.

Contribution

It proposes a divide, conquer, and combine strategy that explicitly separates sample semantics and employs a mixture-of-experts mechanism for better zero-shot DST.

Findings

Achieves state-of-the-art zero-shot DST performance on MultiWOZ2.1

Uses only 10M trainable parameters

Significantly improves robustness and generalization

Abstract

Zero-shot transfer learning for Dialogue State Tracking (DST) helps to handle a variety of task-oriented dialogue domains without the cost of collecting in-domain data. Existing works mainly study common data- or model-level augmentation methods to enhance the generalization but fail to effectively decouple the semantics of samples, limiting the zero-shot performance of DST. In this paper, we present a simple and effective "divide, conquer and combine" solution, which explicitly disentangles the semantics of seen data, and leverages the performance and robustness with the mixture-of-experts mechanism. Specifically, we divide the seen data into semantically independent subsets and train corresponding experts, the newly unseen samples are mapped and inferred with mixture-of-experts with our designed ensemble inference. Extensive experiments on MultiWOZ2.1 upon the T5-Adapter show our schema significantly and consistently improves the zero-shot performance, achieving the SOTA on settings without external knowledge, with only 10M trainable parameters1.