Uni-Encoder: A Fast and Accurate Response Selection Paradigm for Generation-Based Dialogue Systems

TL;DR

Uni-Encoder introduces a novel response ranking paradigm for dialogue systems that combines the accuracy of Cross-Encoder with the efficiency of Poly-Encoder, achieving state-of-the-art results with faster inference.

Contribution

The paper proposes the Uni-Encoder paradigm, which encodes context once and all candidates simultaneously, improving efficiency while maintaining high ranking accuracy.

Findings

Achieves new state-of-the-art results on four benchmark datasets.

Improves R10@1 by 2.9% on Ubuntu V2 dataset.

Provides approximately 4X faster inference speed.

Abstract

Sample-and-rank is a key decoding strategy for modern generation-based dialogue systems. It helps achieve diverse and high-quality responses by selecting an answer from a small pool of generated candidates. The current state-of-the-art ranking methods mainly use an encoding paradigm called Cross-Encoder, which separately encodes each context-candidate pair and ranks the candidates according to their fitness scores. However, Cross-Encoder repeatedly encodes the same lengthy context for each candidate, resulting in high computational costs. Poly-Encoder addresses the above problems by reducing the interaction between context and candidates, but with a price of performance drop. In this work, we develop a new paradigm called Uni-Encoder, that keeps the full attention over each pair as in Cross-Encoder while only encoding the context once, as in Poly-Encoder. Uni-Encoder encodes all the candidates with the context in one forward pass. We use the same positional embedding for all candidates to ensure they are treated equally and design a new attention mechanism to avoid confusion. Our Uni-Encoder can simulate other ranking paradigms using different attention and response concatenation methods. Extensive experiments show that our proposed paradigm achieves new state-of-the-art results on four benchmark datasets with high computational efficiency. For instance, it improves R10@1 by 2.9% with an approximately 4X faster inference speed on the Ubuntu V2 dataset.