Incomplete Utterance Rewriting as Semantic Segmentation
This addresses the problem of improving efficiency and accuracy in dialogue systems for users, though it is incremental as it builds on existing rewriting methods.
The paper tackles incomplete utterance rewriting by reformulating it as a semantic segmentation task, predicting a word-level edit matrix instead of generating from scratch, achieving state-of-the-art performance on public datasets and being four times faster in inference.
Recent years the task of incomplete utterance rewriting has raised a large attention. Previous works usually shape it as a machine translation task and employ sequence to sequence based architecture with copy mechanism. In this paper, we present a novel and extensive approach, which formulates it as a semantic segmentation task. Instead of generating from scratch, such a formulation introduces edit operations and shapes the problem as prediction of a word-level edit matrix. Benefiting from being able to capture both local and global information, our approach achieves state-of-the-art performance on several public datasets. Furthermore, our approach is four times faster than the standard approach in inference.