Toward Fast and Accurate Neural Chinese Word Segmentation with Multi-Criteria Learning

TL;DR

This paper introduces a domain adaptive BERT-based Chinese Word Segmentation model that effectively captures diverse annotation criteria, leveraging shared and domain-specific knowledge, and achieves state-of-the-art results efficiently.

Contribution

It proposes a novel domain adaptive segmenter with private and shared layers, enhancing multi-criteria CWS performance and efficiency using distillation, quantization, and compiler optimization.

Findings

Outperforms previous SOTA on 10 CWS datasets

Achieves higher accuracy with improved computational efficiency

Effectively captures diverse annotation criteria

Abstract

The ambiguous annotation criteria lead to divergence of Chinese Word Segmentation (CWS) datasets in various granularities. Multi-criteria Chinese word segmentation aims to capture various annotation criteria among datasets and leverage their common underlying knowledge. In this paper, we propose a domain adaptive segmenter to exploit diverse criteria of various datasets. Our model is based on Bidirectional Encoder Representations from Transformers (BERT), which is responsible for introducing open-domain knowledge. Private and shared projection layers are proposed to capture domain-specific knowledge and common knowledge, respectively. We also optimize computational efficiency via distillation, quantization, and compiler optimization. Experiments show that our segmenter outperforms the previous state of the art (SOTA) models on 10 CWS datasets with superior efficiency.