Fast and Accurate Neural Word Segmentation for Chinese
This work addresses the problem of slow training and inference in neural Chinese word segmentation, which is incremental as it builds on existing neural approaches.
The paper tackles the computational inefficiency of neural models for Chinese word segmentation by introducing a greedy neural segmenter with balanced embeddings, achieving faster and more accurate results than state-of-the-art models on benchmark datasets.
Neural models with minimal feature engineering have achieved competitive performance against traditional methods for the task of Chinese word segmentation. However, both training and working procedures of the current neural models are computationally inefficient. This paper presents a greedy neural word segmenter with balanced word and character embedding inputs to alleviate the existing drawbacks. Our segmenter is truly end-to-end, capable of performing segmentation much faster and even more accurate than state-of-the-art neural models on Chinese benchmark datasets.