A Chinese Radical Embedding for Joint Segmentation and POS Tagging
Xu Han, Cong Wang, Jie Liu, Zheng Li · 2018
Word embedding has been proven to play an important role in natural language processing. In this paper, we develop a mixed embedding for Chinese texts to perform joint segmentation and POS tagging task. We regard Chinese characters as pictures, and classify them by a special low-than character level feature called radical. We evaluate our model on different datasets, CTB5 and CTB9, and achieve state-of-the-art performances, getting an F1-score of 98.21% on word segmentation task in CTB5.