Joint Chinese Word Segmentation and Part-of-speech Tagging via Two-stage Span Labeling

Joint Chinese Word Segmentation and Part-of-speech Tagging via Two-stage Span Labeling Duc-Vu Nguyen1,3, Linh-Bao Vo2,3, Ngoc-Linh Tran2,3 Kiet Van Nguyen2,3, Ngan Luu-Thuy Nguyen2,3 1Multimedia Communications Laboratory, University of Information Technology, Ho Chi Minh City, Vietnam 2University of Information Technology, Ho Chi Minh City, Vietnam 3Vietnam National University, Ho Chi Minh City, Vietnam vund@uit.edu.vn, {18520503,20521538}@gm.uit.edu.vn {kietnv,ngannlt}@uit.edu.vn Abstract Chinese word segmentation and part-of- speech tagging are necessary tasks in terms of computational linguistics and application of natural language processing. Many re- searchers still debate the demand for Chinese word segmentation and part-of-speech tagging in the deep learning era. Nevertheless, resolv- ing ambiguities and detecting unknown words are challenging problems in this field. Previ- ous studies on joint Chinese word segmenta- tion and part-of-speech tagging mainly follow the character-based tagging model focusing on modeling n-gram features. Unlike previ- ous works, we propose a neural model named SPANSEGTAG for joint Chinese word segmen- tation and part-of-speech tagging following the span labeling in which the probability of each n-gram being the word and the part- of-speech tag is the main problem. We use the biaffine operation over the left and right boundary representations of consecutive char- acters to model the n-grams. Our experiments show that our BERT-based model SPANSEG- TAG achieved competitive performances on the CTB5, CTB6, and UD, or significant im- provements on CTB7 and CTB9 benchmark datasets compared with the current state-of- the-art method using BERT or ZEN encoders. 1 Introduction Chinese word segmentation (CWS) and part-of- speech (POS) tagging are necessary tasks in terms of computational linguistics and application of natural language processing (NLP). There are two primary approaches for joint CWS and POS tagging, includ- ing the two-step and one-step methods. The two- step approach is to find words and then assign POS tags to found words. Ng and Low (2004) proposed the one-step approach that combines CWS and POS tagging into a unified joint task. The one-step ap- proach was proved better than two-step approach by many prior studies (Jiang et al., 2008; Jiang et al., 2009; Sun, 2011; Zeng et al., 2013; Zheng et al., 2013; Kurita et al., 2017; Shao et al., 2017; Zhang et al., 2018). These studies proposed various methods incorporating linguistic features or contextual infor- mation into their joint model. Remarkably, Tian et al. (2020a) proposed a two-way attention mechanism incorporating both context features and correspond- ing syntactic knowledge from off-the-shelf toolkits for each input character. To our best knowledge, we observed all previous studies for joint CWS and POS tagging following the character-based tagging paradigm. The character- based tagging effectively produces the best combi- nation of word boundary and POS tag. However, this character-based tagging paradigm does not give us a clear explanation when processing overlapping ambiguous strings. From the view of experimental psychology, human perception and performance, Ma et al. (2014) concluded that multiple words consti- tuted by the characters in the perceptual span are activated when processing overlapping ambiguous strings. Besides, Tian et al. (2020b) shown that mod- eling word-hood for n-gram …

Content truncated to 500 words