Handling Cross- and Out-of-Domain Samples in Thai Word Segmentation

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 1003–1016 August 1–6, 2021. ©2021 Association for Computational Linguistics 1003 Handling Cross- and Out-of-Domain Samples in Thai Word Segmentation Peerat Limkonchotiwat School of Information Science and Technology, VISTEC, Thailand Peerat.l s19@vistec.ac.th Wannaphong Phatthiyaphaibun Faculty of Interdisciplinary Studies, Khon Kaen University, Thailand wannaphong@kkumail.com Raheem Sarwar RGCL, University of Wolverhampton, United Kingdom R.Sarwar4@wlv.ac.uk Ekapol Chuangsuwanich Department of Computer Engineering, Chulalongkorn University, Thailand ekapolc@cp.eng.chula.ac.th Sarana Nutanong School of Information Science and Technology, VISTEC, Thailand snutanon@vistec.ac.th Abstract While word segmentation is a solved prob- lem in many languages, it is still a chal- lenge in continuous-script or low-resource lan- guages. Like other NLP tasks, word segmen- tation is domain-dependent, which can be a challenge in low-resource languages like Thai and Urdu since there can be domains with in- sufficient data. This investigation proposes a new solution to adapt an existing domain- generic model to a target domain, as well as a data augmentation technique to combat the low-resource problems. In addition to domain adaptation, we also propose a framework to handle out-of-domain inputs using an ensem- ble of domain-specific models called Multi- Domain Ensemble (MDE). To assess the ef- fectiveness of the proposed solutions, we con- ducted extensive experiments on domain adap- tation and out-of-domain scenarios. More- over, we also proposed a multiple task dataset for Thai text processing, including word seg- mentation. For domain adaptation, we com- pared our solution to the state-of-the-art Thai word segmentation (TWS) method and ob- tained improvements from 93.47% to 98.48% at the character level and 84.03% to 96.75% at the word level. For out-of-domain scenarios, our MDE method significantly outperformed the state-of-the-art TWS and multi-criteria methods. Furthermore, to demonstrate our method’s generalizability, we also applied our MDE framework to other languages, namely Chinese, Japanese, and Urdu, and obtained im- provements similar to Thai’s. 1 Introduction Word segmentation (WS) is a crucial upstream pro- cess for most natural language processing (NLP) tasks such as named entity recognition (NER), ma- chine translation (MT), and part-of-speech tagging (POS). Nguyen et al. (2017) showed POS perfor- mance increased from 87% to 93% when the WS was improved. WS can also enhance the perfor- mance of MT, such as the work done by Chang et al. (2008) for Chinese-English MT. While word segmentation is considered a solved problem in many languages, the task is still a chal- lenge in continuous-script languages. A great num- ber of writing systems have no word boundary, e.g., Thai, Chinese, and Japanese. Deep learning has been effective in performing WS in these lan- guages. However, it requires a large amount of training data to construct a reliable model, which can be a limitation for low-resource languages like Thai and Urdu. Furthermore, like other NLP tasks, word segmentation is domain-dependent (Fu et al., 2020). To handle a variety of data domains, there should be a substantial amount of data for each of them, exacerbating the low-resource problem. To make the matter worse, we may also need to handle …

Content truncated to 500 words