Similarity Based Auxiliary Classifier for Named Entity Recognition

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 1140–1149, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 1140 Similarity Based Auxiliary Classifier for Named Entity Recognition Shiyuan Xiao1,2, Yuanxin Ouyang1,2, Wenge Rong1,2, Jianxin Yang1,2, Zhang Xiong1,2 1Engineering Research Center of Advanced Computer Application Technology, Ministry of Education, Beihang University, Beijing, China 2School of Computer Science and Engineering, Beihang University, Beijing, China {xiaoshiyuan, oyyx, w.rong, yjx17, xiongz}@buaa.edu.cn Abstract The segmentation problem is one of the fun- damental challenges associated with name en- tity recognition (NER) tasks that aim to reduce the boundary error when detecting a sequence of entity words. A considerable number of advanced approaches have been proposed and most of them exhibit performance deteriora- tion when entities become longer. Inspired by previous work in which a multi-task strat- egy is used to solve segmentation problems, we design a similarity based auxiliary classi- fier (SAC), which can distinguish entity words from non-entity words. Unlike conventional classifiers, SAC uses vectors to indicate tags. Therefore, SAC can calculate the similarities between words and tags, and then compute a weighted sum of the tag vectors, which can be considered a useful feature for NER tasks. Empirical results are used to verify the ratio- nality of the SAC structure and demonstrate the SAC model’s potential in performance im- provement against our baseline approaches. 1 Introduction Named entity recognition (NER) focuses on ex- tracting a specific sequence and then classifying the sequence as a predefined category. As a funda- mental and important task in natural language un- derstanding and information extraction, NER can be considered a sequence labelling task, which includes part-of-speech (POS) tagging, chunk- ing, and semantic role labelling (SRL) (Collobert et al., 2011). Owing to the growing popularity of deep learning techniques, a considerable num- ber of neural network-based approaches and tradi- tional conditional random fields (CRF) have been proposed and are widely used in many sequence labelling tasks to obtain magnificent results (Ye and Ling, 2018; Wu et al., 2018; Ghaddar and Langlais, 2018). 2019/3/3 example1.xml State Street Bank and Company Trust Organization Organization? Organization? Figure 1: The segmentation problem in NER. “State Street Bank and Trust Company” is an organization en- tity with six words, it would be unsatisfactory if we consider “State Street Bank” and “Trust Company” to be two entities To enhance NER’s performance, significant ef- fort has been devoted in this area, and one direc- tion is to employ useful information at the input end. This mechanism can be realized by includ- ing hand-crafted features (Collobert et al., 2011; Passos et al., 2014; Huang et al., 2015). To even- tually reduce heavy feature engineering, some re- searchers have argued for the exclusion of task- specific features and for the automatic mining of character level features and word level features (Ma and Hovy, 2016; Liu et al., 2018). Such end-to-end models can potentially improve perfor- mance with the exclusion of hand-crafted …

Content truncated to 500 words