DM_NLP at SemEval-2018 Task 12: A Pipeline System for Toponym Resolution

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 917–923 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 917 DM NLP at SemEval-2019 Task 12: A Pipeline System for Toponym Resolution Xiaobin Wang, Chunping Ma, Huafei Zheng, Chu Liu, Pengjun Xie, Linlin Li, Luo Si Alibaba Group, China {xuanjie.wxb, chunping.mcp, huafei.zhf, chuci.lc, chengchen.xpj, linyan.lll, luo.si}@alibaba-inc.com Abstract This paper describes DM-NLP’s system for to- ponym resolution task at Semeval 2019. Our system was developed for toponym detec- tion, disambiguation and end-to-end resolu- tion which is a pipeline of the former two. For toponym detection, we utilized the state- of-the-art sequence labeling model, namely, BiLSTM-CRF model as backbone. A lot of strategies are adopted for further improve- ment, such as pre-training, model ensemble, model averaging and data augment. For to- ponym disambiguation, we adopted the widely used searching and ranking framework. For ranking, we proposed several effective features for measuring the consistency between the de- tected toponym and toponyms in GeoNames. Eventually, our system achieved the best per- formance among all the submitted results in each sub task. 1 Introduction The toponym resolution task is aimed to detect to- ponyms in scientific papers and link them to en- tities in a geographical knowledge base (GeoN- ames1 in this task). A toponym is a proper name of a place or geographical entity that is named, and can be designated by a geographical coordinate, including cities, countries, lakes or monuments. We developed an end-to-end toponym resolu- tion system (for subtask 3) which is a pipeline of toponym detection (for subtask 1) and disam- biguation (for subtask 2). We model the detection task as a Named Entity Recognition (NER) and address it with popular sequence labeling frame- work. For disambiguation task, we adopted the searching and ranking framework which is widely used in Entity linking task. Toponym is a special type of entity similar to the location entity in the general NER task. Thus, 1https://geonames.org the well-studied NER models may be effective for detecting toponyms. The most successful NER models (Chen et al., 2006; Lample et al., 2016; Huang et al., 2015; Yao and Huang, 2016) are se- quence labeling models, including the traditional CRF (Conditional Random Field (Lafferty et al., 2001)) and some variants of RNNs (Recurent Neu- ral Networks) proposed recently, like LSTM-CRF, BiLSTM-CRF, BiLSTM-CNN-CRF, etc. In this paper, We utilize the most popular model, i.e., BiLSTM-CRF for toponym detection. Beyond the model, a prevalent pre-training embedding named ELMo is used after fine-tuning. Model averaging and model ensemble are used for avoiding over- fitting. Data sets from other NER tasks are ex- ploited to augment the training data. We also pro- posed a dictionary based method for detecting to- ponyms in tables separately. Since tables have some peculiarities, i.e., well formatted yet without meaningful context for toponyms in them. Toponym disambiguation can be seen as a vari- ant of entity linking (EL) problem, which links entity mentions in articles to entities in knowl- edge base (KB) like Wikipedia. …

Content truncated to 500 words