YNUNLP at SemEval-2023 Task 2: The Pseudo Twin Tower Pre-training Model for Chinese Named Entity Recognition

Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 1619–1624 July 13-14, 2023 ©2023 Association for Computational Linguistics YNUNLP at SemEval-2023 Task 2:The Pseudo Twin Tower Pre-training Model for Chinese Named Entity Recognition Jing Li Yunnan University 18313855625@163.com Xiaobing Zhou* Yunnan University zhouxb@ynu.edu.cn Abstract This paper introduces our method of develop- ing a system for SemEval 2023 Task 2: Multi- CoNER II Multilingual Complex Named Entity Recognition, Track 9-Chinese. In this task, we need to identify entity boundaries and category labels for the six identified categories. The fo- cus of this task is to detect fine-grained named entities whose data set has a fine-grained taxon- omy of 36 NE classes, representing a realistic challenge for NER. We use BERT embedding to represent each character in the original sen- tence and train CRF-Rdrop to predict named entity categories using the data set provided by the organizer. Our best submission, with a macro average f1 score of 0.5657, ranked 15th out of 22 teams. 1 Introduction NER (Named Entity Recognition) is a traditional NLP task designed to identify text fragments be- longing to predefined categories in free text(Curran and Clark, 2003). NER is an important basic tool for many NLP tasks such as information extrac- tion, question answering system, syntactic analysis and machine translation. Named entities generally refer to entities with specific meaning or strong reference in the text. Academically, it usually in- cludes three categories: entity, time and number, and seven categories: personal name, place name, organization name, time, date, currency and per- centage(Chen et al., 2022). NER extracts the above entities from the unstructured input text and can identify more categories of entities based on busi- ness requirements. Chinese named entity recognition(Gui et al., 2019) plays an important role in the field of natural language processing. Compared with English, Chi- nese NER is more challenging. First of all, Chinese word boundaries are vague, and there are no delim- iters, such as Spaces, to clarify word boundaries. If Chinese NER adopts character-level model, there will be semantic loss and boundary information loss. On the other hand, if we use the word-level model, the wrong word segmentation can also de- grade performance. There are also more complex properties in Chinese, such as complex combina- tions, entity nesting, indefinite length, and neolo- gisms on the Internet. In addition, Chinese is not case-sensitive and root-affix, and lacks the expres- sion of a lot of semantic information. Named en- tity recognition is very important because in many applications we must extract the entities in our vo- cabulary(Wang et al., 2022). SemEval 2023 Task 2: MultiCoNER II Multilin- gual Complex Named Entity Recognition(Fetahu et al., 2023b). This task aims to solve the prob- lem of fine-grained named entity recognition. The mission has 13 tracks, of which tracks 1-12 are monolingual, including English, Spanish, Hindi, Bangla, Chinese, Swedish, Farsi, French, Italian, Portuguese, Ukrainian and German(Malmasi et al., 2022b). Participants trained a model that only worked in one language. Track 13 is a multilingual track, …

Content truncated to 500 words