DANGNT-SGU at SemEval-2022 Task 11: Using Pre-trained Language Model for Complex Named Entity Recognition

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 1483 - 1487 July 14-15, 2022 ©2022 Association for Computational Linguistics DANGNT-SGU at SemEval-2022 Task 11: Using Pre-trained Lan- guage Model for Complex Named Entity Recognition Dang Tuan Nguyen, Huy Khac Nguyen Huynh Saigon University, Ho Chi Minh City, Vietnam {dangnt, hnkhuy}@sgu.edu.vn Abstract This paper describes a system that we built to participate in the SemEval 2022 Task 11: MultiCoNER Multilingual Complex Named Entity Recognition, specifically the track Mono-lingual in English. To construct this system, we used Pre-trained Language Models (PLMs). Especially, the Pre-trained Model base on BERT is applied for the task of recognizing named entities by fine-tuning method. We performed the evaluation on two test datasets of the shared task: the Practice Phase and the Evaluation Phase of the competition. 1 Introduction In natural language processing, Named Entity Recognition (NER) is traditionally a funda- mental task that aims to recognize groups of words used to identify people, organizations, places, times, etc. in documents. SemEval- 2022 Task 11 focuses on detecting semanti- cally complex and ambiguous entities (Mal- masi et al., 2022b). This is a challenging NLP task that has not sufficiently received atten- tion from the research community. In prac- tice, available named entity recognition mod- els find it hard to recognize complex named entities, such as titles of works (movie/book/song/software titles) that are not simple nouns (Malmasi et al., 2022b). Type Description Complex entities Not all entities are proper noun -Noun phrases: Eternal Sun- shine of the Spotless Mind -Gerunds: Saving private Ryan -Full clauses: Mr. Smith goes to Washington… Ambigu- ous enti- ties and Contexts Not always entities -Inside out (movie), -Bonanza (game) -On the Beach (movie)… Emerging Entities New books/songs/movies re- leased daily, weekly… Table 1: SemEval-2022 Task 11 illustrates some types of complex and ambiguous entities To enable the recognition of such entities, we believe that it is necessary to make full use of Pre-trained Language Models (PLMs) to in- crease efficiency as well as reduce training time, which is also the method we applied in this study. 2 Related Work Nowadays, there are many NER works and dif- ferent approaches to solving this problem, such as LSTM (Long Short-Term Memory) (Nut Limsopatham and Nigel Collier, 2016), Trans- formers (Ashish Vaswani et al., 2017). In par- ticular, the BERT model (Jacob Devlin et al., 2019). Many relevant studies also approach the direction of combining deep learning models and adjusting model parameters to improve the effi- ciency of named entity recognition in the text (Zheng Yuan et al., 2021). However, entities with complex and ambiguous names appear con- tinuously in increasing numbers. This has posed emerging challenges for the natural language processing field. Many research show that clari- fies that named substance acknowledgement is particularly troublesome in circumstances with a low context or in scenarios where the named sub- stances are especially complex (Meng et al., 2021). The misrecognition of entities with com- plex and ambiguous names significantly affects the performance of natural language …

Content truncated to 500 words