A Multi-Task Approach for Improving Biomedical Named Entity Recognition by Incorporating Multi-Granularity information

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 4804–4813 August 1–6, 2021. ©2021 Association for Computational Linguistics 4804 A Multi-Task Approach for Improving Biomedical Named Entity Recognition by Incorporating Multi-Granularity Information Yiqi Tong, Yidong Chen*, Xiaodong Shi School of Informatics, Xiamen University, Xiamen, China yqtong@stu.xmu.edu.cn {ydchen, mandel}@xmu.edu.cn Abstract Neural biomedical named entity recognition (BioNER) methods usually require a large amount of annotated data, while the anno- tated BioNER datasets are often difficult to obtain and small in scale due to the limita- tions of privacy, ethics and high degree of specialization. To alleviate the lack of train- ing samples, unlike conventional methods that only use token-level information, this paper proposes a method that simultaneously utilize the latent multi-granularity information in the dataset. Concretely, the proposed model is based on a multi-task approach, which lever- ages different training objectives by introduc- ing auxiliary tasks, i.e. binary classification, multi-class and multi-token classification. Ex- perimental results over three BioNER datasets show that the proposed model produces better performance over the BioBERT baseline and can get more than 3% improvements of F1- score in low-resource scenarios. Finally, we released our code at https://github.com/ zgzjdx/MT-BioNER. 1 Introduction Biomedical named entity recognition (BioNER) aims to identify entity mentions such as gene/protein, disease and chemicals from unstruc- tured text. Such information is useful for down- stream natural language processing (NLP) tasks like relation extraction (Zhou et al., 2014), auto- matic abstracting (Mishra et al., 2014) and question answering (Athenikos and Han, 2010), etc. Differ- ent from those named entity recognition (NER) tasks for general domain like news, BioNER is particular challenge due to the naming complex- ity (Liu et al., 2015), large variations in same en- tity names (Jia et al., 2019; Kim et al., 2019), and new entity mentions rapidly reported in scientific *Corresponding author. Figure 1: Examples from our constructed dataset. In our work, we designed three auxiliary tasks to help im- proving the main NER task. Two of them are sentence- level tasks and the other one is a token-level task. Con- cretely, the first sentence-level task predicts whether or not a sentence contains entities; the second sentence- level task predicts how many entities a sentence con- tains; and the token-level task predicts whether or not a given token belongs to a multi-token entity. Clearly, to support training the auxiliary tasks, additional labels have been added in our data. However, please note that, the additional labels could be derived from the original NER labels and do not need additional manual anno- tations. In a word, what we have done in this paper is try to use the multi-granularity information implied in the original dataset to improve the performance of BioNER. publications (Luo et al., 2018). These various fac- tors lead to the small number and size of current BioNER datasets. In recent years, neural BioNER has become a main approach because of its out- standing performance (Lample et al., 2016; Habibi et al., 2017; Yadav and Bethard, 2019). Some re- searchers have …

Content truncated to 500 words