Using Sentence-level Classification Helps Entity Extraction from Material Science Literature

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 4540–4545 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 4540 Using Sentence-level Classification Helps Entity Extraction from Material Science Literature Ankan Mullick† Shubhraneel Pal† Tapas Nayak†† Seung-Cheol Lee♠ Satadeep Bhattacharjee♠ Pawan Goyal† †Department of Computer Science and Engineering, IIT Kharagpur, India. ††TCS Research, India. ♠Indo-Korea Science and Technology ankanm@kgpian.iitkgp.ac.in, shubhraneel@iitkgp.ac.in, tnk02.05@gmail.com, seungcheol.lee@ikst.res.in, s.bhattacharjee@ikst.res.in, pawang@cse.iitkgp.ac.in Abstract In the last few years, several attempts have been made on extracting information from material science research domain. Material Science research articles are a rich source of information about various entities related to material science such as names of the materials used for experiments, the computational software used along with its parameters, the method used in the experiments, etc. But the distribution of these entities is not uniform across different sections of research articles. Most of the sentences in the research articles do not contain any entity. In this work, we first use a sentence-level classifier to identify sentences containing at least one entity mention. Next, we apply the information extraction models only on the filtered sentences, to extract various entities of interest. Our experiments for named entity recognition in the material sci- ence research articles show that this additional sentence-level classification step helps to improve the F1 score by more than 4%. Keywords: Material Science Information Extraction, Material Science Entity Detection, Material Science Research 1. Introduction Every year, a large number of research articles are pub- lished in the material science domain. It is very diffi- cult to find informative entities in these articles as they spread across the articles. Recently, researchers try to solve this problem by using information extraction in the material science domain. Various models are pro- posed to extract informative entities such as the materi- als used in the research work, simulation software used and its parameters, the method used in the study, the used code languages, and finally, the outcome of the study, etc. These entities are very common in mate- rial science articles. It is critical to extract and store them in a structured way. But the task of information extraction from the entire articles becomes complex as the articles are quite long and they contain too many re- dundant sentences. Most of the text in the articles does not contain any such entities thus leading to erroneous extraction. In this paper, we address this issue of iden- tifying the text in the articles that are devoid of such entities and the text that contains the informative enti- ties. Successfully eliminating the text from the articles that do not contain any entities can help to apply the information extraction models to the entire articles. In this work, we target five types of informative en- tities from the material science domain such as ma- terial names, method names, code or simulation soft- ware names, parameters of the simulation software, and structure type of the materials. We consider a sentence as informative …

Content truncated to 500 words