Method Entity Extraction from Biomedical Texts

Proceedings of the 29th International Conference on Computational Linguistics, pages 2357–2362 October 12–17, 2022. 2357 Method Entity Extraction from Biomedical Texts Waqar Bin Kalim Department of Computer Science The University of Western Ontario London, Ontario, Canada waqaar.199@gmail.com Robert E. Mercer Department of Computer Science The University of Western Ontario London, Ontario, Canada mercer@csd.uwo.ca Abstract In the field of Natural Language Processing (NLP), extracting method entities from biomed- ical text has been a challenging task. Sci- entific research papers commonly consist of complex keywords and domain-specific termi- nologies, and new terminologies are contin- uously appearing. In this research, we find method terminologies in biomedical text us- ing both rule-based and machine learning tech- niques. We first use linguistic features to ex- tract method sentence candidates from a large corpus of biomedical text. Then, we construct a silver standard biomedical corpus composed of these sentences. With a rule-based method that makes use of the Stanza dependency pars- ing module, we label the method entities in these sentences. Using this silver standard cor- pus we train two machine learning algorithms to automatically extract method entities from biomedical text. Our results show that it is pos- sible to develop machine learning models that can automatically extract method entities to a reasonable accuracy without the need for a gold standard dataset. 1 Introduction Method Entity Extraction from unstructured biomedical text has been an important, yet chal- lenging and somewhat under-examined, Natural Language Processing (NLP) task. Especially in the field of biomedicine, the automatic extraction of methodology names and terminologies is impera- tive. With thousands of research papers published each week, the biomedical research community is constantly creating new terminologies. As a result, it becomes difficult for researchers to find relevant information. Wang et al. (2022) give a good discussion of method entities and provide the following defini- tion: “named entities that represent specific meth- ods”. More specifically, “method entities in the academic literature are nouns or noun phrases rep- resenting the specific ways, means, and channels used to solve tasks or problems proposed by the au- thors, including sub-categories such as discipline- specific methods, software, models, algorithms, and metrics”. Biomedical method entities are a type of biomed- ical named entities. Biomedical named entity recognition (NER) is defined as the task of recog- nizing and categorizing entity names in the biomed- ical domain (Lee et al., 2004). As mentioned by Song et al. (2018), biomedical NER faces diffi- culties for various reasons. The first one is the increasing rate of newly created terminologies and keywords requiring new rules and patterns to be manually added to the rule-based methods, which can be a tedious and time-consuming task. Regard- ing biomedical method entity recognition, because it has been little studied (e.g., (Houngbo and Mer- cer, 2012)), a comprehensive definition of what constitutes a biomedical method entity beyond that given by Wang et al. (2022) still needs to be de- termined. Secondly, with information extraction tasks, the same words can have different meanings and significance in terms of the …

Content truncated to 500 words