2016/10/26 by A. K. M. Sabbir, Sabbir, A. K. M., Antonio Jimeno Yepes +3 · 1 citation
Biochemistry, Genetics and Molecular Biology · Computer Science · #Biomedical Text Mining and Ontologies #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Topic Modeling
paper · pdf · doi:10.48550/arxiv.1610.08557
openalex publication_date 2016/10/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Biomedical word sense disambiguation (WSD) is an important intermediate task\nin many natural language processing applications such as named entity\nrecognition, syntactic parsing, and relation extraction. In this paper, we\nemploy knowledge-based approaches that also exploit recent advances in neural\nword/concept embeddings to improve over the state-of-the-art in biomedical WSD\nusing the MSH WSD dataset as the test set. Our methods involve weak supervision\n- we do not use any hand-labeled examples for WSD to build our prediction\nmodels; however, we employ an existing well known named entity recognition and\nconcept mapping program, MetaMap, to obtain our concept vectors. Over the MSH\nWSD dataset, our linear time (in terms of numbers of senses and words in the\ntest instance) method achieves an accuracy of 92.24% which is an absolute 3%\nimprovement over the best known results obtained via unsupervised or\nknowledge-based means. A more expensive approach that we developed relies on a\nnearest neighbor framework and achieves an accuracy of 94.34%. Employing dense\nvector representations learned from unlabeled free text has been shown to\nbenefit many language processing tasks recently and our efforts show that\nbiomedical WSD is no exception to this trend. For a complex and rapidly\nevolving domain such as biomedicine, building labeled datasets for larger sets\nof ambiguous terms may be impractical. Here, we show that weak supervision that\nleverages recent advances in representation learning can rival supervised\napproaches in biomedical WSD. However, external knowledge bases (here sense\ninventories) play a key role in the improvements achieved.\n