Association extraction from biomedical literature based on representation and transfer learning.
Journal:
Journal of theoretical biology
Published Date:
Dec 25, 2019
Abstract
Extracting biological relations from biomedical literature can deliver personalized treatment to individual patients based on their genomic profiles. In this paper, we present a novel sentence-level attention-based deep neural network to predict the semantic relationship between medical entities. We utilize a transfer learning based paradigm which considerably improves the prediction performance. The main distinction of the proposed approach is that it relies solely on sentence information, putting aside handcrafted biomedical features. Sentence information is transformed into embedding vectors and improved by the pre-trained embedding models trained on PubMed and PMC papers. Extensive evaluations show that the proposed approach achieves a competitive performance in comparison with the state-of-the-art methods, while do not require any domain-specific biomedical feature. The evaluation data and resources are available at https://github.com/EsmaeilNourani/Deep-GDAE/.