Using Recursive Neural Networks to Detect and Classify Drug-Drug Interactions from Biomedical Texts

摘要

The purpose of this paper is to explore in detail how a Recursive Neural Network can be applied to classify drug-drug interactions from biomedical texts. The system is based on MV-RNN, a Matrix-Vector Recursive Neural Network, built from the Stanford constituency trees of sentences. Drug-drug interactions are usually described by long sentences with complex structures (such as subordinate clauses, oppositions, and coordinate structures, among others). Our experiments show a low performance that may be probably due to the parser not being able to capture the structural complexity of sentences in the biomedical domain.