End-to-end Answer Selection via Attention-Based Bi-LSTM Network

Many people ask medical questions online, finding the most suitable answer from candidate answers is an important research area in health care. The IEEE HotICN Knowledge Graph Academic Competition given a question and several candidate answers, then sort the candidate answers to get the best answer. We treated this subtask as a binary classification task, sorted the answers by calculating similarity between the question and each answer. In this work, we proposed a neural selection model trained on the training dataset. Our network architecture is based on the combination of Bi-LSTM and Attention mechanism, extended with biomedical word embeddings. Based on this fact, our model achieve state-of-the-art results on answer selection of medical community.