Automatic Acceptance Prediction for Answers in Online Healthcare Community

Predicting whether an answer of doctors would be accepted by a patient on online healthcare communities plays an important role in the development of e-Health. In this paper, we proposed a framework combining different types of features to predict the acceptance of answers. In order to extract textual features from both the questions posted by patients and answers posted by doctors, we first trained a sentence encoder on a held- out dataset, which encodes a pair of questions and answers into a co-dependent representation. and then both numerical features and textual features are combined to predict the acceptance of answers. The experimental results on our dataset demonstrates that our framework is able to extract additional features from text and make a better prediction.