Extracting Temporal Information from Online Health Communities

In order to extract structured medical information and related temporal information from online health communities, an integrate method based on syntactic parsing was proposed in this paper. We treated the extraction of medical and temporal phrases as a series tagging problem and trained two conditional random filed model respectively. The temporal relation identification is considered as a classification task and several support vector machine classifiers are built in the proposed method. For the feature engineering, we extracted some high level semantic features including co-reference relationship of medical concepts and the semantic similarity among tokens. The experiment results show that the proposed method has good performance in both phrase recognition and relation classification and could helped to automatically display a patient's clinical situation in chronological order.