Multi-label Disease Classification of Patient's Tweets using Automatically Acquired Training Corpus

Morbidity determination using tweet data is currently drawing huge attention within the field of influenza outbreak prediction. This work aims to develop a multi-label disease classification system targeting real tweets. One of the difficulties associated with developing such a system is how to collect a large amount of training data. In this paper, we propose a method to acquire the training data for the multi-label disease classifier automatically from real tweets. Our experimental evaluation revealed that the proposed method can be applied to construct the multi-label disease classifier successfully. We also investigate the method of using both the automatically acquired data and the manually annotated data together to further improve the classification performance.