Figurative Health-mention Classification from Social Media using Graph Convolutional Networks

The recent advancements in information and communication technologies paved the way for the widespread availability and use of social media networks. People use social media networks like Twitter and Facebook to share their views, opinions, and criticisms on any topic, from gaming to space science. Recently, the amount of health-related content shared by social media users witnessed an exponential increase. Users share information about many health contents, such as treatments, medications, and adverse drug reactions. Identification of such health mentions is crucial for devising health surveillance measures, but there are challenges associated with the same. We must differentiate personal health references from other metaphoric uses of those terms, such as figurative mentions where words are used in other contextual situations. This work proposes a figurative health mention classification approach using a graph convolutional neural network to classify health mentions better. When compared with the chosen baselines, our proposed method showcased better results in terms of precision, recall, and f-measure.