TextCoherenceOnSocialMedia

This repository contains the project done during CS535. With social media becoming the number one platform for individuals to find a global audience, being coherent is more crucial than ever. However, the colloquialisms of social media make assessing coherence a problematic task. In this project, we look to automate the generation a dataset of coherent and incoherent social media posts. We also describe the preprocessing pipeline, a critical step given the varied nature of social media. Furthermore, we develop a sliding window convolutional neural network (CNN) architecture to capture the semantics of social media. Finally, we compare our results with existing architectures on both social media and non-social media-based corpora.