Detecting negativity in user comments using emotional maps and convolutional neural networks

In this paper we present a new approach to negativity detection in online user comments - an emotional image model. This model mimics image processing paradigm, where a comment is represented as a sentiment map retracing the sequence and proportions of various emotions in the text extract. We use 1D convolutional neural networks (CNN) to process 1D multichannel emotional maps which represent the emotional/sentiment image of a comment. The results show that our approach is capable of modeling and processing complex emotional patterns and detecting specific sentiments within the text image (negativity in our case) in a way similar to a classical CNN in object detection/image classification tasks.