A preliminary investigation of deep emotion-based classification from natural language text

In the Social Web age, the role of the web user has evolved from a simple consumer of web content to the main actor who interacts with other users, shares data and cooperates in social networks, online communities, blogs, wikis, feeds, and chats. His opinions, comments, and suggestions have an amazing influence on the online communities and users that can be inadvertently influenced in decision-making activities such as buying a certain product or trusting the recommendations of the blog, etc. Big corporations, as well as scientific communities, study the user behavior trying to capture human feeling and emotions, aimed at guessing the 'client' preferences and then attend his expectations. Emotion extraction using natural language is a complex activity that needs to understand the content and capture the sentiment hidden in the written text. To this end, the work proposes a text analysis based on Deep Learning (DL) to capture the emotions that regulate human feeling in the natural language. The work shows the effectiveness of this approach presenting a comparative analysis of emotion-based text classification by DL neural networks methods, with different datasets and feature settings.