Building Sophisticated Deep Learning Framework for Text Sentiment Analysis 

Social listening is the process of mining insights from online conversations about the market and leveraging them to discover business opportunities or create effective content for certain audiences. Sentiment analysis on social media conversations is crucial to understand customer opinions on products, services, brands or industries. The main objectives of this project are to build (1) the sentiment labelling (SL) system to create high quality labelled text data sets, (2) develop the deep learning (DL) framework for sophisticated sentiment analysis on texts, (3) integrate them into our social listening tool, called SocialMind, and (4) evaluate the system performance of the extended SocialMind. For the DL framework, we mainly design a DL model with attention mechanism to learn post-specific text embeddings for exploiting the explicit influence of posts/comments/replies and data sources, and a neural collaborative filtering model based on multilayer perceptron to capture the implicit influences implied in the interactions between users and data sources. This project extends SocialMind to provide much more accurate insights on the market for companies based on textual conversations in an automatic manner. The key benefits of the extended SocialMind are to have more effective and efficient management and discover more valuable business opportunities.