Edinburgh Napier University

  We propose a deep-learning-based framework for multimodal sentiment analysis and emotion recognition. In particular, we leverage on the power of convolutional neural networks to obtain a performance improvement of 10% over the state of the art by combining visual, text and audio features. We also discuss some major issues frequently ignored in multimodal sentiment analysis research, e.g., role of speaker-independent models, importance of different modalities, and generalizability. The framework illustrates the different facets of analysis to be considered while performing multimodal sentiment analysis and, hence, serves as a new benchmark for future research in this emerging field.