Ontology-Based Model for Mining User’s Emotions on the Wisdom Web

Abstract

The task of automatically detecting emotion on a web is challenging. This is due to the fact that a traditional web cannot directly interpret the meaning of semantic concepts or assess users emotions. We describe an ontology-based mining model for representation and integration of affect-related knowledge and apply it to detect user's emotions. This application is a typical use case of the broad-based Wisdom Web of Things (W2T) methodology. The model (named BIO-EMOTION) acts as an integrated framework for: (1) representation and interpretation of affect-related knowledge, including user profile, bio-signal data, situation and environment factors, and (2) supporting intelligent reasoning on users' emotions. To evaluate the effectiveness of the mining model, we conduct an experiment on a public dataset DEAP and capture a semantic knowledge base expressing both known and deduced knowledge. Evaluation shows that the model not only reaches higher accuracy than other emotion detection results from the same dataset but also achieves a comprehensive affect-related knowledge base which could represent things from both social world, physical world and cyber world in semantics. The ultimate goal of present research is to provide active, transparent, safe and reliable services to web users through their inner emotion. The model implements crucial sub-processes of W2T data cycle: from Things (acquisition of things in the hyper world) to Wisdom (performing intelligent reasoning on web users' emotion). A long-term goal is to achieve the whole W2T data cycle and to realize a holistic intelligent mining model used on the Wisdom Web.