Affectional Ontology and Multimedia Dataset for Sentiment Analysis

摘要

Ontology is able to understand the association between concepts and the relationships within contents. We argue that, ontology could compete with machine learning in detecting sentiments contained in textual messages. Current ontology-based sentiment models are domain specific, which limits their ability to adapt to different domains. In this work, we propose a general sentiment ontology (Affectional Ontology) using various sentiment lexicons and psychological-based resources. To provide an efficient evaluation on the Affectional Ontology, we propose a domain-free sentiment multimedia dataset (DFSMD). Our DFSMD was constructed with high standard annotation criteria. The results of our work show the effectiveness of the proposed ontology in capturing the sentiment, when compared to the machine learning approach. The proposed DFSMD is publicly available and can be used in various sentiment analysis problems without the restrictions of particular domains or aspects.