Exploiting sequential patterns to detect objective aspects from online reviews

Aspect or feature extraction, from online customer reviews, is the key task of aspect-based sentiment analysis. This task aims to extract fine-grained aspects from online customer reviews, which is a challenging task. Most of the early work focused on the extraction of subjective aspects associated with some opinion words but very few have focused on objective aspects. These objective aspects can affect the overall performance of the aspect extraction task resulting in wrong results. In this paper, we proposed a sequential pattern-based approach to detect such aspects which are objective in nature. We defined several sequential patterns, on the basis of word occurrence in a sentence, to detect objective aspects. The experimental evaluation demonstrated significant improvement in the aspect extraction phase by pruning such aspects which are objective in nature.