Towards Automatically Extracting Contextual Valence Shifters in Reviews of Saudi Universities

Abstract

Text sentiment orientations are affected by valence shifters. Therefore, it must be taken into account when a sentiment analysis system is developed. This paper presents an overview about existence of contextual valence shifters, such as intensifiers, negators, and connectors, in Saudi university reviews in order to seek a method for automatically extracting frequent shifting patterns. For achieving this study, a dataset containing 85,658 tweets written in Saudi dialectal Arabic language about Saudi universities was collected from the Twitter platform. It was found that many types of valence shifters exist in the dataset. As a result, for achieving accurate sentiment analysis the sentiment or polarity shifters needs to be first addressed.