Abstract

Yasser El Sonbaty
Finding Opinion Strength Using Rule-Based Parsing for Arabic Sentiment Analysis
With increasing interest in sentiment analysis research and pinionated web content always on the rise, focus on analysis of text in various domains and different languages is a relevant and important task. This paper explores the problems of sentiment analysis and opinion strength measurement using a rule-based approach tailored to the Arabic language. The approach takes into account language-specific traits that are valuable to syntactically segment a text, and allow for closer analysis of opinion-bearing language queues. By using an adapted sentiment lexicon along with sets of opinion indicators, a rule-based methodology for opinion-phrase extraction is introduced, followed by a method to rate the parsed opinions and offer a measure of opinion strength for the text under analysis. The proposed method, even with a small set of rules, shows potential for a simple and scalable opinion-rating system, which is of particular interest for morphologically-rich languages such as Arabic.