Discourse Based Opinion Mining on Roman Urdu Data

Abstract

The use of Roman Urdu (in the form of web and user-generated content) is a common mode of communication on social media. Content like comments, reviews, feedbacks and social networking posts have been generated in Roman Urdu in large volumes. But this area is not much worked on in terms of sentiment and opinion analysis. Roman Urdu (the scripting style for Urdu language) is one of the limited resource languages that brings forward the challenges and problems for performing Opinion Mining. Adequate opinion mining is not just about understanding the overall sentiment of a document or a single paragraph, but it is also important to be able to extract sentiments on a very granular level and relate each sentiment to the aspect it corresponds to. On the more advanced level, the analysis can go beyond only positive or negative attitude and identify complex attitude types. We, therefore, developed a model for performing discourse-based opinion mining, so we could also consider the impact that various discourse elements have on the overall sentiment of the text. Our work differs from the existing body of knowledge in that not much work has been carried out on processing of Roman Urdu data for opinion mining considering discourse elements. Since our work focuses on performing discourse-based opinion mining it can be considered as first attempt in this direction as none of the literature surveyed revealed discourse[1]based analysis of Roman Urdu text. The overall gist of this research work is to have insights of the nature of user-generated content in Roman Urdu and to build necessary resources and devise algorithms to make an advancement in Sentiment Analysis and Classification for Roman Urdu