Lexicon-based sentiment analysis for Urdu language

Social media has recently become as a powerful weapon people use for online discourse, creating content, share it and network with other individuals at a phenomenal frequency. With social media and user-generated content exploding the web/blogs/social networking forums, vendors/critiques/socialist and influential individuals got enthusiastic to mine this substantial data set for obvious meaning, but they soon discovered a novel challenge: to know that someone is talking about a particular topic/service/brand or social event is far less important in comparison to know how they are feeling and conversing about it. This is known as sentiment analysis or opinion mining. Numbers divulge that people are extensively using social media, expressing their positive opinions or negative apprehensions online. As an aftermath the concept of sentiment analysis/opinion mining is broadly being acknowledged and employed by society as a whole to enhance their business/products/services or just to assess the overall prevailing environment. Acknowledged work is being done in this area converging towards exploring sentiment analysis, its definite requirement in this era, different frameworks for sentiment analysis and their comparison with other previously proposed techniques, but unfortunately Urdu language is not considered comprehensively in this context. As Urdu is one of the prevalent languages, this paper aims at creating an application for sentiment analysis of Urdu comments on various websites. Elaborated system architecture is discussed in detail with techniques employed; experimentation procedure and proven results of 66% accuracy are also deliberated. The F-measure achieved by this proposed system is 0.73. Challenges faced in sentiment analysis with respect to this neglected language are also highlighted for future considerations.