Challenges in developing opinion mining tools for social media

Diana Maynard,  Kalina Bontcheva,  Dominic Rout
University of Sheffield


Abstract

While much work has recently focused on the analysis of social media in order to get a feel for what people think about current topics of interest, there are, however, still many challenges to be faced. Text mining systems originally designed for more regular kinds of texts such as news articles may need to be adapted to deal with facebook posts, tweets etc. In this paper, we discuss a variety of issues related to opinion mining from social media, and the challenges they impose on a Natural Language Processing (NLP) system, along with two example applications we have developed in very different domains. In contrast with the majority of opinion mining work which uses machine learning techniques, we have developed a modular rule-based approach which performs shallow linguistic analysis and builds on a number of linguistic subcomponents to generate the final opinion polarity and score.