Sentiment Analysis: A Market-Relevant and Reliable Measure of Public Feeling?

This paper critically examines emergent research with sentiment analysis tools to assess their current status and relevance to applied opinion and behaviour measurement. The rapid spread of online news and online chatter in blogs, micro-blogs and social media sites has created a potentially rich source of public opinion. Waves of public feeling are vented spontaneously on a wide range of issues on a minute-by-minute basis in the online world. These online discourses are continually being refreshed, and businesses and advertisers, governments and policy makers have woken up to the fact that this universe of self-perpetuating human sentiment could represent a valuable resource to guide political and business decisions. The massive size of this repository of emotional content renders manual analysis of it feasible only for tiny portions of its totality, and even then can be labour intensive. Computer scientists have however produced software tools that can apply linguistic rules to provide electronic readings of meanings and emotions. These tools are now being utilised by applied social science and market researchers to yield sentiment profiles from online discourses created within specific platforms that purport to represent reliable substitutes for more traditional, offline measures of public opinion. This paper considers what these tools have demonstrated so far and where caution in their application is still called for.