Why I don’t like the term “sentiment analysis”
Sentiment analysis has become a hot topic in the fields of natural language processing and affective computing. Sentiment analysis aims to obtain a user’s attitude or feeling about some topic using digital information, like product reviews, ratings, language, emotions, and social networks.
While I like a lot of the research and the promises that sentiment analysis offers, I think the name is an unfortunate example of how the pressures of academics and marketing cause people to give new names to old ideas, in the process obscuring their intellectual roots. Instead “sentiment”, I believe it should be called “utility” to connect it with earlier ideas in economics and AI. According to wikipedia (http://en.wikipedia.org/wiki/Sentiment_analysis) the “early work” in sentiment analysis goes back to 2002. However, the idea of utility goes back to the late 1700’s and 1800’s under the pioneering work of the Bernoulli brothers, Jeremy Bentham and John Stuart Mill, and has been treated by major researchers such as Pareto, Edgeworth, and von Neuman.
To me, it seems that sentiment analysis, especially for product reviews, is measuring classical utility. The novelty is that it uses digital information that is often unstructured, like raw text. So if we could start fresh with a name that I feel is appropriate, I would call it “text-based utility measurement”.
Originally published on abewrites.
clustrmaps.html