Sentiment analysis
The emergence of sophisticated data analysis was perhaps the biggest single development in marketing in the past fifteen years. The ability for a company to mine its own data, and enrich it with external data bought from other organisations, meant it could precisely identify and target potential buyers.
One of the leaders in business analytics software is SAS, which is now turning its attention to social media. It’s a challenge, though. Obtaining and analysing information about individual consumers – their demographics, interests, purchasing history and so on – is complex but do-able, as you are dealing with facts and numbers. But how on earth do you analyse something as nuanced as what people are saying about you in informal conversations?
Not that there’s any shortage of tools that purport to do this. Products such as Twitfeel, Twendz and Twitrratr use a variety of methods and approaches – text mining, natural language processing and sentiment analysis technology – to create a consumer’s-eye view of a company.
SAS’s Jennifer Major argues in a new article that sentiment analysis, though still in its infancy, is the way forward.
‘The simplest algorithms work by scanning keywords to categorise a statement as positive or negative based on simple binary analysis (‘love’ is good ‘hate’ is bad). However, such an approach fails to capture the subtleties that bring human language to life: irony, sarcasm, slang and other idiomatic expressions. Social media, which are by nature dynamic and based on unstructured forms of information, do not fit neatly into traditional database-driven analytics systems. You need reliable sentiment analysis capabilities that require the ability to understand many linguistic shades of grey.’
This sort of technology is going to become increasingly vital as social media’s impact on marketing grows. Major’s piece is worth reading in order to get a sense of what’s involved.
And any business that is serious about getting good data on its social media imprint could do worse than get in touch with SAS to see what it can offer.
December 18th, 2009 at 8:32 pm
You are absolutely right Rob, reliably discerning sentiment through the noise of social media is not adequately addressed by simplified algorithms. Unstructured text cleansing is gaining more and more attention – Seth Grimes article (http://www.b-eye-network.com/view/12072) examines this problem in more detail. What also radically helps is using both statistical rigor with well-defined semantics – together, as a hybrid combination. This combination of subject matter with the scientific rigor of statistical models generates superior rules for extracting the true emotions behind the words.