I had the good fortune of being involved with the 2nd Annual Sentiment Analysis Symposium in New York on April 12. This event, hosted and chaired by Seth Grimes of Alta Plana, brings together the users, creators and providers of sentiment technology to discuss trends and share real-world implementation of sentiment analysis.
As a marketing professional, I’m potentially the prospect for many of the companies that were represented at the event because the goal of sentiment analysis is to extract key trends and triggers about a company, brand, stock or message from unstructured text, and help businesses digest and understand what those triggers mean. Attending the sessions reminded me how analytic marketing tasks can be – and how trends, momentum and themes help large businesses (and some small) to measure a variety of metrics including churn, leads and prospects.
There is the automated analysis method,the semi-automated method and the manual method. And while all of those methods were explored and represented at the Symposium, I think the decision to implement sentiment varies widely from organization to organization. At least that’s what I heard from several of the end users. This group seemed very “automated” focused – but as was evident by a presentation by Katie Delahaye Paine, there seems to be several instances where humans are needed to do the work.
It was great to see how some of the companies are using sentiment, instead of just discussing how it could be used – including RavenPack who uses it to analyze news for automated trading, AOL for ranking movie reviews using Twitter data, and eBay with community and identifying site outages.
There was a lot of information shared over the three days I was there. Between a tutorial hosted by eBay on Monday, the Sentiment Symposium on Tuesday and a Lexalytics Users Group on Wednesday I know quite a bit more the future about text analytics and sentiment. In fact, Lexalytics announced their next release, Salience 5.0 has ingested all of Wikipedia to help in automated concepts and topics. Pretty cool if you have several streams of unstructured data flowing into your company for analysis.
I think one quote that stood out to me was by Armando Gonzalez, CEO Ravenpack, who stated, “People get tired but computers do not. Computers are good at following rules, people are not.” The essence of sentiment analysis is letting the machines do the heavy lifting for you, and then bringing in your experts to review and deliver the messages internally and refine the ones externally. In marketing, how effective your message is and the perception of your brand directly plays into your sales and success.