You don’t have to be an expert to understand that what people feel is as important as what they think — and more important than what they say they think. The reality is that emotions, opinions and attitudes are universal, continual and potentially beneficial for organizations with the technology and solutions to understand them.
Just this week, feelings came to the forefront at the Sentiment Analysis Symposium in New York City. The two-day event began Wednesday with a half-day of workshops on technical and business topics and continued Thursday with a full agenda of speakers.
I bet you're already wondering: "What was the sentiment around the Sentiment Analysis Symposium?" And I plan to answer that question. But let's start at the beginning.
Setting the Stage
The conference was organized by Seth Grimes, who specializes in strategic IT analysis, architecture and planning with a focus on business intelligence (BI) and text analytics. Grimes consults for the Washington, DC-based Alta Plana Corp., organizes the Sentiment Analysis Symposium and writes for several publications, including CMSWire.
Held at the Academy of Sciences at 7 World Trade Center in lower Manhattan, it attracted a good crowd.
As I explained in my Social Media Week (SMW) post recently, I’m no newbie to analytics. And just like SMW, I'm a veteran of the Sentiment Analysis Symposium. I've attended at least four before this one and spoke at the first one in 2010.
In the past year, Sentiment Analysis has gone a little more mainstream, partly thanks to online security concerns resulting from disclosures by former National Security Agency (NSA) contractor Edward Snowden. So I was especially interested in attending the symposium this year.
Sentiment Analysis 101
Jason Baldridge, associate professor of linguistics at University of Texas and co-founder of People Pattern, presented a three-hour class on Practical Sentiment Analysis. It introduced the concepts of sentiment analysis and opinion mining from unstructured text, looking at why they are useful and what tools and techniques are available.
It was designed for advanced users, developers and consultants. While I was familiar with most of what Baldridge presented, I was intrigued by what he shared about specific algorithms to run as well as information on public datasets that can be used to test the algorithms for sentiment accuracy.
What becomes apparent is that many linguists also program their own sentiment analysis programs. One thing Baldridge pointed out is that sentiment analysis algorithms need to be trained against the specific channels (such as Twitter) to be the most effective.
In My Humble Opinion
To be honest, the efficacy of automated sentiment analysis only seems to me to be somewhat better than a flip of the coin. Baldridge agreed sentiment programs should deliver better results than a panel of humans who are asked to vote on a question.