Seth Grimes, well-known authority on text analysis and sentiment, spoke with me about sentiment analysis and the upcoming Sentiment Analysis Symposium on April 11 and 12, 2011 in New York City (click link to register). I’ve read Seth for years and included some of his ideas in the research chapters of Listen First! Seth founded this event last year and readied a compelling second edition. If you’re responsible for text analytic solutions, analyzing or acting on sentiment, the conference will give you software to consider, ideas to apply, and people and companies to meet. I’m honored to be moderating the Visionaries panel on April 12.
Sentiment started with “positive, negative or neutral,” but advances in science and software keep it evolving and making it more valuable. I posed 4 questions to Seth about sentiment, the conference’s value and the near future. Here goes:
Steve Rappaport (SR): Let’s start with the fundamental question: What is sentiment analysis?
Seth Grimes (Grimes): I’ll offer a dictionary-style definition: Sentiment analysis is an effort to systematically detect and evaluate opinions, attitudes, and emotions in a spectrum of personal, online, social, and enterprise information sources. I’ll also tell you that sentiment analysis is often misunderstood. Human coding isn’t the only path to accuracy, and systems can be useful even when not fully automated. Typically a hybrid human-machine approach will produce the best results, but there are many approaches that will work, and your choice will depend on your business goals and on the type and volume of material you plan to analyze. The need for market education — to help users understand choices and help vendors understand needs — is what led me, in 2010, to create the first Sentiment Analysis Symposium, and now we have “round 2″ set for April 12.
SR: What key concerns drove program development this year compared to other years?
Grimes: My only preconceived notion was that users and vendors need a place to meet and share knowledge about interesting and useful sentiment solutions. So the program features major applications — social monitoring & measurement, customer service & support, review mining, sentiment-driven trading strategies — and also allows podium time for new and innovative research and applications. I actually did very little speaker recruiting, just the eBay Research Labs folks to develop the Practical Sentiment Analysis tutorial, Katie Paine as a speaker, and the Visionaries Panel members and moderator who is, of course, yourself. (Thanks!) We ran an open Call for Speakers and got 39 proposals for 8 presentation slots. The result is a program that meets a key concern: Speakers are experts with the ability to communicate their experience in a range of real-world problems. Just like last year, we’re addressing a second key concern: We’ve built in lots of networking time for attendees to learn from one another and, if they’re so-inclined, do a bit of deal-making. And new this year, we’re offering an optional Practical Sentiment Analysis Tutorial, taught by eBay staff, Monday afternoon preceding the symposium. The tutorial is specifically designed for folks who are new to sentiment analysis, addressing the need for market education.
SR: Sentiment analysis has become a checklist feature when brands consider text analytic applications, yet the quality varies from one offer to another and that ultimately affects decisions. What do buyers need to learn about judging sentiment capabilities and how will the Symposium help?
Grimes: Yup, quality certainly does vary, especially accuracy and usefulness, issues I’ve addressed in a number of articles, most recently in What I Look For In A Social Analysis Tool, which appeared a few weeks ago in InformationWeek. Buyers can learn from others about experiences with particular tools — the good, the bad, and the ugly — and from vendors about solution capabilities and plans, and from experts (whether users, developers, or analysts) about evaluation criteria and implementation best practices. I’ll just hint that tools that score sentiment based solely on keywords and only at the message or document level are serious laggards. The leading providers now offer beyond-polarity analysis capabilities ability to discern emotion (for instance, angry/happy/sad) and intent signals (e.g., purchase or service-cancellation plans). If you want to learn more, drop me a line, or attend the symposium.
SR: What will attendees be able to do differently or better as a result of attending the Symposium?
Grimes: The short answer is that attendees will gain a better understanding of sentiment technology and its role in larger solutions for targeted advertising, customer relationship management, marketing and market research, social engagement, etc., and they’ll get a chance to meet leading providers. Really, the symposium is designed to put solutions in perspective, to educate users and providers, and to foster deal-making.
SR: Thanks Seth! See you at the Symposium.
Stephen D. (Steve) Rappaport, ARF Knowledge Solutions Director, creates the knowledge resources, tools and services which help members build brands. Listen First! draws upon his decades of experience in listening for market research and for creating listening-based business strategy, services and companies.
Read all Steve's's MediaBizBloggers commentaries at ARF Insights.
Check us out on Facebook at MediaBizBloggers.com
Follow our Twitter updates @MediaBizBlogger
MediaBizBloggers is an open-thought leadership blog platform for media, marketing and advertising professionals, companies and organizations. To contribute, contact Jack@mediadvisorygroup.com. The opinions expressed in MediaBizBloggers.com are not those of Media Advisory Group, its employees or other MediaBizBloggers.com contributors. Media Advisory Group accepts no responsibility for the views of MediaBizBloggers authors.