Sentiment Analysis or Text Analysis?

Sentiment Analysis or Text Analytics, what’s the Difference Anyway?

Sentiment, and the term ‘Sentiment Analysis’ specifically, has been growing rather surprisingly in popularity and usage according to Google Trends. In fact, ‘Sentiment Analysis’ has overtaken ‘Text Mining’ in search popularity (see chart below).

I mention this because the Annual Sentiment Analysis Symposium is coming up again on June 27^thin NYC. I think the founder, Seth Grimes has been a driving force for the terms’ popularity.

Another reason is the rise of social media monitoring. Unlike the more general terms ‘text mining’ and ‘text analytics’, which are data agnostic terms, social media monitoring is very Twitter centric (though review mining is growing a bit in popularity). Typically 80% or more of a social media monitoring analysis are comprised of twitter data. Twitter data is  limited, just 140 characters, with relatively little accompanying data. Therefore one of the first thing most social media monitoring software attempts to do is to enrich that tweet by assign a sentiment score to it.

A sentiment score is just a that, a semantic judgement as to whether the tweet is likely to be positive, negative, or neutral. In my experience only about 30% of records tend to contain sentiment, usually about 20% positive and 10% negative. The clear majority, or about 70%, don’t contain enough information and so are classified as neutral.

So what?

Because twitter data is relatively simplistic, it doesn’t usually lend itself to more sophisticated text analysis, and so ‘sentiment’ has become a key part of social media monitoring output/reporting.

Is Sentiment Analysis the Same as Text Analytics?

No.

While Text Analytics is synonymous with Text Mining, and these two are almost synonymous with Natural Language Processing (NLP), although NLP is even broader and really related to anything speech or text related (think voice commands to Siri or your car radio/phone system, even when said software doesn’t involve analytics as an output), ‘sentiment analysis’ is just one small feature of most text analytics software.

Over the course of the last 10+ years, text mining practitioners and software vendors slowly moved away from the terms ‘Text Mining’ and ‘NLP’ toward a user friendlier more approachable term, “Text Analytics”.

Therefore, if you are referring to any data source other than social media, using the term Sentiment Analysis is probably too simplistic. In fact, most good text analytics software has more useful and important features than sentiment analysis.

So, What About the Sentiment Analysis Symposium?

I often get asked by those just getting into text analytics which if any conferences they should attend.

I believe there is a misconception about what the Sentiment Analysis Symposium is about. Not just among attendees, but sometimes even first-time presenters who think the event if just about Sentiment.

In my opinion, it wouldn’t really make sense to have a conference about just sentiment. The Sentiment Analysis Symposium has always been about a lot more. The event is about text analytics in general, whether a talk includes anything on positive or negative sentiment identification specifically, or on the somewhat broader but closely related emotion analysis. Talks at the event typically include feature identification, topic clustering and reduction, etc. My talks have often dealt with actionable or applied text analytics.

As a long-time industry observer I believe the only reason the term ‘Sentiment’ was used in original event title was that when this event was created, there were already a couple of events which had “Text Analytics” in their names. The concept of Sentiment in the title helped differentiate. Seth’s event has been very successful, arguably too successful, because the event has long ago outgrown its name, and as I mentioned is about a lot more than just sentiment and social media monitoring. In fact, it is now the top text analytics conference today.

I reached out to Seth Grimes and he confirmed my suspicions “really I think the misconception is that the Sentiment Analysis Symposium is only, or even primarily, about technology… This is a conference from anyone in the research and insights industry who wants to learn advanced measurement methods, linked to business outcomes, or deepen their understanding of same”

Sentiment Symposium 2017

So, back to this years’ event.

Since the first Sentiment Symposium Seth has usually asked me to speak or be involved in some way to champion the marketing research practice area with practical examples. This year I’ll be taking part in a panel discussion on Research & Insights Futures together with Kirsten Zapiec of bbb Mavens and Jen Kern of TracX. (is there a link to the agenda?)

When the first Sentiment Analysis Symposium event was held I don’t recall any marketing researchers, client or supplier side, involved. Things certainly have changed. Everyone has now realized that text analytics is not just a nice to have, but a critical part of any serious data analysis. This year there will be speakers from well-known research suppliers such including Kantar TNS, Ipsos, Mediabrands, Nielsen, and FleishmanHillard, as well as client side practitioners from YouTube, Verizon, Uber, U.S. Bank, IBM, Airbus, and Capital One.

This year especially, there will be quite a few talks from the research & insights business perspective, looking at the link between emotion — a key sentiment aspect, newly machine accessible given the growing computing power — and motivation, loyalty, advocacy, and consumer decisions.

There also looks to be some great text analytics content — in the half-day workshop on Text Analytics for Market Research & Consumer Insights and spread throughout the conference — and also facial coding, neuro methods (Dr. Carl Marci from Nielsen Consumer Neuroscience!), speech analysis, and behavior modeling. Complete agenda here.

If you’d like to attend feel free to use my speaker code for a 20% discount [ODINTEXT]. I hope to see you there!

As usual, if for whatever reason you can’t make it to the event, I’m more than happy to chat about text analytics any time.

@TomHCAnderson

About Tom H. C. Anderson

Tom H. C. Anderson is the founder and managing partner of OdinText, a venture-backed firm based in Stamford, CT whose patented SAS platform is used by Fortune 500 companies like Disney, Coca-Cola and Shell Oil to mine insights from complex, unstructured and mixed data. A recognized authority and pioneer in the field of text analytics with more than two decades of experience in market research, Anderson is the recipient of numerous awards for innovation from industry associations such as CASRO, ESOMAR and the ARF. He was named one of the “Four under 40” market research leaders by the American Marketing Association in 2010. He is also founder and moderator of Next Gen Market Research (NGMR), a professional networking group for consumer insights practitioners, as well as the editor of the NGMR blog. He tweets under the handle @tomhcanderson.