Cookies on the Fluid Rock website

  • Our FR_og

    Our Fluid Rock Blog. Where we keep an eye on the latest technologies & trends that are flying about.

Sentiment - What is it & how is it gauged

First published in the Irish Marketing Journal in April 2011. This article examines sentiment analysis and available products in the market at that time, in the 12 months since publishing, Radian 6 has been the most commonly mentioned product out there, but there still isn't a recognisable leader.

My technology colleagues have been doing the same thing in different ways for the past ten years.  Over the past six months, there’s been a buzz in the office, something is grasping all of our imaginations, actually not something, too many things to take account of.  There’s a shake-up, disruptive technologies is how it’s reported in the press and technology media.  We’re taking out and dusting off our internal blog, we’re in information overload, new products, new platforms and amazing ideas and our collective creative brain is in overload.

What is just a feeling becomes reality when I’m asked to login to a new system that one of our technologists has developed in ‘his spare time’.  I don’t think we’re keeping him busy enough,  actually on second thoughts, he can probably take care of routine work with his eyes closed at this stage, in fact he has probably written programmes to take care of his responsibilities. 

I take a look at the computer screen and before my eyes I’m presented with a solution to the big problem I have with Social Media -  measurement.    It’s a sentiment engine.  OooKay!   I scan the screen and take in terms like actors, sentiment, influencers.  I see parameters like keywords, social media websites and brands.  Underneath,  I see listings of posts and comments in realtime, he has even visually presented some of the data in graphs.  As usual I start asking questions which are answered with links to Wikipedia and Stanford University. Your job, I’m told, is to figure out what marketers need. Then, begins my journey of discovery, using the tried and trusted methods of market and competitive analysis. 

Some of our clients have already been supplied with sentiment reports from agencies around the world.  I’ve looked at them, and questioned them, how do they come up with that conclusion, what was their starting point, is that not very subjective and where exactly did they get this data? 

As I delved into the companies and products in this space, read the whitepapers, watched the videos and viewed the products already on the market, I realised that the easy part is collecting the data, and many of the products are very good at doing this.  When you move into measuring the sentiment it’s a different story.  

Sentiment analysis tries to determine the attitude, opinion, emotional state, or intended emotional communication of a speaker over time.  This is done by classifying subjectivity, polarity, tonality  and emotion.  Applying this analysis to the volume of online user generated content can only be done with the help of computers.  Most of the products I’ve reviewed provide sentiment on a graph that illustrates positive or negative sentiment, and in some instances, neutral.  Depending on the product, sentiment  is controlled by parameters input by humans, either at the start or end of the process.  For example the ScoutLabs  product recently acquired by Lithium Technologies, has a Human override of machine scored sentiment values which it says will continually improve sentiment accuracy as the machine "learns" from human override.  

Since I’ve started researching this area, many of the companies I had identified have been bought by larger entities.  From a marketing perspective the most interesting is probably the acquisition of Radian 6 by in late March 2011.  In press releases following the acquisition, Radian 6 is said to provide detailed dashboards and basic sentiment analysis.  Analysts predict will have to acquire one more company to bolster the Radian 6 sentiment analysis capabilities. For many of the products, I think the same applies to the sentiment element of their product offerings. 

So to answer my colleagues’ question, what do marketers need from a sentiment engine, I would surmise that they won’t know exactly what they need until they start experimenting and using what’s available.  They won’t be able to judge the effectiveness of the analysis until it’s been tried and tested in their company. 

Marketing managers do need something beyond Likes and Share counters to measure effectiveness and the new evolving products will help them identify opportunities and address challenges.  They will need in-house resources to set-up, analyse and take action based on sentiment.  And above all the will need to have confidence in the product they choose, as it will be another few years before the market leader emerges.

Follow us :