Will 2015 answer the big questions about sentiment?

"What is gamer gate?" was the sixth most asked 'what is' question on Google in the US during 2014.

Sentiment analysis is suffering a credibility crisis, argues Sarah Little
Sentiment analysis is suffering a credibility crisis, argues Sarah Little
A continuing controversy that exploded on Twitter surrounding misogyny, harassment and ethics in video game culture, ‘gamer gate’ was still generating a massive 316,669 tweets in a 72-hour period, two months after it first erupted. 

More recently, it has becoming famous for demonstrating the continued difficulty in accurately measuring sentiment. 

Newsweek enlisted a sentiment analysis agency to unravel the social conversation surrounding gamer gate and reported that 90 per cent was neither positive or negative (a red flag given the heat of the discussion) while, in response, other analysis conducted showed that 90-95 per cent did in fact take a clear side. 

A vast discrepancy between sources who must surely have methods of analysis they trust?

In PR, misinformation is dangerous. 

It can be worse than no information at all. Current sentiment analysis is suffering a credibility crisis.

Will we ever reach a point at which we can actually rely on our findings? 

Happily, I believe 2015 will prove a watershed year, for reasons I’ll come on to. But first, let’s look at how these gamer gate fiascos arise.  

Understanding how we measure sentiment is to understand the source of the problems.

Human beings aren’t always accurate in sentiment analysis, let alone machines. 

Ever received an email and wondered why your colleague was being short with you, only to have them greet you warmly in the kitchen? Or had that awkward "Oh, I thought you meant…" discussion when a text message wasn’t read quite as planned? 

Get a computer analysing messages and the same mistakes arise tenfold. Take the following example:  "How can you hate iPhones?? Crazy!"

The message is hugely positive for Apple. 

But a machine interpreting keywords can’t currently see past ‘hate’ and ‘crazy’. Differences in results don’t stop there. 

Are positive and negative sentiments enough? How do you compare results if your advocates prefer to complain about other brands, instead of overtly applauding your own?

Thankfully, questions like these are on more notable minds than my own. It looks as though 2015 will be a year of great leaps in reliable analysis.

In 2014 Hootsuite purchased and integrated analytics start-up uberVu, as well as purchasing social voice technology start-up Zeetl, raising $60m in financing and following on from $165m a year ago. 

Meanwhile, the Stanford professor who claimed to have solved true sentiment analysis in 2013 was recently unveiled as part of a new deep learning start-up, MetaMind, that just raised $8m including funding from VC leaders, Khosla Ventures. 

In addition, Twitter revamped its advanced search to facilitate searching all historical tweets, and Facebook rolled out its graph search functionality onto its mobile app. 

All of these point to a far better understanding of brand perception.

This is an exciting time in social media, where the work of key players like Hootsuite could soon take us beyond a binary thumbs up/thumbs down view of public opinion, gaining a real qualitative measure of social success. 

Social media will move from being simple quantitative data and shallow demographics indicators, to become a pivotal KPI mapped against all our activity – online and offline – with real-time response, global reach and a true understanding of our audience as humans, not numbers.

Sarah Little is social media manager at Captive Minds Communications 

Have you registered with us yet?

Register now to enjoy more articles and free email bulletins

Already registered?
Sign in