Data Frontiers: Subjectivity, Sentiment, and Sense
By Seth Grimes on June 7th 2017Read this article on our full site
Seth Grimes explores what distinguishes advanced social insight tech from basic - covering the complexity of language, sentiment, opinion, and emotion.
The secret of advanced social insight is that there is no secret.
The data, analytics, solutions, and guidance you need are available and accessible, whatever your industry. Your challenge is to choose the right tools – the right methods and sources and of course the right software – for the job at hand.
That is no small caveat, however. Success is correlated with subject-matter expertise and analysis judgment and an open-to-exploration attitude.
Given the Brandwatch context for this article, you very likely have strong social analysis at your disposal – that’s my take-away from a look at Brandwatch Vizia 2 and the Audiences product at the recent Now You Know Conference – so we’ll focus on key elements that distinguish advanced social insight tech from foundational varieties: the ability to cope with the complexity of written (and spoken) language and to quantify and exploit sentiment, opinion, and emotion.
Make sense of subjective data
First point: In working with subjective, “affective” information, there’s no substitute for human judgment.
The social sources we’re analyzing were designed for people-to-people interactions. Machines can be trained to analyze social and online text and enterprise feedback – a job for natural language processing (NLP) – but they’re not yet up to gold-standard human understanding.
So we seek AI in the sense of Augmented Intelligence rather than (just) Artificial Intelligence.
AI in either sense, artificial or augmented, must go beyond simplistic counting, cookie-cutter dashboards, data-source limits. You’ll want to consider both numbers and also the sentiment, emotion, and intent signals captured in social text.
There are complications. Ambiguity and accuracy are two of them. Troy Janisch from U.S. Bank has grappled with these challenges.
“U.S. Bank is a very general brand name, so we pull a lot of references to ‘US Bank’ that aren’t specific to our brand. Sentiment can be hit and miss. To address these problems, we created an audit dashboard that we use to verify every brand mention for relevance and sentiment. We have more than 20,000 mentions a month that are verified by someone on the team during the month.”
Women in Tech: A Study on the Gender Gap in 50 Tech Companies
Cascade Insights share the findings of their study on the gender gap in tech, taken from LinkedIn data on 50 technology giants
Sentiment is essential
Now, sentiment is a tricky thing, subjective attitudinal information.
I liked that one Now You Know speaker, Brandwatch’s Kate Hoffmire, emphasized the importance of looking at outliers, at relative sentiment across market categories, and at changes over time with insights conveyed via visual story-telling.
Sometimes you need to go deep.
Again quoting Troy Janisch: “Sentiment is essential. It’s even better when you can look at subtleties such as emotion and intensity.” (Troy will be speaking at an upcoming conference I organize, the Sentiment Analysis Symposium, which focuses on the role of emotion in consumer insights and market understanding.)
Some providers do differentiate emotion, expressed in categories such as surprised, afraid, disgusted, angry, and sad, from positive/negative/neutral sentiment assignments. Again, subjectivity is tricky. For example, I’d argue that the use of the term ‘sadness’ with regards to Gregg Allman’s recent death represents positive sentiment about Allman, rather than negative.
U.S. Bank works with Brandwatch partner Converseon to get at deeper sentiment/emotion/intent signals. It’s a natural partnership for the companies, one that lets shared clients grow into advanced insight capabilities when they’re ready.
Quick side note: Converseon is a Sentiment Analysis Symposium sponsor. CEO Rob Key, whom you might have met at the recent Brandwatch Now You Know conference, will speak alongside client Thomas Wilson of Uber, on Measuring Emotions in Brand Health. Troy Janisch’s symposium talk is titled “Social Benchmarking You Can Bank On,” and he’ll participate in a panel discussion with Verizon analytics strategist Jonathan Schwedel, who’ll be speaking on Social Listening. Oh, and you can contact Converseon for a symposium registration discount.
A key to advanced social insight
But wait, there’s more! Troy’s take on another advanced capability, one you may also be using: “Visual search, the ability to find photos that include our logo without any specific mention is also great. It’s the kind of thing you don’t realize you need until you have it.”
In fact 80% of the images that contain logos, don’t mention that brand in the accompanying text. So, if you’re only searching for text mentions, you’re probably missing thousands of images about your brand or product.
Brandwatch has spent a while building their product, but it’s worth the wait. They’ve created a solution that’s 2x more accurate than the nearest competitor. Cool stuff, no? Technologies that wrangle human language and mine images in order to produce advanced social insights.
There’s no secret involved, just having the right tools available – the right methods and sources and of course the right software – for the job at hand.