Correctly analyzing sentiment around brands is easier said than done and has become a big business—Market Research Future projected that the global sentiment analytics market is expected to reach a validation of $6 billion dollars between 2017-2023.
Innovations surrounding artificial intelligence (AI) have been greatly impacting the evolution of Natural Language Processing (NLP) and sentiment analysis. We sat down with ListenFirst’s VP of Engineering, Ben Kao, and Chief Product Officer, Jon Farb, to discuss how machine learning is changing the industry’s approach to sentiment analysis.
So, what should brands be focusing on? Here are some ListenFirst recommendations on which Instagram Stories metrics and insights are most important.
How do you see AI advancements informing the future of sentiment analysis on social media?
Jon Farb: AI and machine learning (ML) empower sentiment analysis to scale at speed while our language evolves. It seems like every month there is a new term or phrase that works its way into everyday conversation (and further, is used by brands engaging their audiences and vice versa). With AI and ML, sentiment analysis can adapt and classify phrases like “hot girl summer” shortly after their first use. In many cases, a machine can determine the sentiment of a new phrase even before its use in language is broadly understood.
Going forward, do you think supervised machine learning or unsupervised lexicon-based methods will dominate the future of sentiment analysis?
Ben Kao: I’d go with supervised machine learning. With unsupervised learning, generally you’re thinking of things like clustering (i.e. grouping). While you might be able to group conversations/texts together, you’d still have to go back and manually apply a label for that grouping to have any meaning. Therefore, unsupervised learning is fairly difficult to solely rely on.
Do you think sentiment analysis will get to the point where it can correctly identify tone?
Ben Kao: I think that sentiment analysis could ultimately detect tone. In terms of tone and context, you need to not just analyze data but actually label data at a high enough volume to train a classifier, enabling it to identify patterns and sub-classifications, such as sarcasm.
What are the biggest pain points for brands currently utilizing sentiment analysis?
Jon Farb: Brands tend to fall into traps of viewing sentiment analysis in its own bubble, and often weigh it above or outright disregard other marketing signals. Sentiment helps brands gauge audience perception and intent, enriching their understanding of social media audiences and consumers. However, it’s just one of the many important signals needed to completely understand how audience engagement with brands should be analyzed on social.
What differentiates ListenFirst’s Brand Reputation Index from other social media sentiment analysis tools?
Ben Kao: It’s about benchmarking. With ListenFirst’s Social NPS, the Brand Reputation Index, brands have the ability to create a cohort (i.e. set of brands) and then see how their particular brand performs. That’s a very useful research tool. Analyzing a brand’s sentiment without knowing how it relates to sentiment around competitors is less actionable—they need to have context.
Why has it taken so long for the industry to produce a Social NPS and why did ListenFirst get there first?
Ben Kao: ListenFirst has collected billions of consumers signals from every social channel for every major brand (client or not), giving us the most complete dataset out there. Establishing social NPS requires having the most complete cross-platform social analytics solution and collecting all major social media signals over an extended period of time – all boxes that ListenFirst checks.
Want to learn more about how ListenFirst approaches sentiment analysis? Request a demo today!