When it comes to social media, determining sentiment isn’t as clean and straightforward as it is in a journal or newspaper. In social media people write like they speak. This means heavy uses of slang, shorthand sentences and words, poor grammar, idioms, and industry/brand/regional specific dialect.  There is a big difference between automated Generic Sentiment analysis and Adaptive Sentiment™ analysis and how it relates to Social Media Monitoring.

Generic Sentiment Analysis

Almost every 1st Generation Social Media Monitoring tool uses generic sentiment analysis algorithms with a majority using the same outsourced technology company to do so. This technology is based on understanding sentiment from proper sentence structures from a textbook or newspaper article. If it says “good, delicious, fast” it’s going to be positive. If it says “gross, disgusting, sick, slow”, it’s going to be negative.  Generic Sentiment analysis is an approach that just doesn’t lend itself to social media and all of its informal communications and industry contexts.  Because of the language nuances and industry contexts, Generic Sentiment analysis typically produces accuracy results in the 50%-65% range, sometime much lower depending on the industry topic.  That’s “flip of the coin” accuracy.

Adaptive Sentiment™ Analysis

Our product, RESONATE, uses automated Adaptive Sentiment™.  Adaptive Sentiment is our patent pending technology that incorporates machine-learning and human feedback to fine-tune sentiment algorithms and continuously LEARN the unique language of your brand or product within social media.  The language and medium for every brand is different, which means the words that mean positive, neutral, or negative will also be different.

Examples:

For an electronics manufacturer, a product seen as “Sick”, “Nuts”, “Dope”, “Crazy”, or even “The Bomb” is a good thing, it’s positive.  Using Generic Sentiment, these words are typically an indication of something very wrong, but it depends on the context and industry.  Context and meaning changes from product to product, brand to brand, and industry to industry. Another example from one of our clients, the words “nom nom nom” are meaningless, unless they’re used in the food industry, where “nom nom nom” means “delicious.” Generic off-the-shelf sentiment analysis would score this typically as a neutral comment.

What does this mean for your brand?

If you’re running a social media monitoring campaign it’s critical to ACCURATELY know what people are saying about your brand, market and competitors.  Adaptive Sentiment technology LEARNS your brand and the unique language used by your consumers.  It learns that “nom, nom, nom” is good if you’re in the food industry, and “leaking” is bad if you’re in the plastic bag industry. Your brand is accurately measured and the algorithms constantly fine-tune themselves with human feedback as opposed to the generic model which simply runs a basic sentiment algorithm and never learns.  Because of its continuous learning intelligence, Adaptive Sentiment delivers sentiment accuracy in the 90% range vs 50%-65% for Generic Sentiment.

At ListenLogic, we service each client with a dedicated analyst that custom builds your Adaptive Sentiment models and provides daily optimization of your account to deliver highly accurate brand sentiment in real-time.  We help your organization identify new threats and opportunities each business day. Social media monitoring is a powerful way to gain insight into your business, but if the information you’re receiving isn’t accurate you’re wasting the effort.

See it for yourself

If you’d like to take a demo and see how our enterprise listening solution can better serve your brand or clients, contact us today and request a demo.

THE VERDICT IS IN.  According to ListenLogic’s Voice of the Consumer Report, Audi was the talk of social media during Super Bowl XLIV:

FINDINGS:

Audi was the most talked about and Dodge had the most positive sentiment.

Methodology:

ListenLogic used its proprietary social media listening platform and analytic methodology to tap into the conversations relating to Super Bowl XLIV automotive ads across message boards, forums, microblogs, blogs, news and social-networking sites.

The first step in this study was to gather a clear signal of the Super Bowl viewer and remove irrelevant noise. Our analysts used a combination of automated and manual techniques to identify accurate mentions, removing commercial editorial from the data.