Social Media Monitoring or "listening" to what's getting said on popular social media channels about your company or products has become an important use case in recent years.
Even the United States Secret Service released a work order last month to procure a social media analytics tool. They are looking to automate and streamline their social media (mainly Twitter) monitoring process. The solicitation lists standard functional requirements like search, trend analysis, visual reports, influencer identification, audience segmentation, and so forth.
Interestingly, the Secret Service also added a requirement for the "ability to detect sarcasm and false positives." It's is easy to see the rationale for this requirement in a medium known for both the volume of chatter and the prevalence of sarcasm. Given the current state of maturity of the software, however, I'd categorize that requirement as more as a "wish list" item.
Our Marketing Automation & Social Technology Report, in a section titled “A Caution on Sentiment Analysis,” explains how this functionality is fraught with many perils, with the accuracy of sentiment analysis usually coming in far lower than vendor claims.
Tone, meaning, and intent are complicated. You can perhaps tell when a friend or a colleague is being sarcastic given your history of interaction and the presence of physical and visual clues. But can you be so certain when dealing with a stranger? If it's difficult in face-to-face interactions, mining text to predict sentiment is further limited in the absence of body language and tonal clues. Algorithms can perhaps infer tone from lengthy pieces of text but tweets with their 140 character limit pose additional challenges – there simply may not be enough context. Not to mention multiple sentiments being expressed in a single post. Or that a single tweet can have phrases in multiple languages. And don't get me started on slang.
When tool vendors claim a very high rate of success with sentiment prediction, they typically refer to plain vanilla scenarios. No doubt the algorithms for sentiment analysis are getting better as we gain more experience in this field, but these are still early days. If your enterprise is looking for social media monitoring technology, make sure your testing is comprehensive enough to yeild a realistic picture of what the tool can do and what it cannot do.
To know more about the intricacies of social media analysis, which of the services are mature and which are immature, the caveats and how different vendors stack up against social media monitoring and analysis use cases, consult our Marketing Automation & Social Technology Report.