Most of you know that I serve as Chief Strategist for Converseon, who offers social media listening services, among many other consulting and activation services. Most of you probably also know that I have over 20 years’ experience in text analytics (OK, I am old). So, you might be wondering how social media and text analytics work together. If so, you have come to the right place. Dr. Philip Resnik (Converseon’s Lead Scientist) and I co-authored a white paper that Converseon released yesterday that tackles social media and text analytics.
Text analytics technologies are critical for social media listening because the sheer volume of social conversation makes it impossible for people to monitor what is going in, even in niche subject areas, without automation. And computers happen to be very good at identifying conversations that contain certain words, for example, even if it has to do it with millions of conversations every minute.
But, and it pains me to say this as a certified text analytics geek, computers don’t do a very good job (yet) at knowing whether the conversations that contain those words are actually relevant to what your business wants to know. I mean, if you work for T-Mobile, you’re fine. Anyone who types “t-mobile” in a tweet is talking about that company. But if someone enters “sprint,” it could be about some high school race. Computers aren’t all that good at detecting that difference.
So, you need people to fix the computer’s mistakes. The question is, are they your people or are they the people that work for your listening vendor? (Guess which one we recommend at Converseon?) One reason that you might want your listening vendor to fix these errors is that it costs less than for you to fix them, because the listening vendor can set up an assembly line technology with trained people who do this work all the time.
But there’s another reason, too. With the right technology–machine learning technology–text analytics software can look at those human corrections and detect patterns in what kinds of conversations are being incorrectly identified. The program then “learns” how to do a better job with new conversations that match those patterns.
It might sound like magic, but it actually works. Read the white paper for more details, so that you know what kind of technology you need in your listening solution.