IBM’s New Analytics Software Interprets Slang and Even Emoticons for Better Service

Sarcasm, emoticons, slang, and idioms are usually difficult for software to interpret, which makes it tricky for companies to analyze huge amounts of customer feedback through comments, blogs, and emails. IBM’s new software might be able to help.

IBM logo


Here’s an example of a sentence that would be tough for most software to interpret: “The iPad’s great…if you don’t care about Wi-Fi. :-p” The sentence isn’t actually pro-iPad; it uses sarcasm, implicit criticism, and an emoticon to convey distaste. But if, say, Apple tried to use most kinds of analytic software to glean reaction to the iPad, that would show up as positive. It says “iPad” and “great”!

IBM‘s new predictive analytics software scours the Web for feedback on the product or service of your choice, and is able to interpret slang and emoticons from blogs, comment boards, and other far-flung corners of the Internet to give businesses a more accurate view of what their customers are saying.

Recognizing that each industry has unique priorities and its own vernacular, the new software analyzes trends and captures insights from industry-specific terminology. Within these domains, the software includes new semantic networks with 180 vertical taxonomies (from Life Sciences to Banking and Insurance, and Consumer Electronics), and more than 400,000 terms, including 100,000 synonyms and thousands of brands. This allows customers to draw better links and understanding between sentiment and products without having to spend time building their own definitions.

For instance, in the banking industry, the semantic network knows that a “floating rate” is a “Mortgage Loan,” and “Variable Rate Mortgage” and “Adjustable Rate Mortgage” are synonyms. It can also detect that “estate planning,” “older people,” and “retirement planning” are related to “reverse mortgage.”

IBM’s pitching the new software as a means to better understand customer feedback without having to hire actual humans to go read through various emoticon-laden and misspelled comment boards, which is better for everybody’s sanity, really. It’s a little bit creepy in a Big Brother kind of way–I’m glad businesses will be better able to understand the subtleties of my compositions, but the thought of my ideas being indexed and farmed out to whoever pays is a bit discomfiting.

The software is called the IBM SPSS Modeler, and it’s available now.

Dan Nosowitz, the author of this post, can be followed on Twitter, corresponded with via email, and stalked in San Francisco (no link for that one–you’ll have to do the legwork yourself).


About the author

Dan Nosowitz is a freelance writer and editor who has written for Popular Science, The Awl, Gizmodo, Fast Company, BuzzFeed, and elsewhere. He holds an undergraduate degree from McGill University and currently lives in Brooklyn, because he has a beard and glasses and that's the law