ITBusiness.ca

Watson plays Cupid: new set of APIs give IBM’s AI an emotional intelligence upgrade

Connectidy and IBM Watson

LAS VEGAS — Thanks to a new set of algorithms, now IBM’s Watson technology can not only beat you at Jeopardy, but it can probably score more dates than you too.

Well, Watson is no Don Juan just yet, but with several new APIs that IBM released this week at its InterConnect conference, Watson has definitely seen an emotional intelligence upgrade. The new cognitive APIs are being released on the Watson Developer Cloud on Bluemix for developers to use in their own programs. One IBM beta partner – a soon-to-launch online dating site based on relationship science – is already putting some of these capabilities to the test, with the aim of helping its users craft that perfect introduction message in hopes of scoring a date.

Connectidy, the online dating service that plans to launch in New York within the next four weeks, has been testing Watson’s new Tone Analyzer. Released in beta this week, the API aims to give users feedback on the tone they’re conveying in their written communications. It can rate your text on how it conveys a wide set of emotions such as joy, disgust, fear, sadness, or cheerful. It also gives feedback on what social propensities are conveyed – are you sounding open and agreeable, or conscientious? The tool analyzes text at the sentence level.

One of Connectidy’s co-founders was involved with IBM research for many years, explains Dineen Tallering, president of Connectidy. So when IBM learned that he was working on an online dating site that was going to use algorithms to help people make connections, it was a perfect opportunity to put Tone Analyzer to the test.

On Connectidy, Tone Analyzer is used to display real-time feedback on a user’s tone as they write. Tallering expects it will help users overcome that nerve-wracking moment when they’re writing that first message to that cute guy or girl they want to date.

“Somebody writes us something and we don’t often get what they’re saying. Sometimes we take it the wrong way,” she says. “Tone analyzer is more helpful when you haven’t seen someone face to face and you haven’t had an exchange yet. This is to help you get that first level of communication right.”

Also being added to Watson’s suite of emotional capabilities is an Emotional Analysis tool. It serves a similar purpose to the Tone Analyzer, but is meant to help users understand the underlying emotion of written content from others. It goes beyond the typical sentiment analysis that might be part of your social media analytics package, distinguishing a broader range of emotions beyond positive or negative.

Developers also have a new Expressive Text to Speech tool that aims to make those interactive voice response (IVR) systems more dynamic. Rather than sounding flat and emotionless, developers could program a response that varies its inflection to sound emotionally sensitive.

“They’re all different flavours of getting underneath the emotion of language and reading patterns of emotion over long periods of time,” says David Kenny, general manager of IBM Watson. “We’re getting smarter and communicating more effectively.”

Business applications for Watson’s new and improved emotional intelligence quotient go beyond playing Cupid. A company could better understand the overall sentiment coming from social media feeds and review websites, Kenny says. Users of IVR systems could program a tone and cadence that is appropriate for different segments of customers.

“A developer could say ‘start off apologetic and then end with good news,'” he gives as an example. “The voice actually sounds more empathetic.”

But he cautions against actually trying to fool people with Watson’s advanced ability to empathize. “We never want to post like a person when you’re not,” he says.

At Connectidy, Tallering is also interested in honesty, and sees Watson as a way to help people communicate their true selves on their profiles.

“Most people misrepresent themselves online,” she says. “It’s because people don’t really truly understand themselves.”

So instead of relying on people to fill out an online profile, Connectidy offers to analyze text they’ve written to generate a personality matrix. Users can plug in their Facebook feeds or Gmail to feed Watson the data it needs to do the assessment, and the data isn’t stored on any server afterward. All users are also authenticated to make sure they’re representing their real identity.

Watson’s new emotional capabilities are available to developers interested in getting in touch with their more sensitive sides via the Watson Developer Cloud on Bluemix. IBM has also released new SDKs so Watson’s APIs are now easier to call upon for developers using Swift and Unity (already supported SDKs include Node, Java and Python.)

 

Exit mobile version