Coding IBM's Watson

October 11, 2017

As an Extreme Blue intern at IBM Michelle Morales, a Ph.D. candidate in linguistics, worked with the famous computer.

As a member of IBM's selective Extreme Blue internship program this past summer, Michelle Morales got to work with Watson - the supercomputer of Jeopardy! fame - to find a new way for IBM to listen to customer feedback and, ultimately, improve its products. For Morales, a Graduate Center Ph.D. candidate in linguistics and a self-described sci-fi nerd with a deep interest in artificial intelligence and natural language processing, it was "an amazing experience."

Essentially, Morales and her three teammates - an M.B.A. student and two undergraduate computer science majors - enhanced Watson's ability to process human language. They fed Watson reams of unstructured data from all sorts of sources - feedback forms, online review forums, question and answer websites, etc. Using code, they instructed Watson to determine the keywords and the sentiment of those words, e.g., overpriced, reliable, fast, etc. - all standard capabilities for the supercomputer.

To make the outcome specific for each product, though, Morales and her team extended Watson's capabilities by writing new computer code. The result was a dashboard that managers can use to understand how "to really push the innovation" of their cloud computing, security, and machine learning products, Morales said.

At the end of the 12-week internship, Morales, who was the technical lead on the project, and her teammates pitched their product to about 150 top IBM executives. Weeks of rehearsing yielded a satisfying result. "The Chief Analytics Office decided to continue the project full time," Morales said.

Morales described the culture within IBM's Chief Analytics Office, where she spent most of her time, as "wonderful" - an interesting mix of M.B.A.'s and Ph.D.'s, including a number from the GC. "I couldn't have asked for a better internship experience," she said.

 * * *

Morales credits the GC with helping her discover both the internship - through a posting in the Office of Career Services and Professional Development - and her field.

A first-generation American - her father immigrated from Guatemala, her mother from Argentina - Morales had long been interested in how people produce and understand language. In her first semester at the GC, she took an introduction to computational linguistics course. "That was a really mind-blowing, eye-opening experience," she said, "and it completely changed my entire journey."

Morales says she'll "be pretty happy" if she can continue to pursue natural language processing work, involving innovation and technology.

"I think the GC has done a really nice job of supporting academic alternate careers," she said. "I feel really supported going forward. I've had this wonderful opportunity here at IBM because of the GC, so I feel pretty blessed."