It's a contemporary cliché in almost every field of endeavor: we're drowning in data. From the simplest Google search to the most recondite statistics for scholars, the new flood of information must be filtered and channeled and managed.
Enter the "data scientist." That's what Azza Abouzied, an assistant professor of Computer Science, calls herself and her colleagues at NYU Abu Dhabi's Design Technology Lab (DTL). "We want," she said, "to make it simpler for ordinary people to use their data."
That effort is leading the DTL to some interesting places — including the UAE's iconic shopping malls. Under the title of "mall science," Abouzied and others in the DTL are working on helping shoppers navigate through enormous, confusing malls.
Abouzied, born in Bahrain to Egyptian parents, grew up in Dubai, where an A-level computer science course in her British-system secondary school sparked her interest. After earning her Bachelor of Science and Master of Computer Science degrees at Canada's Dalhousie University, she received her Ph.D. from Yale University in 2013.
Through those years she grew increasingly aware, she said, that although data abounds, for any given problem "the amount of relevant data is actually tiny. We can make the machines go faster, but the people using the data need help." Her approach is to "ask people to tell us what they need from the data and then we build the software tools to let them do that."
She offered a simple example: if a car salesman wants to know what color is most popular in, say, Dubai, he can hire people to count vehicles on the streets. But geo-spatial data processing might be simpler and faster — summon up Google satellite photos, scan them for vehicle shapes, tally those up by color, and extrapolate. In India, she said, urban poverty is being assessed by satellite-photo identification of the building materials used for roofing.
But such efforts demand "very sophisticated programs," she said, because "the process is not intuitive…you can't use natural language, since there are ambiguities."
For these methods to become widely useful, she said, they must first become simple. "We want anyone to be able to pull up Google Maps or some other map data and get what they want using our tool." She calls it an "inference machine."
We want to make it simpler for ordinary people to use their data.
Then there is PackageBuilder, software Abouzied developed with three colleagues from the University of Massachusetts before she arrived at NYUAD. Any search engine can provide a list of, say, gluten-free recipes. But imagine asking for, and getting, a seven-day, 21-meal gluten-free menu — limited to 2,200 calories a day. PackageBuilder extends the familiar SQL programming language to allow such searches.
There is also DataPlay, another pre-NYUAD project of Abouzied's. This is a system for querying a database and getting responses as clear data visualizations — graphs, for example — that can then be fine-tuned or amended as needed via a touchscreen. The goal, she has said, is "more friendly and interactive query components."
While DataPlay is mainly theoretical, she said, PackageBuilder has more immediate practical possibilities, and so offers interesting commercial prospects. "Whether it's buying from Amazon or building an exercise plan, people want to be able to build packages," she said. With UMass, she's hoping to proceed with a three-year project to develop PackageBuilder.
The DTL, meanwhile, is moving ahead with CommonTies. Imagine registering for a conference and being given a wristband with a microchip and a tiny light-emitting diode. This lights up whenever you are in the same room with someone who shares your research interests, has attended the same sessions, or has some other connection. Big data makes that possible, as the DTL's prototype is demonstrating. There are non-academic applications, too, wherever people gather.
Abouzied says the DTL is in part modeled after the Massachusetts Institute of Technology's celebrated Media Lab, in tackling projects that offer "magic, impact, and real work" — that is, each effort must provide a "wow factor," have social utility, and involve a genuine advance.
That's where data science goes to the mall. Imagine stepping onto a computer-linked floorplate and asking a touch-screen for directions to a given shop. Instead of a complicated route on a hard-to-grasp diagram, you're given just a general direction. At the next intersection you step on another plate — and it recognizes you, and steers you onward. You repeat until you find your destination.
We all "naturally emit" so much data, Abouzied said, that this is possible. The sensor plate could recognize your foot pattern, your smartphone's digital signature, even your weight, and combine these to identify you and keep steering you the right way.
For the DTL team, led by Abouzied and Jay Chen, also an assistant professor of Computer Science, efforts like these are just the start. A decade from now Abouzied hopes DTL will be a cosmopolitan Middle Eastern hub where people from all over can come to design, build, and test diverse projects of real use.
This article originally appeared in NYUAD's 2013-14 Research Report (13MB PDF).