Beyond Key Words
Assistant Professor Jordan Boyd-Graber says scientists already have the computing horsepower to do enormous tasks, like search through every article ever published in The New York Times. What’s missing, he says, are the proper tools to quickly query specific information, like finding stories only pertaining to political reporting on a given topic.
In the iSchool’s Cloud Computing Center, Boyd-Graber is working to scale up “topic modeling” applications that can quickly analyze massive sets of data.
“Without any human intervention, you want to discover the topic, or theme, that runs throughout these large collections,” he says.
Discovering thematic similarities—rather than just matching key words—allows for a much better search of the entire collection, and even finding items that users were unaware of, Boyd-Graber explains.
Ultimately, he says, this approach to data retrieval could soon lead to “thinking machines” able to understand what humans say and do, instead of just carrying out rote commands.
“There is a lot of information out there but we are not really using it well,” Boyd-Graber says. “Part of our research is looking at how we can we distill it into a useful form. The second part is determining how we enable computers to use it.”