What’s One Way To Spur Innovation? CMU Researchers Say Teach Computers To Understand Analogies

Aug 21, 2017

Analogies are a common problem solving method in research. For example, the Wright Brothers used their knowledge of how balance and weight affect a bicycle to create the first airplane. Velcro was invented when a Swiss engineer took a closer took at the burrs that stuck to his dog's fur.

Carnegie Mellon University researchers have discovered a way for computers to understand and find analogies like these in hopes that computers, not people, will sift through decades of research to solve problems. The computers will search through databases of patents, inventions and research papers to identify problem solving methods that can be repurposed in other ways. The researchers called the inability for computers to think abstractly about comparisons as the "analogy gap."

CMU computer science researcher Niki Kittur says this technology could benefit researchers in many fields.

"We can think of it as a hybrid system where we have the best of what people are good at and the best of what machines are good at," he said. "They work together to promote innovation."

Kittur says the research team's goal is for the technology to stimulate human creativity. The more research is made by the scientific community, the harder it is for people to find analogies that fit their needs.

"Tools like this are going to be increasingly necessary to keep our rate of innovation high," he said.

The software was developed by crowdsourcing information on how people found analogies, and then developing the technology based on that knowledge.

The researchers hope closing the "analogy gap" will speed up innovation and connect researchers across the world.