Statement on Research, Teaching and Service

“Naturalness ” of Software. This is a recent, very exciting discovery, with substantial scientific and practical implications. We showed that software has striking similarities to human linguistic behavior, in that it is highly repetitive and predictable [1] using models adapted from Natural Language Processing. Indeed, we found that code is 8-16 times more predictable than English— this, together with the large mixed corpora of code and English that are widely available, opens up a large realm of research, rife with scientific questions and engineering promise. This work has led to a total of about $2,900,000 funding from the NSF (3 grants, including two EAGERs and one NS