Dr. Francisco Pereira, staff scientist at Siemens Healthcare, leads the IARPA KRNS team which comprises 12 people across five institutions (Siemens, Princeton University, MIT, MGH and Harvard University).

Their models are based on learning distributed representations of individual words from text corpora, as well as resources such as FrameNet and WordNet, and using recursive neural network approaches and other techniques to assemble them into the representation of a sentence. The models are validated by decoding mental content from brain imaging data acquired with our own experiments.