Abstract: Teaching machines to read, and how it will save lives

Abstract: Teaching machines to read, and how it will save lives

One of the oldest challenges of artificial intelligence is to have machine read documents created for humans, also known as OCR (Optical Character Recognition). It has taken 60 years to solve easy problems, such as accurately transcribing high quality printed text, but new problems crop up in various applications, and the field presents a rich collection of challenges for algorithmic mathematics, machine learning and a variety of related areas.  Currently, I and several math graduate students are engaged in joint research with the College of Medicine that aims at automated reading of medical forms used in organ transplantation, which has the potential to save thousands of lives per year and affect medical policies nationwide. In this talk I will give an overview of this project and the kind of research questions that it brings about.