VP for Research & Economic Development,
SUNY Distinguished Professor, University at Buffalo
Dr. Venu Govindaraju is Distinguished Professor and Vice President for Research and Economic Development at the University at Buffalo, State University of New York. Renowned for his expertise in Artificial Intelligence, Dr. Govindaraju has made significant contributions to the field, particularly in handwriting recognition. His groundbreaking work was instrumental in developing the first handwritten address interpretation system used by the U.S. Postal Service.
With a prolific career spanning over three decades, Dr. Govindaraju has led research projects funded by federal agencies and high-tech companies totaling nearly $100 million, holds six patents, and authored close to 500 refereed scientific papers. He is the principal investigator of the National AI Institute for Exceptional Education at UB, a prestigious $20 million initiative funded by the National Science Foundation and the Institute of Educational Sciences.
Dr. Govindaraju is the Chief Research Officer at UB, where he oversees the university’s $570M research enterprise, fostering industry relations and driving economic development.
Dr. Govindaraju is a Fellow of the American Association for the Advancement of Science (AAAS), Association of Computing Machinery (ACM), International Association of Pattern Recognition (IAPR), Institute of Electrical and Electronics Engineers (IEEE), and National Academy of Inventors. In 2024, he received the University at Buffalo President’s Medal; as well, he was named Person of the Year by the Council of Heritage and Arts of India for his contributions to AI.
” The Evolution of AI: A perspective on language communication “
Our journey through the handwriting recognition landscape will highlight the transition from lexicon-based to lexicon-free approaches, and from heuristic-driven techniques to the principled methodologies we pioneered. These include applications that span the spectrum of the “use of context” in reading and interpreting documents. On the one extreme we have postal address recognition, which is highly constrained with a database of zip codes, city names and street names; on the other, we have the task of reading a 20th Century, American poet’s handwritten diary with no fixed rules or format. Clearly, AI approaches benefit from the availability of context, which led to some early success as in the case of postal automation.
The challenge of reading children’s handwriting adds a new dimension of difficulty because of lack of training data and privacy concerns. Furthermore, identifying children with dyslexia requires OCRs to recognize patterns and characters that are outside the alphabet. In fact, we argue that it requires the ability to use context in a dynamic fashion rather than a simple binary where transcription must preserve the spelling errors without correction by a dictionary. We explore the AI advances to address “scalable context” to solve this problem. We contrast with the superficial adaptation by LLMs, which lack mechanisms to “choose” the appropriate level of context.