At a time when the idea of practical artificial intelligence systems seemed more like science fiction than reality, Dr. Venu Govindaraju led a team of graduate students and research scientists who pioneered the world's first autonomous system capable of deciphering handwritten addresses. Operating at a remarkable speed of 13 postal mail pieces per second, the software system interpreted natural handwriting without the need for rigid guidelines, structured forms, or meticulously printed text. Initially funded by the U.S. Postal Service, it was subsequently adopted by the Australia Post and the UK Royal Mail. It was a transformative development that ultimately impacted the entire field of DAR. Today, the automated software system proficiently reads addresses on nearly all handwritten mail in the United States in real-time, resulting in billions of dollars in savings for the postal service and is widely regarded as one of the first success stories in artificial intelligence.
Govindaraju's approach hinged on innovative algorithms that addressed the seemingly insurmountable challenge of determining the destination for each mail piece amidst a staggering array of nearly 170 million possible choices. The strategy was to leverage the contextual information derived from postal directories, thereby narrowing down the choices (classes) to a small set of possibilities based on dynamically generated lexicons. The intricacies of natural cursive writing, which often rendered individual characters illegible without additional context, were overcome through an interactive and iterative A*-like algorithm. He developed a stochastic recognizer, which uses continuous attributes of structural features so that writing styles can be clustered and distinguished. The success of these algorithms in the postal domain spawned a wide array of new research areas around topics from lexicon-driven and lexicon-free text recognition to pre- and post-processing techniques, multilingual OCR, and writer identification. His work catalyzed a significant shift from heuristic-driven approaches to principled methodologies across the entire document-analysis pipeline.
The impact of Govindaraju's research on DAR is far-reaching. It extends beyond revolutionizing the postal service industry – and well into allied fields such as Digital Libraries and Multilingual OCR. From detecting early indicators of illness outbreaks by processing healthcare forms for the New York State (NYS) Department of Health and enhancing patient safety through automated reading of faxed medical prescriptions, to enabling efficient access of historical documents (especially Sanskrit and Arabic) and retrieval of lecture videos segments with substantial use of white boards and handwritten content, Govindaraju’s techniques of word spotting, transcript mapping, text retrieval and writer identification have created powerful new methods to advance the technology in many applications.
Currently, Govindaraju is applying his expertise in handwriting recognition to the problems of Dysgraphia and Dyslexia among children facing challenges with communication and language impairment. As the Director and Principal Investigator of the $20M National Science Foundation-funded National AI Institute for Exceptional Education – with a mission of social good – he is developing AI tools for the early screening of children to identify their special needs and then giving the teachers AI-based intervention tools to efficiently help them overcome their limitations.
Download Govindaraju's curriculum vitae to review his career achievements, explore his Wikipedia bio, or read the Post Automation Hightlights. For more information or questions please contact Research and Economic Development at firstname.lastname@example.org
UB | AI is a two-year series exploring how UB faculty across disciplines are harnessing artificial intelligence for the public good. Launched on September 6, 2023, the series will discuss AI's role in advancing societal good in the realms of education, health care and the arts, among others.
Brice is joined by UB’s VP of Research & Economic Development, Venu Govindaraju, to discuss UB’s rich history with AI research, how UB is leading the way in AI development for good, and how AI will change education in the future.
Govindaraju explores a different perspective to biometric research. Addressing the challenges and advancing biometric technologies for civilian and homeland security.
I always wanted to become professor. Becoming a professor means you have to do research. I did that and I loved it.
My approach to good research is to recruit good, passionate students and give them the freedom to solve their own research problems.
Wanting to study in the U.S., UB offered me a great package as a grad student earning my PhD. I accepted it knowing only that Buffalo was near Niagara Falls.
Govindaraju’s leadership as head principal investigator (PI) and director of the newly-established National AI Institute for Exceptional Education aims to close the educational gap – leveraging and developing AI as a primary tool – for the nearly 3.4 million U.S. children with speech and language processing challenges.
The Institute for Artificial Intelligence and Data Science (IAD) brings together researchers, labs, institutes, and centers of excellence at the University at Buffalo that are focused on advancements in AI, data science, computational science and related areas of research to tackle these complex problems.
The Center for Unified Biometrics and Sensors (CUBS) was established in October 2003 with the mission of advancing the science of biometric technologies for both civilian and homeland security applications by integrating pattern recognition and machine learning algorithms with sensors technology.
Govindaraju has contributed significantly to the advancement of his fields by mentoring post-doctoral fellows and supervising dozens of graduate students. Upon graduation, the fellows and students have been employed globally in industry-leading companies and prestigious universities.
Their research has ranged from handwriting analysis and recognition to cybersecurity to statistical modeling for medical image segmentation. His students have worked on fingerprint detection, transfer learning for probability density estimation and language motivated approaches for human action recognition and spotting. His post-doctoral fellows have focused on Arabic handwriting recognition and fusion of classifiers in biometric systems.
UB researchers are committed to using AI for social good, including developing new technology that addresses shortages of speech-language pathologists in K-12 education, online deepfakes, or improvements to medical imaging and more!