John Prager

Also known as John M. Prager

John Prager is a retired IBM Research computer scientist whose question-answering work fed into Watson, the system that beat two former champions on national television in February 2011. After degrees from the University of Cambridge and a 1979 PhD in low-level vision at the University of Massachusetts, he joined IBM's Cambridge Scientific Center, where he led the REASON project on natural-language help. In 1992, he moved to IBM's Watson Research Center, where he co-developed predictive annotation for answering fact-seeking questions and led IBM's entries in the TREC question-answering evaluations. As a DeepQA team member from 2006, he worked on Watson's question analysis, wordplay and game strategy, shared the 2013 AAAI Feigenbaum Prize as part of the IBM Watson team, and later coauthored WatsonPaths, a 2017 system for medical questions.

Badges: Laureate
Bridge

Timeline

1974–1979

Cambridge degrees and a vision PhD

He earns a BA at the University of Cambridge in 1974, a Diploma in Computer Science with Distinction in 1975 with a dissertation on intelligent help systems, and an MA in 1977; his Cambridge degrees span mathematics and computer science. In 1979, he completes an artificial-intelligence PhD at the University of Massachusetts on low-level vision, exploring how a moving subject can use rapidly changing visual input to analyze the scene ahead.
1979–1992

Leading the REASON project

He joins IBM in 1979 and works until 1992 at the IBM Cambridge Scientific Center in Cambridge, Massachusetts, on projects involving intelligent user interfaces. There he leads the REASON (Realtime Explanation And SuggestiON) project, which helps users by processing natural-language questions with an inference engine tied to a large repository of facts and rules about network-wide resources.
1992–2006

Predictive annotation and TREC

In 1992, he moves to the IBM T.J. Watson Research Center to work on interfaces, search and question answering, including language identification, web search and categorization. With Eric Brown, Anni Coden and Dragomir Radev, he presents predictive annotation at SIGIR 2000, a technique that finds potential answers to questions in text, annotates them and indexes them. He leads IBM's entries in the TREC question-answering tasks, an annual evaluation at NIST.
2006–2011

Building Watson for Jeopardy!

He joins IBM's DeepQA research team in 2006 to develop Watson, a system designed to play Jeopardy! at champion level, and coauthors the 2010 AI Magazine overview “Building Watson: An Overview of the DeepQA Project.” In February 2011, Watson plays two former champions on national television and wins. KQED interviews him in San Francisco in February 2011 as a developer who helped Watson understand human language.
2012–2013

Question analysis and game strategy

His Watson work centers on question analysis, wordplay and game strategy. In May 2012, he coauthors “Question analysis: How Watson reads a clue” in the IBM Journal of Research and Development. In 2013, he coauthors a Journal of Artificial Intelligence Research article on Watson's wagering, square-selection and buzz-in strategies, built with a simulation model, learning and Monte-Carlo methods.
2013

The AAAI Feigenbaum Prize

He is one of the IBM Watson team members who receive the 2013 AAAI Feigenbaum Prize for showing that a synthesis of symbolic knowledge representation, natural language understanding and statistical machine learning can achieve human-level performance in real-time factual question answering.
2016–2017

WatsonPaths for medical questions

In May 2016, speaking for the IBM T.J. Watson Research Center and IBM Watson Health, he lectures in Vienna on Watson's adaptation to clinical decision support in healthcare; by 2016, his publications include fifteen patents and a book on Turing. In 2017, he coauthors an AI Magazine article on WatsonPaths, a system that breaks a patient summary into pieces, asks Watson subquestions and runs probabilistic inference over a graph to answer medical questions.
As of October 2026

Retired from IBM Research

As of 2026, he is retired from IBM Research, with research interests in AI and natural language processing.