Wlodek Zadrozny

Timeline

1. 1996

   NLP task complexity

   He holds a PhD in Mathematics with distinction from the Polish Academy of Science. In 1996, as a researcher at IBM's T. J. Watson Research Center in Yorktown Heights, New York, he introduces a method for measuring the difficulty of natural language processing tasks; its complexity measures use the Kolmogorov complexity of meaning automata.

2. 1999

   Minimum description length and compositionality

   His paper “Minimum Description Length and Compositionality” appears in the 1999 Kluwer volume Computing Meaning. It defines compositionality by combining the minimum description length principle with the idea that a whole’s meaning depends on its parts, linking linguistic and corpus-based statistical approaches to natural language understanding.

3. 2008–2013

   Textual resources for Watson

   From 2008 to 2013, he is responsible for textual resources in IBM's Watson, the Jeopardy! playing machine, and is technical leader for building Watson's first customer care application. Before Watson, he works on semantic search for sales professionals and coordinates research and development of UIMA, IBM's Unstructured Information Management Architecture.

4. 2013

   A Feigenbaum Prize with Watson

   In 2013, AAAI awards its Feigenbaum Prize to IBM's Watson team for demonstrating 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. He appears among the team's listed members.

5. 2013

   Joining UNC Charlotte

   After a 27-year career at IBM's T. J. Watson Research Center, he joins the faculty of the University of North Carolina at Charlotte in 2013. In July 2013, the Ribarsky Center for Visual Analytics announces him as a faculty member focused on natural language understanding and deep question answering.

6. 2015

   Explaining Watson

   His 2015 AAAI paper “Explaining Watson: Polymath Style,” with Valeria de Paiva and Lawrence Moss, argues that IBM's Jeopardy! playing machine needs a formal semantics. It proposes a collaborative effort to develop formal tools for question-answering systems and large-scale natural language processing efforts.

7. February 2017

   Concept embeddings for text

   In February 2017, he and Walid Shalaby propose two neural embedding models that learn continuous concept vectors. They also give a linear-time vector aggregation method for continuous bag-of-concepts representations and test the models on semantic relatedness, concept categorization and document classification.

8. Since 2026

   Professor emeritus at Charlotte

   Since 2026, he is a professor emeritus of computer science at UNC Charlotte. His recent work spans concept embeddings, topological data analysis, medical guideline text analysis and intelligent information retrieval.