Bonnie Dorr
Bonnie Jean Dorr is an American computer scientist specializing in natural language processing, machine translation, automatic summarization, social computing, and explainable artificial intelligence. She is a professor and director of the Natural Language Processing Research Laboratory in the Department of Computer & Information Science & Engineering at the University of Florida. Gainesville, Florida She is professor emerita of computer science and linguistics and former dean at the University of Maryland, College Park, former associate director at the Florida Institute for Human and Machine Cognition,, and former president of the Association for Computational Linguistics.
Education
Dorr is a graduate of Boston University, and earned both a Master's and a Ph.D. from the Massachusetts Institute of Technology. Her dissertation, Lexical Conceptual Structure and Machine Translation, was supervised by Robert C. Berwick.Academic career
Dorr joined the University of Maryland faculty in 1992. At Maryland, she became the founding co-director of the Computational Linguistics and Information Processing Laboratory, and associate dean of the university's College of Computer, Math, and Natural Sciences. She has also worked as a program director at DARPA beginning in 2011 while on leave from Maryland.She joined the Florida Institute for Human and Machine Cognition in 2014. In January 2022, she joined the University of Florida as a professor, where she founded and now serves as director of the Natural Language Processing Research Laboratory.
Book
Dorr is the author of Machine Translation: A View from the Lexicon, a revision of her doctoral dissertation. It describes an approach to interlingual machine translation in which, rather than directly translating text from one language to another, it goes through an intermediate form represented using conceptual semantics. The translations between the syntax of each natural language handled by the system and this form are made using government and binding theory, in contrast to the more typical approach from that time which performed this sort of translation using phrase structure grammars and the unification of feature structures. Her system was embodied in the UNITRAN system, and translated between English, Spanish, and German. However, her work was criticized for its lack of completeness.Subsequently to Dorr's work, rule-based machine translation systems such as hers have largely been supplanted by statistical machine translation and neural machine translation, and some of Dorr's own later work instead focuses on data-driven approaches to machine translation and prioritization of explainability in the face of the recent push for large language models.