Server: Netscape-Communications/1.1 Date: Wednesday, 20-Nov-96 23:03:11 GMT Last-modified: Tuesday, 29-Oct-96 19:42:17 GMT Content-length: 6056 Content-type: text/html Michael R. Brent

Michael R. Brent

Assistant Professor of Cognitive Science; joint appointment in Computer Science; Ph.D., Computer Science, MIT, 1991

Research Areas

Computational models of language acquisition, machine learning of natural language, lexical acquisition, Bayesian learning.

Contents

Research Summary
Preprints and Manuscripts
Selected Publications
Postdoctoral Position Available
Courses
Cognitive AI I: Reasoning
Machine Learning
How to contact me

Research Summary

My research focuses on developing and testing theories of how people acquire and process language. My approach is to analyze language acquisition and language processing tasks as problems in reasoning under uncertainty. To carry out such an analysis for a particular linguistic task, one must:
  1. Make a hypothesis about the sources of information people use to perform the task.
  2. Work out a formula for the way in which information of this type affects the probabilities of various linguistic analyses, according to normative rules of conditional probability. This yields a mathematical hypothesis at Marr's computational level about the implications of the information source for the linguistic task.
  3. Develop an algorithm by which the language processor could evaluate various linguistic analyses, according to the formula. This requires specifying how the language processor determines which analyses to evaluate, since there are typically far too many to evaluate all of them.
  4. Using computer simulation, investigate the extent which the algorithm can perform the task.
  5. If the simulations show that the algorithm is effective, use the evaluation formula to make predictions about human behavior and test them by psycholinguistic methods.
So far, my collaborators and I have applied this method to:
  1. Segmentation and word discovery by young children ( short abstract format )
  2. The discovery of syntactic substitution classes by young children ( short abstract html format )

Preprints and Manuscripts

Brent, M.R. (in press). "Advances in the Computational Study of Language Acquisition." Cognition, volume 61. (PS)

Brent, M.R. (draft). "A Unified Theory of Lexical Acquisition and Lexical Access." (PS) .

Brent, M.R., and T. A. Cartwright (in press). "Distributional Regularity and Phonotactic Constraints are Useful for Segmentation. Cogntion, volume 61. (PS)

Cartwright, T.A., and M. R. Brent (submitted). "Early Acquisition of Syntactic Categories: A Formal Model." (view PS)


Selected Publications

Brent, M. R. (1994) ``Acquisition of subcategorization frames using aggregated evidence from local syntactic cues.'' Lingua, 92, 433-470. (PS) Reprinted in Acquisition of the Lexicon, L. Gleitman and B. Landau, eds. MIT Press, Cambridge, MA.

Brent, M. R., A. Gafos, and T. A. Cartwright (1994) ``Phonotactics and the lexicon: Beyond bootstrapping.'' (PS) In Proceedings of the 1994 Standford Child Language Research Forum, Cartwright, T. A. , and M. R. Brent (1994) ``Segmenting speech without a lexicon: The roles of phonotactics and speech source.'' In Proceedings of the 1st Meeting of the Association for Computational Phonology.

Brent, M. R. (1993) ``From grammar to lexicon: Unsupervised learning of lexical syntax.'' Computational Linguistics, 19, 243-262. Reprinted in Using Large Corpora, S. Armstrong, ed. MIT Press, Cambrdige, MA.


How to contact me

  1. email: brent@jhu.edu
  2. phone: 410-516-6844
  3. fax: 410-516-8020
  4. campus coordinates: Krieger 241 (please make an appointment)
  5. physical deliveries: Dept. of Cognitive Science, Johns Hopkins University, Batlmore, MD 21218, USA