🚀 go-pugleaf

                         UNIVERSITY AT BUFFALO
                      State University of New York

                     DEPARTMENT OF COMPUTER SCIENCE

                               Colloquium


                            JONATHAN J. HULL

       Center of Excellence for Document Analysis and Recognition
                     Department of Computer Science
                State University of New York at Buffalo


       LANGUAGE-LEVEL CONTEXTUAL CONSTRAINTS IN TEXT RECOGNITION

  The use of constraints derived from the language in which  a  document
  is  written  to  improve  the recognition of the text contained within
  that document is explored.  These constraints span a range from graph-
  ical context to the syntax and semantics of the underlying text.  This
  talk surveys several techniques that have recently been developed  for
  utilizing language-level contextual information to improve the perfor-
  mance of a word recognition method.

  An algorithm that uses patterns of word repetitions within a  document
  to  recognize  function and content words is discussed.  Another tech-
  nique that employs an information retrieval method to improve the per-
  formance  of a word recognition algorithm by matching partial recogni-
  tion results to a free text database is explored.  Also, a hidden Mar-
  kov  model  that  uses  statistics about the co-occurrence of part-of-
  speech tags to improve recognition performance is presented.

                    Place:  20 Knox, Amherst Campus
                   Date:  Thursday, November 12, 1992
                      Time:  3:30 P.M. - 4:45 P.M.

     Refreshments will be served in 224 Bell Hall, after the talk.
                   For further information, contact:
    Dr. William J. Rapaport, (716) 645-3193, rapaport@cs.buffalo.edu