The ICICLE Project: An Intelligent Written English Tutoring System for Deaf Students

Status:

(interface pic)

ICICLE's current implementation is a prototype application implemented using Wx Windows and developed in parallel on the UNIX and Windows platforms. It makes use of the TRAINS text parser (Allen, 1995), the COMLEX Syntax 2.2 lexicon (Grishman, Macleod, and Meyers, 1994), and a mal-rule-augmented custom grammar (Schneider and McCoy, 1998). At this time, the system is able to recognize many morphosyntactic errors, but it selects between competing parses for each utterance by choosing arbitrarily - selecting the first grammatical parse it encounters, if any, or the first overall if no grammatical parses were found. The interface allows the user to type in or load a text file, request an analysis, and view the results. Sentences containing problems are highlighted in colors corresponding to the type of error and canned one-sentence explanations of the error can be accessed.

Present areas of research include:

  • The development of the user model, including preliminary statistical analysis of samples from our corpus of writing from the learner audience.
  • The use of machine learning to train a text planning model to replace the canned explanations.
  • The implementation of a dialogue interface to add proper nouns to the lexicon, including features extracted from icons selected by the user.
  • The development of a spell-checker algorithm with rules modified by those in "aspell" in order to capture the unique spelling behavior of the user population.
  • The integration of explanatory material as video-recorded ASL performed by an ASL interpreter, and the investigation of "signing avatars" to incorporate sign language instruction generated by the tutorial component of the system.