An Intelligent Tutoring System for Deaf Learners of Written English
Lisa N. Michaud, Kathleen F. McCoy, Christopher A. Pennington · 2000 · Proceedings of the Fourth International ACM Conference on Assistive Technologies (Assets '00) · doi:10.1145/354324.354348
Summary
This paper from the University of Delaware describes ICICLE (Interactive Communication and Correction of Language Errors), an intelligent tutoring system designed to improve written English literacy among deaf students who are native or near-native signers of American Sign Language (ASL). The core challenge is that deaf students' reading comprehension level is considerably lower than their hearing counterparts — about half of deaf 18-year-olds read at or below a fourth-grade level, and only about 10% read above eighth-grade level. The paper frames this as fundamentally a second language acquisition problem: ASL is a complete, independent language with its own grammar and syntax distinct from English, so deaf students learning to write English face challenges similar to any second language learner, but without the auditory access to the target language that hearing learners have. ICICLE takes student-written text, analyses it for grammatical errors using an augmented English grammar with error-production rules (mal-rules), and engages students in tutorial dialogue to help them generate corrections. The system has four main components: an Error Identification module, a User Model, a Response Generation module, and a Domain Knowledge Base that includes information about ASL grammar to inform error analysis.
Key findings
The grammar analysis component was built on a COMLEX Syntax 2.2 lexicon with approximately 38,000 syntactic word entries, augmented with mal-rules that allow the parser to recognise common grammatical errors made by deaf writers — such as dropped articles, incorrect verb agreement, missing prepositions, and issues with relative clauses. The user model, called SLALOM (Steps of Language Acquisition in a Layered Organisation Model), tracks each student's proficiency across grammatical features organised hierarchically by acquisition order, tagging each feature as "acquired," "unacquired," or in the "zone of proximal development" (ZPD). The system uses these tags to focus tutoring on errors within the student's ZPD — grammatical features the student is currently in the process of learning — rather than features too advanced or already mastered. The response generation component structures tutorial feedback across five dimensions: content, method, form, history, and manner, allowing it to adapt explanations based on the learner's proficiency level and previous interactions. The team also explored multimodal instruction, investigating when an animated sign agent might be more effective than text-only explanations.
Relevance
This research addressed the critical and often underappreciated literacy gap facing deaf students, grounding its approach in linguistic theory rather than treating deafness as a deficit. By framing English literacy as second language acquisition — where ASL is the legitimate first language — the system respected deaf culture and linguistic identity while tackling a real educational need. The adaptive user modeling approach, focusing tutoring on each student's individual zone of proximal development, was sophisticated for its time and anticipated modern adaptive learning systems. The inclusion of ASL as a potential medium of instruction (via animated signing agents) rather than forcing all interaction through English demonstrated commitment to bilingual educational principles. The work highlighted an important intersection of accessibility, education, and natural language processing — building NLP tools that understand not just correct English but the specific error patterns of a particular learner population. This approach of characterising systematic errors as reflections of first-language transfer rather than random mistakes has implications for any literacy technology serving linguistically diverse populations.
Tags: deaf literacy · intelligent tutoring system · American Sign Language · English literacy · second language acquisition · educational technology · deaf and hard of hearing · natural language processing