Composition Corrector to Support Deaf Professionals and Writers with Hearing Problems

At the 10th ICCHP Conference on Computers Helping People with Special Needs on July 12-14 at the University of Linz, Austria, Martha Birnbaum of Gallaudet University, Washington DC, presented a paper on an innovative assistive technology solution that enables deaf and hard-of-hearing students and professionals improve their writing skills.

The Composition Corrector (CC) is an online, browser-based tool, aimed at deaf students and professionals who are conversant with sign languages like ASF (American Sign Language) but are not very fluent in English. It detects and simultaneously corrects any grammatical errors found in the user’s written English composition.

The writing skills of the deaf are sometimes poorer compared to that of their non-deaf peers. This leads to disparities in their professional and social standing. The CC gives them the accessibility and resources to excel in both of these arenas. The user types an English sentence into the browser window and the system immediately gives an analysis of grammatical errors and corrections in Standard English. Other programs, like the ICICLE and the Microsoft Grammar Checker, facilitate grammar support to the deaf, however they are not as exhaustive and effective as the CC.

The creators of the CC have been diligent in the selection of their source data. It was made sure that the algorithms were developed based on data representative of the eventual users. To this end, a special data collection tool was devised, through which 8233 sentences and an additional 1239 sentences in a free essay task were harvested. This has made the CC the only system with a unique body of data in the textual language arena, specifically for the deaf.

The basic premise of CC’s design is to take into consideration the “non-standard patterns” of English grammar used by the deaf, instigated by the influence of sign language. These deviations from the Standard English rules are considered as “errors” in CC and the algorithms are based on an analysis of the errors from the data collected from deaf students. These algorithms use natural language understanding technology to parse a sentence, detect error patterns and map the errors to correct Standard English patterns.

The CC detects errors involving subject-verb agreement, verb conjugation errors and auxiliary-verb agreement, errors involving infinitives, and nominal errors including qualifier-noun agreement. The CC, besides being a parsing engine, is equipped with a 4000-word dictionary, a database, an easy to use online graphic interface and a statistical platform that calculates word and phrasal probabilities.

A beta-tested version of CC is being considered for introduction to American schools with sizable population of deaf students. The developers are looking at introducing CC as a tool for deaf students in high schools and colleges. It will be adapted as a pedagogical tool for ESL students in universities, also as a tool of document correction for various businesses and government functions. Spanish, Chinese and Japanese are huge foreign language markets for the CC, and many other avenues are opening up.