Helping Aphasic People Process Online Information
Siobhan Devlin, Gary Unthank · 2006 · Proceedings of the 8th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '06) · doi:10.1145/1168987.1169027
Summary
This short paper from the University of Sunderland describes the HAPPI (Helping Aphasic People Process Information) project, a web-based text simplification system designed to help people with aphasia access online content such as news stories. Aphasia is a language disorder most commonly caused by stroke (cerebrovascular accident), affecting an estimated 130,000 people annually in England and Wales alone, with 120,000 being post-retirement age. The condition affects speech, reading, writing, and comprehension in varying ways across subtypes (Wernicke's, Broca's, Global aphasia). HAPPI specifically targets alexia — the reading difficulty associated with aphasia — by simplifying text to improve comprehension. The system works by "jogging the memory": when a user encounters a difficult word, the system offers a simpler synonym selected from psycholinguistic databases that contain word frequency, familiarity, and age-of-acquisition data. The prototype was built using existing NLP tools including the LT CHUNK Part of Speech Tagger, a MySQL port of Princeton's WordNet, and the Irvine Phonotactic Dictionary, and was tested with local newspaper content from North East England.
Key findings
The HAPPI prototype takes three input parameters controlling the number of synonyms considered when querying WordNet, allowing tuning of simplification aggressiveness. The system builds on two prior research projects: Devlin's PhD work on automatic lexical simplification (replacing difficult words with simpler synonyms) and the PSET (Practical Simplification of English Text) project that explored syntactic simplification using full sentence parsers. The key advance enabling HAPPI was that NLP tools previously available only for academic research had become freely available in formats suitable for web deployment — particularly WordNet being ported to MySQL, enabling a PHP-based web application without expensive local installation. Future directions include word sense disambiguation (to select correct synonyms), named entity recognition, morphological analysis, image retrieval to illustrate difficult concepts, allowing users to select which words to simplify, and incorporating psycholinguistic measures of age-of-acquisition and imageability. The researchers also plan to investigate allowing users to click difficult words to see images or hear pronunciation — a multimedia approach informed by the fact that aphasia affects individuals differently day to day.
Relevance
This paper addresses a significant accessibility gap: while much web accessibility work focuses on visual, motor, and sensory disabilities, people with acquired language disorders like aphasia face profound barriers to accessing text-based online content that are not addressed by standard accessibility guidelines. The prevalence of aphasia (affecting about a third of stroke survivors, with stroke being a leading cause of disability) makes this a substantial population. The HAPPI approach of automated lexical simplification — replacing difficult words with simpler synonyms based on psycholinguistic properties — represents an early example of NLP-powered cognitive accessibility that has become increasingly feasible with modern language models. The insight that the system should provide multiple support strategies (simpler words, images, audio pronunciation) rather than a single simplification approach reflects the reality that aphasia manifests differently across individuals and even fluctuates day to day within the same person. For accessibility practitioners, this work highlights that cognitive accessibility for acquired language disorders requires different strategies than accessibility for developmental cognitive disabilities or low literacy.
Tags: aphasia · text simplification · cognitive accessibility · lexical simplification · natural language processing · stroke · reading comprehension · alexia