Download PDF

Workshop on Improving Social Inclusion using NLP: Tools and Resources (ISI-NLP), Date: 2016/05/23 - 2016/05/23, Location: Portorož, Slovenia

Publication date: 2016-05-01
Pages: 11 - 19
ISSN: 978-2951740891
Publisher: Paris; ELRA

Proceedings of 1st Workshop on Improving Social Inclusion using NLP: Tools and Resources

Author:

Sevens, Leen
Vanallemeersch, Tom ; Schuurman, Ineke ; Vandeghinste, Vincent ; Van Eynde, Frank ; Schuurman, Ineke ; Vandeghinste, Vincent ; Saggion, Horacio

Keywords:

Automated Spelling Correction,, Intellectual Disabilities, Pictograph Translation, Alternative and Augmentative Communication

Abstract:

We present the first version of an automated spelling correction system for Dutch Internet users with Intellectual Disabilities (ID). The normalization of ill-formed messages is an important preprocessing step before any conventional Natural Language Processing (NLP) process can be applied. As such, we describe the effects of automated correction of Dutch ID text within the larger framework of a Text-to-Pictograph translation system. The present study consists of two main parts. First, we thoroughly analyze email messages that have been written by users with cognitive disabilities in order to gain insights on how to develop solutions that are specifically tailored to their needs. We then present a new, generally applicable approach toward context-sensitive spelling correction, based on character-level fuzzy matching techniques. The resulting system shows significant improvements, although further research is still needed.