Creating artificial intelligence solutions in E-health infrastructure to support disabled people

Abstract

Recently, the creation of a barrier-free environment for disabled people is becoming more and more important. All this is done so that people do not feel difficulties in filing their ordinary needs, including communication. For this purpose, a communicator application was developed that allows communication using card-pictograms for people with speech and writing disorders, particularly people with ASD. According to the US National Center for Health Statistics and the Health Resources and Services Administration, in 2011–2012 Autism was detected in 2% of schoolchildren worldwide, and this problem is very relevant. This article discusses several approaches of using Artificial Intelligence to simplify text typing with pictogram based cards by predictive input, which allows users faster compose messages and simplify communication process. A tool for analyzing the texts semantics - Word2Vec, was used, which is a neural network of direct distribution. Two approaches are considered: Continuous Bag of Words and Skip-gram. Also quality measures of advisory systems were used, and an approach giving the best results was identified. Besides that, quality measurements were carried out to identify optimal solutions of sentiment analysis to automatically detect suspicious messages sent by the users with such disabilities, which will help doctors to enhance their capabilities of monitoring and behavioral control and take appropriate actions if undesirable behavior of patient is detected by the system.