Abstract:One in every 200 people worldwide cannot express orally because of cognitive, motor, neurological, or emotional problems. Assistive technologies can help people with impairments to use computers to perform their daily life activities independently and to communicate with others. This paper presents a Hidden Markov Model-based word prediction method that allows keyboard emulation software to predict words so that children with disabilities can type texts more quickly. The proposed system involved the development of a keyboard emulator, the construction and processing of a corpus, as well as a word prediction algorithm. Children with different cognitive profiles had to produce a text and type it twice: first with free typing, second using the virtual keyboard's word prediction. Results indicated the word prediction of the keyboard emulator software reduced typing efforts. However, the software initially increased the typing time when the corpus was not well adapted to users. The total amount of clicks with word prediction decreased by around 26.2%. Regarding execution time using prediction, 61% typed the text in less time. The tests performed with literate volunteers indicated a reduction in the number of clicks by up to 51.3%. This result surpasses the 15% achieved in the previous study by Free Virtual Keyboard with word prediction based on pure statistics. Moreover, all volunteers required fewer clicks to perform the task. People with impairments, especially children, could use the system and demonstrate their knowledge and abilities. The entire system is available on the Internet and users have unrestricted and free access to it.

Federated Learning for Mobile Keyboard Prediction

Federated Learning of N-gram Language Models

Optimize Scheduling of Federated Learning on Battery-powered Mobile Devices

Collaborative Learning Between Cloud and End Devices

Close the Gap Between Deep Learning and Mobile Intelligence by Incorporating Training in the Loop

Applied Federated Learning: Improving Google Keyboard Query Suggestions

Two Models are Better than One: Federated Learning Is Not Private For Google GBoard Next Word Prediction

Federated Learning Of Out-Of-Vocabulary Words

Communication-Efficient Learning of Deep Networks from Decentralized Data

Private Federated Learning in Gboard

Federated Evaluation of On-device Personalization

Federated Learning Based Multilingual Emoji Prediction In Clean and Attack Scenarios

Cross-device free-text keystroke dynamics authentication using federated learning

Pretraining Federated Text Models for Next Word Prediction

Federated Learning of Gboard Language Models with Differential Privacy

Federated Learning Based on Dynamic Regularization

DeepType

Federated Reconstruction: Partially Local Federated Learning

FLaaS: Cross-App On-device Federated Learning in Mobile Environments

Efficient Language Model Architectures for Differentially Private Federated Learning

Virtual keyboard with the prediction of words for children with cerebral palsy