Algorithm to support accurate efficient text generation within an onscreen keyboard app

Abstract The use of writing to interact with others is of critical importance to persons with severe physical impairments SPI and persons with severe speech and physical impairments SSPI For these people written communication tools such as email texting and social media play essential roles in supporting social closeness the exchange of information with others and exercising self determination Many people with SPI however experience difficulty both in generating and in editing text using traditional onscreen keyboards Our Phase research has two goals The first goal is to correct letter targeting and spelling errors when a person uses an alternative input method to write To accomplish this goal we will create and evaluate an input correction algorithm that recognizes input errors and may increase the writing rate of people with SPI who use scanning as their text input method Our second goal is to address text editing challenges when using scanning to correct composition errors e g inserting missing words To accomplish this goal we will use the language model within our onscreen keyboard to predict the location of composition errors within the writer s recent text When the writer indicates that they want to edit their text the onscreen keyboard will create an error prediction list that replaces the word prediction list available when the writer is composing text The error prediction algorithm will evaluate the text created by the writer and predict the five most likely locations of errors within the text When the writer selects one of the predictions the cursor will jump to the associated location within the text to ease the process of editing We will evaluate the effectiveness of the error correction algorithms three ways mathematically using an Alternating Treatment Design and through evaluation by assistive technology AT experts We will establish feasibility if our algorithms permit faster scanning rates correct a substantial number of errors and AT experts rate the technology as having important benefit Project Narrative The proposed onscreen keyboard app corrects the errors that occur when people with severe physical impairment generate written text The system may enable people with severe physical impairment to write more quickly efficiently and effectively