[Colloq] PhD Defense - Karl Wiegand - Fri, Dec 5 @ 3:30pm - 366 WVH - Disambiguation of Imprecise User Input Through Intelligent Assistive Communication

Fong, Andy a.fong at neu.edu
Mon Dec 1 15:58:39 EST 2014


Speaker: Karl Wiegand



Date: Friday, December 5 @ 3:30 pm EST

Location: 366 WVH



Title: Disambiguation of Imprecise User Input Through Intelligent Assistive Communication



Link: http://www.karlwiegand.com/defense



Abstract:



Augmentative and alternative communication (AAC) systems are used by people with speech impairments severe enough to preclude the use of spoken communication. While communication systems for non-disabled users often implement intelligent prediction, correction, and behavior adaption, current AAC systems are relatively passive conduits for translating user intentions into spoken output. This dissertation seeks to shift the burden of communication from the user to the system by leveraging knowledge of the user's abilities, usage patterns, and contextual needs. The ultimate goal is to create an assistive communication prosthesis that enables users to seamlessly engage in timely and meaningful interactions. Towards that end, this dissertation makes the following contributions in the areas of natural language processing (NLP) and human-computer interaction (HCI):



   * A word-level language model - semantic grams - that bridges the gap between syntax and semantics by leveraging an author's own syntactic delimiters of semantic content. This model is more effective than similar n-gram-based language models for prediction tasks with unusual ordering or syntax.



   * An empirical comparison of contextual language predictors, showing that the use of statistics from a global corpus, such as the New York Times, is sub-optimal. Instead, situational context can provide more accurate background probabilities for pervasive speech and language processing tasks.



   * Results and observations from a touchscreen tablet study with current and potential AAC users, quantifying the challenges faced by people with upper limb motor impairments and showing how they can be addressed through intelligent interfaces.



   * Three user-driven interface designs and prototypes, including an approach to icon-based AAC that can be controlled effectively with a single input signal and leverages semantic frames to accommodate different screen sizes and user abilities.



Committee:



Rupal Patel (Advisor)

Javed Aslam

David Smith

Shaun Kane, University of Colorado Boulder (External Member)


Andrew W. Fong
Assistant Director for Graduate Admissions and Enrollment

Northeastern University
College of Computer and Information Science
360 Huntington Avenue
202 West Village H
Boston, MA 02115
617-373-8493
a.fong at neu.edu

Follow us on Twitter - @CCISGrad<https://twitter.com/CCISGrad>
Like us on Facebook - CCIS Graduate School<https://www.facebook.com/CCISGradSchool?ref_type=bookmark>



More information about the Colloq mailing list