[Defense] NLP-enabled Design Assistance For Visual Communication
Friday, May 7, 2021
10:00 am - 12:00 pm
In
Partial
Fulfillment
of
the
Requirements
for
the
Degree
of
Doctor
of
Philosophy
Amirreza
Shirani
will
defend
his
dissertation
NLP-enabled
Design
Assistance
For
Visual
Communication
Abstract
In visual communication, a wide range of design components are typically used to increase the comprehension of content and to convey the author’s intent. Different authoring and graphic design applications perform automatic design assistance that include images and text in different forms and shapes. However, a majority of publicly available tools are mainly driven by some basic heuristics in assisting users during authoring. Prior research has begun to use Artificial Intelligence (AI) to provide users with interface suggestions. Considering a wide range of applications and its unique challenges, this interdisciplinary area has not been fully studied and has little cross-disciplinary collaboration.
In this dissertation, we aim at advancing new technology to employ AI-based models to assist users during authoring by recommending appropriate design components based on the content. In particular, the first part of this dissertation focuses on the task of emphasis selection, i.e., choosing candidates for emphasis, where we propose label distribution learning methods to capture the ambiguity of the input. These models are designed to comprehend the most common interpretation of a written text, so the right emphasis can be achieved automatically or interactively.
In the second part of this dissertation, we focus on the task of font recommendation, i.e., suggesting fonts based on input text, where we model the associations between visual font attributes and textual context, with the final goal of better font recommendation during text composition. Specifically, we propose different end-to-end models that exploit contextual and emotional representations of the input text to recommend fonts. We introduce a total of three new language resources for both tasks of emphasis selection and font recommendation. We focus on social media data as well as presentation slides, and by examining the challenges of the language resources, we provide different analysis components. Besides a wide range of applications, the methods discussed in this dissertation are meant to benefit relevant tasks such as machine-human interaction, reading comprehension, graphic design, and user experience.
Friday,
May
7,
2021
10:00AM
-
12:00PM
CT
Online
via
Zoom
Dr. Thamar Solorio, dissertation advisor
Faculty, students and the general public are invited.
