A Study Exploring the Use of Comparative Terms in Image Understanding Wins Best Paper Award
"Lighter Can Still Be Dark: Modeling Comparative Color Descriptions” won one of the best short paper awards at the Association for Computational Linguistics 2018 conference.
The paper, by Columbia graduate student Olivia Winn, presents a method for modeling comparative terms, such as brighter or deeper, as quantitative changes in color. Winn merges Natural Language Processing and Computer Vision techniques in her research.
Her study, advised by Smaranda Muresan, a Research Scientist at the Data Science Institute, is the first to explore the use of comparative terms in image understanding, taking the first step towards incorporating descriptions of comparisons between objects into object-recognition methods.
You can view her presentation here: https://bit.ly/2QnFj47.
By Robert Florida