Robotics Colloquium: Human and Data in the loop of NLP Pipeline

This week's speaker, Dongyeop Kang, will be giving a talk titled "Human and Data in the loop of NLP Pipeline."

Abstract

NLP systems trained on standard machine learning pipelines; annotation, learning, and evaluation, are limited to causing various problems; for instance, the dataset collected from crowd workers often contains annotation artifacts or repeating patterns; as the systems are deployed to real-world users, they are not well controlled, interpreted, or interacted with real users. To address these problems caused by the ML pipeline, I will discuss recent work from the Minnesota NLP group on human-centric and data-centric approaches. For the human-centric aspect, we collect human perception on linguistic styles and then make the model to mimic how humans perceive styles. Then we develop interactive NLP systems that help scholars better read and write academic papers. In the data-centric NLP, we model data informativeness based on various training dynamics and then use them to find new important data points for data augmentation and annotation. We believe more involvement of humans and consideration of data dynamics transforms the traditional ML-driven NLP pipeline to be more robust, interactive, and information-effective.

Biography

Dongyeop Kang is an assistant professor in the Computer Science Engineering department at the University of Minnesota, Twin Cities. He leads the Minnesota Natural Language Processing (NLP) group that aims to develop human-centered language technologies. His group's research lies at the intersection of computational linguistics, machine learning, and human-computer interaction.

He completed a postdoc at the University of California, Berkeley, and obtained a Ph.D. in the Language Technologies Institute of the School of Computer Science at Carnegie Mellon University. During his Ph.D. study, he interned at Facebook AI research, Allen Institute for AI (AI2), and Microsoft Research. He has been awarded the AI2 fellowship, CMU Presidential fellowship, and ILJU Ph.D. fellowship.