Minnesota Natural Language Processing Seminar Series: Explanations in Natural Language: From Theory to Practice

The Minnesota Natural Language Processing (NLP) Seminar is a venue for faculty, postdocs, students, and anyone else interested in theoretical, computational, and human-centric aspects of natural language processing to exchange ideas and foster collaboration. The talks are every other Friday from 12 p.m. - 1 p.m. during the Fall 2021 semester.

This week's speaker, Dheeraj Rajagopal (Carnegie Mellon University), will be giving a talk titled "Explanations in Natural Language: From Theory to Practice."

Abstract

Understanding the reasoning process through explanations is spontaneous, ubiquitous and fundamental to our sense of perceiving the world around us. Scientific progress often relies on explanations to facilitate discovery of hypotheses, identify applications and also identify systematic errors and correct them. An in-depth study of explanations thus helps shed light on core cognitive issues, such as learning, induction and conceptual representation. Current NLP systems, despite significant advances, are usually treated as black boxes with little to no insight into how they reason. Understanding Explanations is an under-explored area in the natural language processing literature due to the lack of a unified formalism. In this talk, we will discuss two kinds of approaches to explanation - data-based and model-based. Next, we present methods for learning explanations via data (both statically and dynamically). We will also discuss an instantiation of model-based explanation approach where we explore how to explain model predictions while also jointly optimizing for the end-task. Finally, we discuss the learnings from the work and how to make the bridge between theory and practice for explanations in NLP.

Biography

Dheeraj Rajagopal is a 5th year PhD student at Carnegie Mellon University, School of Computing Sciences and advised by Prof. Eduard Hovy. His research interests are in the area of explanations, and designing explainable systems for Natural Language Processing. He has been an intern previously at MSR with the Knowledge Technologies and Experience team and at Ai2 with the Aristo team. Previously, he was a masters student at CMU working with Prof. Eduard Hovy and Prof. Teruko Mitamura.