MIS Speaker's Series - Peter Jansen

Event Date

Friday, October 19, 2018 -
1:00pm to 2:15pm

Location

Speaker: Peter Jansen, School of Information, University of Arizona

Presentation Title: What's in an Explanation?  Toward Explanation-centered Inference for Science Exams

Abstract: Modern question answering systems are able to provide answers to a set of common natural language questions, but their ability to answer complex questions, or provide compelling explanations or justifications for why their answers are correct is still quite limited. These limitations are major barriers in high-impact domains like science and medicine, where the cost of making errors is high, and user trust is paramount.  In this talk I'll discuss our recent work in developing systems that can build explanations to answer questions by aggregating information from multiple sources (sometimes called multi-hop inference).  Aggregating information is challenging, particularly as the amount of information becomes large due to "semantic drift", or the tendency for inference algorithms to quickly move off-topic when assembling long chains of knowledge.  Motivated by our earlier efforts in attempting to latently learn information aggregation for explanation generation (which is currently limited to short inference chains), I will discuss our current efforts to build a large corpus of detailed explanations expressed as lexically-connected explanation graphs to serve as training data for the multi-hop inference task.  We will discuss characterizing what's in a science exam explanation, difficulties and methods for large-scale construction of detailed explanation graphs, and the possibility of automatically extracting common explanatory patterns from corpora such as this to support building large explanations (i.e. six or more aggregated facts) for unseen questions through merging, adapting, and adding to known explanatory patterns.

Bio: Peter is a broadly interdisciplinary artificial intelligence researcher specializing in natural language processing and methods inspired by cognition and the brain. He applies these to application areas in science and health care.

A central focus of Peter's science research is on how we can teach computers question answering in the form of passing standardized science exams, as written. In particular, Peter focuses on methods of automated inference that generate explanations for why the answer is correct, largely using graph-based methods.

In terms of health care, Peter studies how we can use natural language processing and inference to improve electronic health records and improve nurse communication, as well as detect potentially dangerous clinical events before they happen.

Peter uniquely has two distinct educational backgrounds, one in natural language processing, cognition, and computer science, the other in physics, electrical engineering, and sensing. Peter maintains active outreach in grounding science education through sensing, largely in the form of open source hardware like the tricorder project, and projects like the open source computed tomography scanner. This work has been widely featured in over 50 international news media articles, including Reuters, Forbes, WIRED, MSNBC, and the Washington Post, as well as an invited talk at TEDxBrussels 2012.  In 2015, Peter's open source science tricorder was honoured by being placed on permanent exhibit at the German Museum of Technology in Berlin.