Luke Zettlemoyer

Bio

I am an Associate Professor in the Allen School of Computer Science & Engineering at the University of Washington, and also lead the AllenNLP project at the Allen Institute for Artificial Intelligence. I am also a PECASE Awardee and an Allen Distinguished Investigator. Previously, I did postdoctoral research at the University of Edinburgh and was a Ph.D. student at MIT.

Research

My current research is in the intersections of natural language processing, machine learning, and decision making under uncertainty. I am particularly interested in designing learning algorithms for recovering representations of the meaning of natural language text. For more research details, please see my publications.

Recent Teaching

• Spring 2016: Introduction to Artificial Intelligence (CSE 473)
• Winter 2016: Artificial Intelligence (CSEP 573)
• Winter 2015: Machine Learning (CSE 446)

Students (Alumni)

• Terra Blevins
• Eunsol Choi (Co-advised with Yejin Choi)
• Nicholas FitzGerald
• Luheng He
• Srinivasan Iyer (Co-advised with Alvin Cheung)
• Mandar Joshi (Co-advised with Dan Weld)
• Kenton Lee
• Victoria Lin
• Julian Michael

Postdoc

• Omer Levy

Announcements

• We launched AllenNLP, a new open-source NLP research library built on PyTorch. AllenNLP includes reference models for an (ever growing) range of NLP tasks and infrastructure to make it easy to develop new deep learning approaches. Please give it a try!
• Congratulations to Mark for winning a best paper award for Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints at EMNLP 2017! Although I wasn't involved in the work, I am excited to help publicize Mark's contributions to this important research!
• Kenton released a new neural coreference resolution model that is state of the art and uses no preprocessing. Check out the demo!
• Luheng released DeepSRL, an end-to-end neural SRL system that performs 4 F1 points better than any other publicly released PropBank model (a nearly 20% error reduction).
• Srini released code for interactive learning of neural semantic parsers that map from English directly to SQL. Now you can build and deploy a semantic parser that improves performance on its own over time, with just a little crowd sourcing!
• Yannis released his Neural AMR code, with models for mapping AMR to English and back. It includes state-of-the-art generation results and the best neural parser to date!
• Omer released his zero-shot relation extraction data and code. This work provides a new way of authoring relation extraction systems, where engineers simply specify natural language questions they want answered about target relationships!
• Congratulations to Kenton and Mike for winning a best paper award for Global Neural CCG Parsing with Optimality Guarantees at EMNLP 2016!
• Mark released ImSitu, a large, new situation recognition dataset. He has formulated this recognition problem as visual semantic role labeling, where images are mapped to structured action representations. Have a look at the data and recognition demo!
• Luheng released a QA-SRL dataset, demonstrating that non-expert annotators can provide supervision for recovering the predicative argument structure of English sentences.
• Congratulations to Yoav and Kenton for winning a best paper award for Broad-coverage CCG Semantic Parsing with AMR at EMNLP 2015!