What?

We are a team of researchers participating in the six-week JSALT 2018 in-person intensive research summer camp at the CLSP at Johns Hopkins University.

Research proposal: Our group will work to improve the ability of artificial neural network models to represent and understand sentence meaning. Candidate systems will be evaluated on their ability to solve a suite language understanding tasks using a single shared sentence encoder component, combined with a set of ten lightweight task-specific attention models. We expect to explore new methods for unsupervised learning, new objectives and techniques for multi-task and transfer learning, and new methods for linguistically sophisticated error analysis.

We use baselines and metrics from the GLUE benchmark suite.

Because we study ELMo, and because JSALT provides cookies generously, our mascot is Cookie Monster.

Results

• Can You Tell Me How to Get Past Sesame Street? Sentence-Level Pretraining Beyond Language Modeling (ACL 2019, earlier version ‘Looking for ELMo’s Friends’ here)
• What do you learn from context? Probing for sentence structure in contextualized word representations (ICLR 2019)
• Probing What Different NLP Tasks Teach Machines about Function Word Comprehension (*SEM 2019)
• The jiant software toolkit.
• This terrible group photo.

The Team

Team Leaders

• Sam Bowman
• Ellie Pavlick

Senior Researchers

• Ben Van Durme
• Edouard Grave
• Yinghui Huang
• Tal Linzen
• Ian Tenney
• Katherin Yu

Graduate Student Researchers

• Jan Hula
• Najoung Kim
• Tom McCoy
• Raghu Pappagari
• Roma Patel
• Adam Poliak
• Alex Wang
• Patrick Xia

Undergraduate Student Researchers

• Berlin Chen
• Shuning Jin
• Alexis Ross

Advisors and Consulting Participants

• Dipanjan Das
• Allyson Ettinger
• Felix Hill
• Jamie Kiros
• Omer Levy
• Ves Stoyanov