Blackbox@NL

Netherlands workshop on understanding neural networks in NLP

Venue

Jheronimus Academy of Data Science, s’Hertogenbosch, 6 May 2019

Program

Morning

• 9:50 - 10:00 Introduction
• 10:00-10:25 Grzegorz Chrupała & Afra Alishahi: Correlating neural and symbolic representations of language
• 10:25-10:50 Dieuwke Hupkes: The compositionality of neural networks: integrating symbolism and connectionism
• 10:50-11:10 Coffee break
• 11:10-11:35 Samira Abnar & Jelle Zuidema: Representational Similarity & Stability Analysis of Neural Language Models and Brains
• 11:35-12:15 Shared task discussion
• 12:15-14:00 Lunch

Afternoon

• 14:00-14:25 Arianna Bisazza: Does Syntactic Knowledge Transfer Cross-Lingually in Multilingual Neural Language Models?
• 14:25-14:50 Madhumita Sushil: Synthetic dataset for explaining and evaluating rules learned by RNNs
• 14:50-15:10 Coffee break
• 15:10-15:35 Mathijs Mul & Jelle Zuidema: Siamese recurrent networks can learn first-order logic reasoning and exhibit zero-shot generalization to novel expressions
• 15:35-16:00 Sandro Pezzelle & Raquel Fernández: Exploring Models’ Abilities in Grounding Vague Expressions

Organizers

• Afra Alishahi, Tilburg University
• Grzegorz Chrupała, Tilburg University

Sponsors

Sponsored by SIKS