Home > Events > CLIP Colloquium: Tal Linzen (Johns Hopkins)

CLIP Colloquium: Tal Linzen (Johns Hopkins)

Time: 
Wednesday, April 11, 2018 - 11:00 AM to 12:00 PM
Location: 
3258 A.V. Williams Building (AVW)

Title: On the syntactic abilities of recurrent neural networks

Abstract: Recent technological advances have made it possible to train recurrent neural networks (RNNs) on a much larger scale than before. These networks have proved effective in applications such as machine translation and speech recognition. These engineering advances are surprising from a cognitive point of view: RNNs do not have the kind of explicit structural representations that are typically thought to be necessarily for syntactic processing. In this talk, I will discuss studies that go beyond standard engineering benchmarks and examine the syntactic capabilities of contemporary RNNs using established cognitive and linguistic diagnostics. These studies show that RNNs are able to compute agreement relations with considerable success across languages, although their error rate increases in complex sentences. A comparison of the detailed pattern of agreement errors made by RNNs to those made by humans in a behavioral experiment reveals some similarities (attraction errors, number asymmetry) but also some differences (relative clause modifiers increase the probability of attraction errors in RNNs but decrease it in humans). Overall, RNNs can learn to exhibit sophisticated syntactic behavior despite the lack of an explicit hierarchical bias, but their behavior differs from humans in important ways.

Bio: Human sentence comprehension is remarkably effective: we can reconstruct the structure of a sentence and extract its meaning with little perceptible delay.  Tal Linzen studies the representations and processes that make this feat possible, using human experiments (behavioral and neural) as well as computational simulations. He is also interested in using ideas from psycholinguistics to understand the strengths and limitations of artificial intelligence systems (in particular, artificial neural networks), with the goal of bringing their linguistic abilities closer to human levels.

Before joining Johns Hopkins, he was a postdoctoral researcher at Ecole Normale Supérieure in Paris. His degrees are a B.Sc. in Mathematics and Linguistics and an M.A. in Linguistics, both from Tel Aviv University, and a Ph.D. in Linguistics from New York University.