Modeling Syntax Acquisition with Cognitive Constraints
- Wednesday, January 29, 2020, 4:00pm-5:30pm
- Muenzinger D430
Children seem to learn their first languages effortlessly, but how they are able to do this has been heatedly debated for many years among linguists.
Recent advances in computational linguistics have presented us a unique opportunity to explore the problem of syntax acquisition with computational modeling. In this talk, I will first introduce syntax acquisition modeling, also known as grammar induction, from a theoretical perspective. I will then present our efforts in modeling syntax acquisition with statistical machine learning models with human memory constraints. Simulations using Bayesian and neural network models on natural data in many languages have provided insight into how language acquisition may happen without universals as inductive biases as well as how cognitive constraints may interact with syntax acquisition. Finally I will discuss some theoretical considerations and future directions for acquisition modeling.
Lifeng Jin is a Ph.D. candidate in Linguistics at the Ohio State University focusing on Computational Linguistics. His research interests include language acquisition modeling with unsupervised Bayesian and deep learning models, computational modeling for sentence processing, graph-based representation learning and NLP in education. He received his M.A. from the University of Sheffield in Intercultural Communication, and a B.A. in Chinese Pedagogy from Beijing Language and Culture University with a minor in Computer Science. He is a recipient of the Presidential Fellowship at OSU, as well as grants and awards from the Center for Cognitive and Brain Sciences and the Ohio Supercomputer Center. He also worked in industry to incorporate natural language processing into education.