Welcome to C.Psyd

We are looking for passionate new PhD students, Postdocs, and Master students to join the team (more info) !

We are a linguistics research group at Cornell University. We are interested in the intersection of computational linguistics and psycholinguistics. By building computational models to predict human language processing behavior (e.g., reading times), we can study the linguistic features that impact human processing decisions. Relatedly, C.Psyd members use psycholinguistic techniques to study the strategies used by neural networks to produce high accuracy in different language contexts, which gives us insights as to when different strategies might be employed by humans.

We are affiliated with the Cornell Computational Linguistics Lab and with the Cornell Natural Language Processing Group.

News

26. Aug 2021

Timkey and van Schijndel (2021; EMNLP) shows that cosine similarity doesn't work for Transformer models. We introduce a simple method to correct the issue without retraining.

25. June 2021

Paper published in Cognitive Science!
Surprisal can only explain the existence of garden path effects in reading times, not the magnitude of the effects themselves.

10. May 2021

2 papers accepted at ACL and ACL Findings:
1) Davis and van Schijndel (2021) shows that linguistic knowledge in language models can be modeled as constraints.
2) Wilber et al. (2021) shows that abstractive summarization is extremely shallow at present, often simply emulating extractive summarization.

... see all News