University Calendar

Tuesday, November 23, 2021 1-2PM  
Andrey Kutuzov, University of Oslo
Grammatical Profiling for Semantic Change Detection

https://montclair.zoom.us/j/88518888648?pwd=VGRwa1hyRUtSd2pLMHZvUlNUcnJlZz09

Semantics, morphology and syntax are strongly interdependent. However, the majority of computational methods for semantic change detection use distributional word representations which encode mostly semantics. We investigate an alternative method, grammatical profiling, based entirely on changes in the morphosyntactic behaviour of words. We demonstrate that it can be used for semantic change detection and even outperforms some distributional semantic methods. We present an in-depth qualitative and quantitative analysis of the predictions made by our grammatical profiling system, showing that they are plausible and interpretable.

Bio: Andrey Kutuzov is a postdoctoral fellow in Natural Language Processing at the University of Oslo, Norway. He obtained a PhD in NLP from the University of Oslo in 2020, with a thesis on modeling diachronic semantic change with distributional word embeddings. Computational detection of semantic change is his main research interest, and he published extensively on that in top-tier NLP venues like CoNLL, EMNLP, COLING and SemEval, as well as in several ACL workshops. In 2021, he organized RuShiftEval, the first shared task on semantic change detection for Russian.