For best experience please turn on javascript and use a modern browser!
You are using a browser that is no longer supported by Microsoft. Please upgrade your browser. The site may not present itself correctly if you continue browsing.

MSc AI student Anna Langedijk and former MSc AI student Verna Dankers of the Informatics Institute have won a best paper award at the Computational Conference for Natural Language Learning (CoNLL) with an analysis of how neural networks learn and process language.

Langedijk and Dankers presented in their paper titled 'Generalising to German Plural Noun Classes, from the persepective of a Recurrent Neural Network' a meticulous analysis of how neural networks acquire the complex German plural system. This is a task historically related to broader questions about generalisation in language and the viability of neural networks as cognitive models of language. They used a variety of different techniques, some of which originally proposed at the University of Amsterdam, such as behavioural analysis, diagnostic classification, representational analysis and causal intervention.

The project was a collaboration with the University of Edinburgh and Facebook AI Research, and was led by UvA alumnus and former employee Dieuwke Hupkes.