19 June 2023
First, we perform focused web crawling for domain-dependent corpus expansion and implicitly supply context by restricting the vocabulary. Second, we retrofit static word embeddings to so-called sense embeddings to model the multiple senses a word can have. Finally, we analyze the compositional abilities of contextualized word embeddings and use them for semantic downstream tasks
To address the ambiguous nature of language, we explore various techniques to incorporate context at different levels.
Date and Time: 19 June 2023, 14:00 CET.
Room G-021/022, Informatikum Campus, Vogt-Kölln-Str. 30, 22527 Hamburg.