NLP Beyond the Top-100 Languages

titleNLP Beyond the Top-100 Languages
start_date2022/11/18
schedule11h
onlineyes
visiohttps://docs.google.com/document/d/1bnuIxo9F2WLPsdI9VIn233fMVwybFSUequrUh4vlYkk/edit?usp=sharing
location_infoOnline
summaryThe availability of large multilingual pre-trained language models has opened up exciting pathways for developing NLP technologies for languages with scarce resources. In this talk I will summarize some of my group's recent work on the challenges of handling new, unseen languages through finetuning, proposing a phylogeny-based adapter solution. Last, as data is paramount for extending into new languages, I will discuss issues relating to data requirements and data representativeness.
responsiblesBawden