- April 19, 2022 This AmericasNLP2022 webpage is online!
Even though deep learning models have resulted in huge progress in the field of natural language processing (NLP), this progress is frequently limited to a tiny set of so-called high-resource languages, for which large amount of unlabeled as well as annotated text data are available. Machine translation (MT), which arguably is the "holy grail" of NLP is not exception to this: Translations between high- resource languages are easily understandable, while translations between under-resourced languages or an under-resourced and a high-resource language – such as English – are often bad or outright unusable. The severity of this problem becomes obvious when one considers that around 7000 languages are spoken on Earth, but Google Translate, which is commonly considered to be highly multilingual, supported only 108 of them as of February 2020.
- TBA
