AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
How to Deal With Quirks and Anomalies in NLP
When you're able to kind of land on accuracy, it sort of doesn't matter if you're sort of following the rules to get there. It's just that you found a way to translate. That's a great question. I think at least my work would definitely fall more into the second camp where you basically are working on building a very abstract, very large pattern recognizer and don't worry too much about the pre-processing. But that still doesn't mean that the data landscape isn't important. Even at Hugging Faces, we have almost half the team is working on this data sets project, which is to make it easier to find an LP data set.