Our submission to Poleval 2019 won 1st price in hate speech detection. That will be the second time in the row! The competition was strong, and congratulations to them. It is nice to see that simpler and faster architecture like ULMFiT beat huge models like BERT.
To answer this question, let's take an example. We all learned to recognize colours by being introduced to the objects in those colours first. Strawberry, tomato, pepper and fire truck were probably enough to understand the concept of redness. Once we catch what the red colour is we can correctly identify it on flowers, cars, abstract paintings and also on many real-world objects that we’ve never come across before. This is transfer learning.
We had a pleasure to win the first prize in Poleval 2018 for language modeling task. This success has largely resulted from the adaptation we did to ULMFiT architecture by Jeremy Howard and Sebastian Ruder. Below you can find a short presentation pointing the recent changes to the Language Modeling, especially the crucial improvement of polish language model and n-waves contribution to this:)
Recently our company, n-waves, has won Poleval, national NLP competition. I’d like to share with you some resource without which it wouldn’t have been possible: the fastai library, fast.ai’s courses, and the community around them. If you’re struggling with finding an acceptable service to address your automation needs or if you are interested in applying recent breakthroughs in artificial intelligence, then these resources can also work for you.