3 datasets found

The dataset used in the paper is not explicitly described, but it is mentioned that the authors used three downstream tasks from the GLUE benchmark: Stanford Sentiment Treebank...

This paper proposes BERT, a pre-trained deep bidirectional transformer for language understanding.

Pre-trained language models (PrLM) have to carefully manage input units when training on a very large text with a vocabulary consisting of millions of words. Previous works have...