license: apache-2.0 | |
This model receives scrambled or incoherent sentences as input and returns a meaningful sentence using the same words in the input . A form of grammar correction if you may . | |
It was trained on a dataset of permutated sentences derived from wikipedia pages as input with the correct arrangement of words as labels . | |
It is an encoder-decoder model that uses BERT's weight in both it's encoder and decoder . | |