Create README.md
Browse files# Albert-XXL-v1 trained on sentences of SQuAD 2.0
## Objective
The model was trained as a sentence pair classification model in order to be used as a re-ranker in a QA pipeline.
## Data
The data was taken from SQuAD 2.0. The texts were sentence-split and the sentence containing the answer + the question were taken as a positive example (full cross-encoder style). The sentences that didn't contain the answer + question were used as hard negatives, since they were much related to the question.