README.md

Reranking

For reranking we use the model proposed by Nogueira and Cho [Nogueira2019] trained on MS-MARCO.

In the notebook Reranking with dl4marco-bert.ipynb we show the steps taken for the reranking phase:

1. converting the retrieval output into the format required by the model
2. converting the input files into TF records using the convert_CAST_to_tfrecord.sh script
3. run the model in inference over the test set using the test_CAST.sh script
4. use trec_eval for evalution

References

[Nogueira2019] Nogueira, R., & Cho, K. (2019). Passage Re-ranking with BERT. ArXiv, abs/1901.04085.