Bert Base Mdoc Bm25

🚀 BERT Reranker for MS-MARCO Document Ranking

A text reranker trained on the MS MARCO document dataset for the BM25 retriever, enhancing retrieval performance.

🚀 Quick Start

To learn how to use this model, please visit our project repo page.

✨ Features

• Trained for BM25: Specifically trained for the BM25 retriever on the MS MARCO document dataset.
• Flexible Compatibility: Can work with other retrievers, but performs best with an aligned BM25 setup.

📚 Documentation

Model Description

A text reranker trained for the BM25 retriever on the MS MARCO document dataset.

Intended Uses & Limitations

It can be used with other retrievers, but using an aligned BM25 setup yields the best results.

We used the BM25 implementation from the Anserini toolkit and indexed with tuned parameters (k1=3.8, b=0.87) following this instruction.

Eval Results

MRR @10: 0.423 on the Dev set.

BibTeX entry and citation info

@inproceedings{gao2021lce,
               title={Rethink Training of BERT Rerankers in Multi-Stage Retrieval Pipeline}, 
               author={Luyu Gao and Zhuyun Dai and Jamie Callan},
               year={2021},
               booktitle={The 43rd European Conference On Information Retrieval (ECIR)},
      
}

📄 License

This project is licensed under the Apache-2.0 license.