Updating score for BM25 batch search

[Fatima-Haouari]

Hi,
I indexed my own collection and I am using BM25 batch search for retrieval. If I want to update the BM25 scores such as multiplying the score by X before retrieving top N documents. Is there a way to achieve this?

[lintool]

Can you explain the use case? You can access Lucene queries directly in Pyserini as follows:
https://github.com/castorini/pyserini/blob/master/docs/usage-querybuilder.md

[Fatima-Haouari]

What I achieved is retrieving top 1000 using BM25 then multiplying the BM25 scores by X value relevant to my problem and reranked again the retrieved documents. What I am trying to achieve is applying my scoring function BM25 score*X while retrieving so the scoring is applied to all documents not only top 1000.

[lintool]

The standard solution is to retrieve top K (> 1000) and rerank those hits to obtain the top 1000. The only other option is the write a custom Lucene Similarity.

[Fatima-Haouari]

Thanks a lot for your quick response.