For LtR features, It should be possible to fetch the, i.e. BM25 and language model, scores from ElasticSearch without needing to change a lot of code, which would be error prone.
Add means to retrieve the scores before LtR training/ranking in a systematic way.
For LtR features, It should be possible to fetch the, i.e. BM25 and language model, scores from ElasticSearch without needing to change a lot of code, which would be error prone.
Add means to retrieve the scores before LtR training/ranking in a systematic way.