# Hybrid Retrieval with approximate KNN

**URL:** <https://discuss.elastic.co/t/hybrid-retrieval-with-approximate-knn/329196>\
**Category:** Elasticsearch\
**Tags:** vector-search\
**Created:** [April 3, 2023, 9:50am UTC](https://discuss.elastic.co/t/hybrid-retrieval-with-approximate-knn/329196 "2023-04-03T09:50:20Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![jinmingteo](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/jinmingteo/32/119351_2.png) [@jinmingteo](https://discuss.elastic.co/u/jinmingteo)\
**Post date:** [April 3, 2023, 9:50am UTC](https://discuss.elastic.co/t/hybrid-retrieval-with-approximate-knn/329196/1 "2023-04-03T09:50:20Z")

</div>

Hi,

I have read through the following documentation: [k-nearest neighbor (kNN) search | Elasticsearch Guide [master] | Elastic](https://www.elastic.co/guide/en/elasticsearch/reference/master/knn-search.html#_combine_approximate_knn_with_other_features)

I have also experimented with a small database and it seems that the results are skewed towards the matching function (which is BM25 (?)).

Matching score seems to be in a much higher magnitude as compared to KNN score. This can be further validated using the Explain API. Thus, is there any normalisation done to the matching values before adding to KNN score? Or we have to tune the values through boosting?

---

<div class="post-metadata">

**Author:** ![BenTrent](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/bentrent/32/33915_2.png) [@BenTrent](https://discuss.elastic.co/u/BenTrent)\
**Post date:** [April 3, 2023, 12:50pm UTC](https://discuss.elastic.co/t/hybrid-retrieval-with-approximate-knn/329196/2 "2023-04-03T12:50:49Z")

</div>

Hey @jinmingteo ,

You must boost the KNN results to make them competitive with BM25.

Or you can inversely boost BM25 (boost by 0.6 or something).

We want to make this experience better. But for now, that is what you have to do.

---

<div class="post-metadata">

**Author:** ![system](https://us1.discourse-cdn.com/elastic/original/3X/1/a/1ac57faf039f6b580b3f104ef42a2a89e41014de.png) [@system](https://discuss.elastic.co/u/system)\
**Post date:** [May 1, 2023, 12:51pm UTC](https://discuss.elastic.co/t/hybrid-retrieval-with-approximate-knn/329196/3 "2023-05-01T12:51:15Z")

</div>

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.