# Compare a vector to a list of vectors contained in each document

**URL:** https://discuss.elastic.co/t/compare-a-vector-to-a-list-of-vectors-contained-in-each-document/263905
**Category:** Elasticsearch
**Created:** [February 10, 2021, 4:55pm UTC](https://discuss.elastic.co/t/compare-a-vector-to-a-list-of-vectors-contained-in-each-document/263905 "2021-02-10T16:55:13Z")
**Posts on this page:** 3
**Page:** 1

<div class="post-metadata">

### Author: ![flavienbwk](https://avatars.discourse-cdn.com/v4/letter/f/5f9b8f/32.png) [@flavienbwk](https://discuss.elastic.co/u/flavienbwk)
#### Post date: [February 10, 2021, 4:55pm UTC](https://discuss.elastic.co/t/compare-a-vector-to-a-list-of-vectors-contained-in-each-document/263905/1 "2021-02-10T16:55:13Z")

</div>

Hi,

Each of my documents contain a list of dense\_vectors inside a nested object : `words.vector`

I try to compare my input vector to all documents' `words.vector` properties.

I currently try with the following but elasticsearch tells me `script_score function must not produce negative scores...` :

```auto
GET test_words/_search
{
  "query": {
    "nested": {
      "path": "words",
      "score_mode": "min",
      "query": {
        "function_score": {
          "script_score": {
            "script": {
                "source": "cosineSimilarity(params.myvector, 'words.vector')",
                "params": {
                    "myvector": [
			-1.310115814,
			-0.399771153,
			-0.420954197,
			1.589795112
		  ]
              }
            }
          }
        }
      }
    }
  }
}

```

How could I achieve that ?

Thank you,

---

<div class="post-metadata">

### Author: ![mayya](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/mayya/32/83147_2.png) [@mayya](https://discuss.elastic.co/u/mayya)
#### Post date: [February 11, 2021, 11:04pm UTC](https://discuss.elastic.co/t/compare-a-vector-to-a-list-of-vectors-contained-in-each-document/263905/2 "2021-02-11T23:04:54Z")

</div>

Elasticsearch doesn't allow for documents to produce scores \< 0. As `cosineSimilarity` function returns values from `[-1;1]` to ensure non-negative scores, you need to add `1` to your function like this:  
`cosineSimilarity(params.myvector, 'words.vector') + 1`.

This is also explained in [our documentation](https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-script-score-query.html#vector-functions).

---

<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: [March 11, 2021, 11:05pm UTC](https://discuss.elastic.co/t/compare-a-vector-to-a-list-of-vectors-contained-in-each-document/263905/3 "2021-03-11T23:05:34Z")

</div>

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