How to get the similarity score by documentDB

Question

I tried vector search on DocumentDB based on the reference below.
https://docs.aws.amazon.com/documentdb/latest/developerguide/vector-search.html

I was able to get the document, but I can't get the score.
I couldn't find any description of how to get the score in the reference either.
Could you tell me how to get the similarity score?
~~~
db.collection.aggregate([
  {
    $search: {
      "vectorSearch": {
        "vector": [0.2, 0.5, 0.8], 
        "path": "vectorEmbedding", 
        "similarity": "euclidean",
        "k": 2,
        "probes": 1
      }
    }
  }
]);

{ "_id" : ObjectId("653d835ff96bee02cad7323c"), "product_name" : "Product A", "vectorEmbedding" : [ 0.2, 0.5, 0.8 ] }
~~~

Answer

**DocumentDB**'s output for vector searches does not include similarity scores, unlike **MongoDB**. To obtain these scores, you can calculate them manually using a function like the following:

```python
import math

def euclidean_distance(vec1, vec2):
    return math.sqrt(sum((a - b) ** 2 for a, b in zip(vec1, vec2)))
```

> This function calculates the Euclidean distance between two vectors, which can be used as a measure of similarity. For example:

```python
query_vector = [0.2, 0.5, 0.8]
document_vector = [0.2, 0.5, 0.8]  # Example vector from a returned document

score = euclidean_distance(query_vector, document_vector)
```

> In this case, the `score` value represents the similarity between the query vector and a document's vector. Smaller score indicate higher similarity.

#### Resources:
- [Atlas Vector Search Score](https://www.mongodb.com/docs/atlas/atlas-vector-search/vector-search-stage/#atlas-vector-search-score)
- [Vector Search DocumentDB](https://docs.aws.amazon.com/documentdb/latest/developerguide/vector-search.html#w5aac21c11c15)

How to get the similarity score by documentDB

Resources:

相關內容