How to restrict number of words to be retrieved to a limit 50 using filter attribute?

Question

{
    "retrievalQuery": {
        "text": "What is aws?"
    },
    "retrievalConfiguration": {
        "vectorSearchConfiguration": {
            "numberOfResults": 1
            }
        }
    }
}

Answer

Hi Guru,

I apologize for responding to your post so late. I've been researching how to restrict the number of words retrieved to a maximum of 50 using the filter attribute. Unfortunately, Amazon Bedrock doesn't natively support truncating text responses based on word count using metadata attributes. Therefore, you would need to manage this requirement through:

1. Calculate the `numberOfWords` for each document during pre-processing. Include this custom metadata in the document. For example, you can add `"numberOfWords": 76` to indicate the total word count.
2. [Amazon Bedrock's filtering capabilities](https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-ds.html#kb-ds-metadata) enable you to filter based on numerical comparisons in metadata values.
3. You can query using a filter based on `numberOfWords < 50` to include the documents that match this attribute.

> ⚠️ I haven't tested this solution, but it may work.

---

> ℹ️ However, I believe preprocessing will be necessary to achieve this.

How to restrict number of words to be retrieved to a limit 50 using filter attribute?

Contenuto pertinente