Boosting chat responses using metadata boosting - Amazon Q Business
Understanding boosting (Legacy)

Boosting chat responses using metadata boosting

Note

Relevance tuning has replaced metadata boosting. For more information, see Tuning the query results based on document attribute relevancy.

Important

This section assumes that you understand document attributes and metadata controls in Amazon Q Business.

Note

Before you configure relevance tuning, you must create a Amazon Q Business retriever and index for your application.

If you choose to use an Amazon Q Business native retriever, you can assign weights to document attributes after mapping them to Amazon Q Business index fields using the Amazon Q Business metadata boosting feature. Then, you can use these assigned weights to fine-tune the underlying ranking of RAG retrieved passages within your Amazon Q application environment to optimize the relevance of chat responses. In Amazon Q, boosting means to raise a document in chat results using these weights.

Important

Boosting document attributes using metadata boosting is an admin-only feature.

Boosting chat responses based on document attributes helps you rank sources that are more authoritative higher than other sources in your application environment. You can assign a higher value to more recent content, specific file types, or specific data sources.

Amazon Q Business automatically boosts specific document attributes, like document title, when retrieving information from your index to generate end user chat responses. You can use the boosting feature to customize and control boosting, and also override any pre-existing boosts applied by Amazon Q Business.

When you use this feature, a Retrieval Augmented Generation (RAG)-generated result is given a boost during retrieval when the query includes terms that match that field or attribute. You specify how much of a boost the document receives when there is a match. When Amazon Q Business generates responses, it prioritizes the sources that are assigned higher rankings.

Choosing to boost document attributes doesn't by itself cause Amazon Q Business to include or exclude a document in the chat response. A boosted document attribute is only one of the factors that Amazon Q Business uses to determine the relevance of a document.

Note

Boosting in Amazon Q Business is only available if you use an Amazon Q native retriever. If you use an Amazon Kendra retriever, you must configure boosting for document attributes in Amazon Kendra. Amazon Q Business supports any boosting that's already configured in your Amazon Kendra index.

Topics

Understanding boosting in Amazon Q Business

Note

Relevance tuning has replaced metadata boosting. For more information, see Tuning the query results based on document attribute relevancy.

To improve retrieved results and customize the end user chat experience, Amazon Q enables you to map attributes to fields in your Amazon Q Business index.

Amazon Q Business offers two kinds of attributes:

  • Reserved or default – Reserved attributes are based on document attributes that commonly occur in most data. You can use reserved attributes to map commonly occurring document attributes in your data to Amazon Q Business index fields.

  • Custom – You can create custom attributes to map document attributes that are unique to your data to Amazon Q Business index fields.

Document attributes can be mapped to index fields using either the Amazon Q console or the API:

  • Use the API – Before you use the API, you must first create an index. Next, create index fields. Then, to ingest documents into your Amazon Q Business index, use the CreateDataSource or BatchPutDocument API operations.

  • Use the console – You can choose to map document attributes directly to your index or from your data sources when you connect your data source to Amazon Q Business.

    You can create and map document attributes to index fields using the Metadata function on the console.

    When you use data sources the your document attributes are added using data sources, Amazon Q Business automatically maps specific data source document fields to Amazon Q Business index fields. You can choose to add custom fields during the data source configuration process as well.

Document attributes—both reserved and custom—can only be of the following data types: DATE, NUMBER, STRING, and STRING_LIST. To use STRING and STRING_LIST type document attributes for boosting on the console and the API, they must be enabled for search. To enable these attributes using API, use the DocumentAttributeConfiguration object of the UpdateIndex API operation.

To customize and control boosting for document attributes, use the boostingOverride parameter of the NativeIndexConfiguration object of the UpdateRetriever API operation.

To enable these attributes using the console, follow the steps in Mapping document attributes directly to index fields.

Note

If you don't enable search on these attributes, you can't boost attributes of these data types on either the Amazon Q Business console or the API.

For more information about Amazon Q Business document attributes and how to map them, see Document attributes and types.