AWS has launched a series of services, including structured data retrieval support and extraction, transform and load for unstructured data, aimed at SSL (Structure Summarization and Language Models) development to ease enterprise data access into retrieval-machine-augmented generation (RAG) pipelines.
To be accessible for RAG, structured data requires more than merely looking up a single row in a table. It needs the translation of natural language queries into complex SQL queries to filter, join and aggregate tables – difficult when working with unstructured data, as there is no pre-defined format.
AWS's extract, transform and load for unstructured data, structured data retrieval support, automatic data automation and knowledge base support services aid this process.
Amazon Bedrock Knowledge Bases service has also been launched, a fully-managed RAG facility that enables contextual, relevant data for customised response.
Structured data retrieval support in Amazon Bedrock Knowledge Bases will provide a fully-managed RAG solution from querying all structured data to enrich the model’s response and improve accuracy, learning from query patterns to tailor customisations.
GraphRG solves the challenge of explaining RAG systems, piecing together distinct pieces of data and connecting them to build a foundation for enhanced Gen AI application accuracy by using knowledge graphs – relationships within and across multiple data sources – and creating graph embeddings for Gen AI applications.
Unstructured data is challenging to extract, transform and load for SSL development as it needs to be processed and restructured. Amazon Bedrock Data Automation technology provides SSL powered ETL for unstructured data.
Amazon Bedrock Data Automation technology handles the processing of enterprise content extraction, transformation and parsing multi-model content for enhanced gen AI applications, with both automatic data transformation and industry-aligned customisation available.
AWS's new structured data retrieval support and GraphRG capabilities in Amazon Bedrock Knowledge Bases will create more complex gen AI applications without the need for graph expertise to retrieve and connect various data sources.
The new AWS services and updates like retrieval augmented generation (RAG) feature solve the challenges of accessing structured and unstructured data, helping build contextually more relevant gen AI applications for enterprise SSL development.