Sure! Here’s the translation into American English:
—
Amazon has taken a significant step in its effort to democratize access to data by introducing its data assistant, the Returns & ReCommerce Data Assist (RRDA). This innovative tool leverages generative artificial intelligence to transform natural language questions into validated SQL queries, thereby facilitating technical users’ ability to obtain the information they need. However, data democratization doesn’t stop at these queries, as a key objective of the Worldwide Returns & ReCommerce (WWRR) team is to enable business users to visualize trends and patterns without having to dive into SQL results.
The urgency of getting quick and definitive answers is vital, especially when conducting specific analyses that are not covered by standard reports. To strike a balance between natural language queries and effective visualizations, the team has integrated Amazon Q into QuickSight. This integration allows users to transform questions like “How many items were returned in the U.S. over the last six months?” into meaningful data visualizations.
The RRDA architecture includes an intent and domain classification system, which allows the assistant to identify whether a query requires a visualization or SQL code generation. When an intention to “show metrics” is detected, the query is redirected to Amazon Q in QuickSight. Additionally, the business domain of the query is classified, helping to find specific contexts relevant to the user.
An essential component is the Q topic retrieval and selection process, which allows the tool to present a list of topics that best fit the requested information. This process uses a machine learning model to evaluate the capability of the topics, thereby enhancing accuracy and optimizing the user experience.
Among the system’s challenges is the need to rephrase user questions to make them suitable for Amazon Q. Thanks to the Converse API of Amazon Bedrock, RRDA can reconfigure questions while preserving their original meaning, thus allowing for effective processing by QuickSight.
Incorrect or imprecise formulations could result in suboptimal visualizations. However, thanks to RRDA’s integration with QuickSight, visualizations are generated directly in the context of the dialogue, improving workflow fluidity and reducing the time needed to reach data-driven conclusions.
The system relies on a database that is updated daily, ensuring that information about Q topics is always relevant. This database is continuously enriched with questions validated by experts, improving the quality of recommendations.
As Amazon continues to refine RRDA and address user feedback, this tool is expected to become an essential resource in decision-making, allowing users to focus on questions and insights while the technology handles the creation of necessary visualizations.
Referrer: MiMub in Spanish
