Case Study: Airbus achieves sub-second accurate QA over text and tables with deepset

How Airbus built a combined table and text QA system that returns accurate results in under a second

Airbus, a global aerospace leader, faced a challenge where pilots struggled to quickly find critical information in their extensive, table-heavy Flight Crew Operating Manuals (FCOMs) using basic keyword search. To improve information discovery in high-pressure cockpit situations, Airbus's AI research unit turned to deepset and leveraged its open-source Haystack framework for natural language processing to build a more intuitive solution.

deepset's solution was a complex question answering system that processes queries through dual pipelines for both text and tables, pinpointing the exact answer from over a thousand pages of documentation. The implemented system delivers highly accurate results by identifying the correct cell in a table or passage in text in under one second. While not yet systematically evaluated, early feedback from operations teams has been very encouraging, with Airbus excited about the potential of this technology to save crucial time.

Airbus

Alexandre Arnold

AI Research Unit