deepset
11 Case Studies
A deepset Case Study
Airbus, a global aerospace leader, faced a challenge in helping pilots quickly find critical information within their extensive, complex Flight Crew Operating Manuals (FCOMs). Their existing digitized manuals relied on keyword searches, which were inefficient and time-consuming, especially in high-pressure cockpit situations. To address this, Airbus’s AI research unit partnered with deepset, leveraging its open-source Haystack framework for natural language processing, to build a more intuitive and faster question-answering system.
deepset's solution was a complex composite QA system capable of extracting precise answers from both plain text and tables within the massive manuals. The system processes a query through dual pipelines for text and table QA, then uses a final node to select the most confident answer. The result is a highly accurate system that can pinpoint the correct cell in a table from over a thousand pages and deliver the answer in under one second. Early feedback from operations teams has been very encouraging, significantly reducing information retrieval time for pilots.