FAA Office of Aerospace Medicine 
Civil Aerospace Medical Institute 

Report No: DOT/FAA/AM-23/22

Title and Subtitle: Improving Computational Usability of Unstructured Pilot Medical Certification Data

Report Date: June 2023

Authors: C.Horowitz

Abstract:Current Federal Aviation Administration's Office of Aerospace Medicine (AAM) operations include scanning paper documentsreceived from various third‑party medical providers to support individual pilot medical certification decision-related matters.These operations lack analytic tools, resulting in a time-intensive effort by subject matter experts to manually search throughdocument sets to find relevant information. MITRE applied human language technologies to a sample set of AAM documentscontaining unstructured pilot medical certification data. MITRE demonstrated that third-party documents received andscanned by AAM could be automatically classified through a combination of computer vision and human languagetechnologies; demonstrated that these documents can have their content extracted with sufficient accuracy to enablefundamental automation and human support tasks including summarization, search, and de-identification; identified aspectsof the documents that present risk to future systems development; and produced a prototypical integrated documentprocessing software pipeline capable of automatic ingest, classification, content extraction, and indexing to support basicsearch as an initial application.

Key Words: Aviation, Safety

No. of Pages: 10