Visual representation of safety narratives. (October 2016)

• Record Type: Journal Article
• Authors: Robinson, S.D.
• Is Part Of: Safety science. Volume 88(2016)
• Page Start: 123
• Page End: 128
• Publication Date: 2016-10
• Subjects: LSA -- Adaptive taxonomy -- Safety -- Isometric mapping -- Machine learning; Industrial accidents -- Periodicals; Accident Prevention -- Periodicals; Safety -- Periodicals; Travail -- Accidents -- Périodiques; 363.11
• DOI: 10.1016/j.ssci.2016.05.005
• Languages: English
• ISSNs: 0925-7535

Abstract

Abstract: A computational method for the visualization of text-based safety narratives on a two dimensional plane is shown. This multi-step approach utilizes latent semantic analysis to first infer higher-order structures and then isometric mapping to reduce the projection to two dimensions. Metadata may then be overlaid on the projection. Demonstrated is the application of this process to the human coded primary-problems identified and the phase of flight for a sample of the Aviation Safety Reporting System database. It is evident that this approach provides additional insight for the analysis of large inter-related corpora commonly found in safety programs.