Visualization of Large-Scale Narrative Data Describing Human Error

Abstract

Objective

Introduced is a visual data exploration technique for compiling, reducing, organizing, visually rendering, and filtering text-based narratives for detailed analysis.

Background

The analysis of data sets provides an increasingly difficult problem. The method of visual representation is considered an effective tool in many applications. The focus of this study was to determine if a latent semantic analysis–based projection of narrative data into a geographic information systems software program provided a useful tool for reducing and organizing large sums of narrative data for analysis.

Method

This approach utilizes latent semantic analysis to reduce narratives to a high-dimensional vector, truncates the vector to a two-dimensional projection through application of isometric mapping, and then visually renders the result with geographic information systems software. This method is demonstrated on aviation self-reported safety narratives sourced from the Aviation Safety Reporting System.

Results

Thematic regions from the corpus are illustrated along with the first five topics identified.

Conclusion

Shown is the ability to assimilate a large number of narratives, identify contextual themes, recognize common events and outliers, and organize resultant topics.

Application

Large narrative-based data sets present in aviation and other domains may be visualized to facilitate efficient analysis, enhance comprehension, and improve safety.