Automated Sentiment Analysis for Personnel Survey Data in the US Air Force Context

Abstract

When surveys are distributed across the Air Force (AF), whether it be an employee engagement survey, a climate survey, or similar, significant resources are put towards the development, distribution and analysis of the survey. However, when open ended questions are included on these surveys, respondent comments are generally underutilized, more often treated as a source for pull-quotes rather than a data source in and of themselves. This is due to a lack of transparency and confidence in the accuracy of machine-aided methods such as sentiment analysis and topic modeling. This confidence reduces further when the text has special context, such as within the Air Force context. No model or methodology has been universally identified as ideal for this use case, nor has any model been universally adapted. The inconsistencies in approaches across analytical teams tasked with assessing the results of these surveys leaves data on the field.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 2021
Accession Number
AD1131133

Entities

People

  • Julia M Haines

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Air Force
  • Artificial Intelligence
  • Artificial Intelligence Software
  • Bayesian Networks
  • Computational Science
  • Data Mining
  • Data Science
  • Engineering
  • Hidden Markov Models
  • Information Retrieval
  • Information Science
  • Information Systems
  • Linguistics
  • Machine Learning
  • Neural Networks
  • Online Communications
  • Probabilistic Models
  • Probability
  • Social Media
  • Supervised Machine Learning
  • United States

Readers

  • Aerospace logistics and air mobility.
  • Organizational Psychology.
  • Systems Analysis and Design