Categorization of Survey Text Utilizing Natural Language Processing and Demographic Filtering
Abstract
Thousands of Navy survey free text comments are overlooked every year because reading and interpreting comments is expensive, time consuming, and subjective. Valuable information from these comments is not being utilized to make important Navy decisions. We provide a new procedure to automate the identification of primary topics in short, jargon laced, topic based survey comments by applying a label to each comment and then using those labels to bin comments into operationally meaningful categories. We apply this method to the Navy Retention Survey to provide the Chief of Naval Personnel with an objective analysis of the questions Why are sailors leaving? and What will make sailors stay on active duty? Furthermore, we introduce an implementation of this method using the Demographic Analysis of Responses Tool for Surveys (DARTS), which allows us to filter comment bins using the over 100 demographic and military status elements associated with each sailor. By targeting critically undermanned specialties, the reports generated with this tool provide quantifiable results that allow retention policy makers the ability to review, modify, and create relevant incentives to retain critically talented sailors to meet fiscal year end strength and operational requirements.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2017
- Accession Number
- AD1046799
Entities
People
- Christine M. Cairoli
Organizations
- Naval Postgraduate School