Jump to navigation
Automated analysis of news articles on hydraulic fracturing in Colorado, New York, and Pennsylvania
University of Colorado Denver. School of Public Affairs.
University of Colorado Denver, School of Public Affairs
This report presents the findings of different approaches to media analysis of newspaper articles that focus on unconventional oil and gas development inclusive of hydraulic fracturing in New York, Pennsylvania, and Colorado. The first objective is to compare the results of manual coding of media articles on hydraulic fracturing issues from newspapers in New York and Pennsylvania to the automated coding results using the latent Dirichlet allocation (LDA) on the same articles. The purpose is to see if the automated coding can be used as a valid alternative to human coding, particularly when researchers are interested in coding a large number of articles. The findings show that, when comparing the automated coding results to the manual coding results, the inter-rater reliability was considered "fair" and the percent agreement for economic, environmental, and social topics is approximately 61%. In addition, some of the articles that were manually coded did not contain any of the codes which the researchers were searching and when these articles were removed from the analysis, the percent agreement increased to 64%.
authors: Benjamin Blair [and 4 others]
Online resource; title from PDF cover (viewed May 2018)
Hydraulic fracturing--Press coverage--Evaluation
Discourse analysis--Data processing
Automatic data collection systems
1 online resource (25 pages) : illustrations