Computational support for media ecosystems research

Abstract

This thesis summarizes the design, implementation, and evaluation of two end-user web tools for automated content analysis of online news data. The first tool is a visualization that displays neural word embeddings data, allowing a user to explore words used in similar contexts within a text corpus. The second tool is an interface that guides users through a supervised machine learning pipeline, enabling novices to train their own binary classification models to detect the presence of a specific frame within the text of a news story. The visualization and interface were evaluated in a user study and think-aloud test respectively. These tools were developed for integration into Media Cloud, an open-source platform for media analysis, which is part of a larger effort to facilitate and advance media ecosystems research.