Unmasking Machine Learning With Tensor Decomposition: An Illustrative Example for Media and Communication Researchers

As online communication data continues to grow, manual content analysis, which is frequently employed in media studies within the social sciences, faces challenges in terms of scalability, efficiency, and coding scope. Automated machine learning can address these issues, but it often functions as a black box, offering little insight into the features driving its predictions. This lack of interpretability limits its application in advancing social science communication research and fostering practical outcomes. Here, explainable AI offers a solution that balances high prediction accuracy with interpretability. However, its adoption in social science communication studies remains limited. This study illustrates tensor decomposition—specifically, PARAFAC2—for media scholars as an interpretable machine learning method for analyzing high-dimensional communication data. By transforming complex datasets into simpler components, tensor decomposition reveals the nuanced relationships among linguistic features. Using a labeled spam review dataset as an illustrative example, this study demonstrates how the proposed approach uncovers patterns overlooked by traditional methods and enhances insights into language use. This framework bridges the gap between accuracy and explainability, offering a robust tool for future social science communication research.

Abstract licence: CC BY

40 more references, whose wording the paper’s licence keeps private.

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.