In a field traditionally driven by intuition and subjective judgment, this study presents a data-driven approach to typography, the art of arranging text. Leveraging a comprehensive dataset of font-use cases across diverse mediums, we employed Non-negative Matrix Factorization to extract three fundamental morphological characteristics of fonts: Serif vs. Sans-Serif, Basic vs. Decorative letterforms, and Light vs. Bold. This analysis demonstrated that different mediums preferentially utilize fonts with distinct morphological features. We also predicted variations between single and paired fonts, contrasting these findings with random pairings from several null models, to identify unique font-pairing trends across various mediums. Furthermore, we utilized a network analysis approach to identify the most authentic font pairings, thereby yielding practical insights for typography applications. The primary contribution of our research lies in significantly enhancing the understanding of typographies. Our work lays the groundwork for the scientific exploration of the systematic categorization of fonts and their pairings. This study establishes foundational principles for the application of typography in visual communication.