Implications of the Computational Social Science Perspective in the Analysis of Hypertextual Conversations on Instagram around Freight Train Graffiti
Abstract
Graffiti on freight trains exists in a constant duality between the physical and digital realms, with artwork displayed on rolling vehicles across the North American subcontinent. This dual circuit of circulation encompasses the geographical movement of trains along the railway network and the digital presence on platforms like Instagram. The study of this phenomenon addresses the increasing datafication of everyday life and employs theoretical perspectives and diverse techniques to explore it.
Computational social sciences offer valuable conceptual guidance for analyzing this kind of phenomena. Rather than pursuing new concepts, researchers modify existing frameworks to interpret data using computational tools and artificial intelligence. These tools provide the benefits of aggregating and analyzing large volumes of data, uncovering hidden patterns and insights.
However, analyzing the specific phenomenon of New York graffiti on freight trains requires specialized approaches. It involves data collection, custom model training, unconventional metric usage, and potentially proposing new metrics. This work aims to elucidate the theoretical, methodological, and empirical decisions made in programming three computational tools for data gathering and computational analysis.
The first tool, Instagrapi, enables data mining to gather Instagram posts relevant to graffiti on freight trains. The second tool employs a TensorFlow object detection model to automatically classify published content, identifying graffiti types in images. The third tool, SpaCy, facilitates text analysis by extracting relevant terms from hashtags, such as graffiti writers’ names, crews, north america cities, and terms from a custom-glossary.