Developing a Taxonomy of Historical Visualizations
The paper "VisTaxa: Developing a Taxonomy of Historical Visualizations" presents an empirical approach to constructing a taxonomy for categorizing historical visualizations, addressing the gap in existing taxonomies which primarily focus on contemporary visualization techniques. This research illustrates the potential of classifying historical visualizations to enhance the understanding and utility of such visual artifacts.
Methodology
The methodology incorporates creating a taxonomy through qualitative data analysis, facilitated by a protocol that guides the coding process for image categorization. The team implemented a systematic workflow involving multiple coders to develop a taxonomy structure, focusing on precise image labeling while accommodating subjective interpretations by researchers. Critical to the coding process were steps involving individual taxonomy creation, conflict resolution to consolidate coding differences, label modification, and further conflict resolution.
The paper leverages a corpus from the OldVisOnline dataset, which consists of visualizations dated before 1950, aiming to capture a diverse range of designs typical of historical practices. This choice is crucial for addressing the research question: "What visual designs have been used in historical visualizations?"
Noteworthy is the integration of VisTaxa, a system developed to assist the coding process. The system supports taxonomy labeling, incorporating machine assistance like clustering and semantic cluster overview to facilitate image classification tasks. This tool plays a pivotal role in improving the efficiency and accuracy of taxonomic development in visual analysis.
Results and Analysis
The resulting taxonomy from coding 400 images is organized hierarchically, pinpointing 51 distinct visualization types. Among the frequently occurring types are map, bar chart, and line chart, paralleling common techniques seen today, yet notably absent are scatter charts which reflect innovations from the later periods in history.
The paper ventures further by predicting taxonomy labels for the remainder of the dataset not directly coded by researchers, using similarity-based matching methods that outperformed alternative zero-shot classification techniques.
Observations on the historical distribution of certain types, such as bathymetric maps and early appearances of parallel-coordinate designs, offer valuable insight into the historical context and evolution of visualization designs.
Implications for Research and Practice
This work serves as a foundational effort towards a more comprehensive taxonomy of historical visualizations, striving to bridge the gap between historical and contemporary visualization research. The proposed taxonomy can significantly enhance retrieval and analysis of historical visualization assets, offering structured perspectives that can stimulate further academic discussion and exploration in digital humanities and visualization studies.
Moreover, by systematically categorizing historical visualization techniques, this research improves accessibility and comprehension among scholars, fostering a deeper understanding of the evolution and foundational practices of visualization. Beyond the theoretical framework, practical applications may include improved data indexing, aiding historians and researchers in tracing the lineage and transformation of visualization techniques.
Future Directions
Future work can expand upon this taxonomy by incorporating non-Western historical visualization practices, further utilizing machine learning techniques to refine predictions and adapt the framework to larger, more diverse datasets. Additionally, extending this approach to analyze the impact of social and technological context on visualization practices could provide richer insights into historical design evolution.
In conclusion, "VisTaxa: Developing a Taxonomy of Historical Visualizations" contributes an essential perspective to the discourse on visualization taxonomy, providing both a methodological framework and practical insights for understanding and utilizing historical visualizations in research.
