Beyond the Search Box: Visualizing Digital Cultural Heritage through Generous Interfaces and Machine Learning

The rapid digitization of cultural heritage collections has produced unprecedented volumes of archival data, yet most digital repositories continue to rely on linear, metadata-driven interfaces that constrain holistic interpretation and exploratory discovery. This paper presents a visualization-centered framework designed to support semantically guided navigation of large-scale cultural heritage datasets, developed through a case study with the Tainacan collection of the Museu Nacional dos Povos Indígenas, the largest digital repository of Indigenous artifacts in Brazil. Our primary contribution lies in the design of an interactive visual analytics environment that provides a global, navigable view of the collection while supporting filtering, clustering, and item-level inspection. The system leverages visual and textual embeddings derived from deep neural models to encode features, which are projected into two-dimensional semantic spaces. These algorithmic similarities are framed not as authoritative classifications but as exploratory hypotheses that encourage curatorial reflection and critical engagement with institutional taxonomies. To ground computational abstractions in culturally situated contexts, the platform also integrates temporal and geospatial visualizations that map acquisition histories alongside Indigenous geographies. Developed as an open source project (https://github.com/Luizerko/indigenous_clusters_and_communities) with full pipeline documentation, the tool supports reproducibility and transferability to other heritage datasets. By bridging deep learning and visualization within a culturally responsible design process, this work advances scalable and interpretable interfaces for the exploration of digital cultural heritage collections. Moving forward, the project's framework opens avenues for technical, curatorial, and experiential enhancements, ranging from dynamic archive integration and AI-assisted metadata completion to immersive exhibition environments.