A Multiscale Geometric Method for Capturing Relational Topic Alignment
This addresses the need for interpretable topic modeling to track research evolution in co-author communities, but it is incremental as it builds on existing geometric and hierarchical methods.
The paper tackled the problem of missing rare topics and failing to capture smooth temporal alignment in topic modeling for scientific corpora, by proposing a geometric method that integrates multimodal data, resulting in effective identification of rare-topic structure and visualization of smooth topic drift over time.
Interpretable topic modeling is essential for tracking how research interests evolve within co-author communities. In scientific corpora, where novelty is prized, identifying underrepresented niche topics is particularly important. However, contemporary models built from dense transformer embeddings tend to miss rare topics and therefore also fail to capture smooth temporal alignment. We propose a geometric method that integrates multimodal text and co-author network data, using Hellinger distances and Ward's linkage to construct a hierarchical topic dendrogram. This approach captures both local and global structure, supporting multiscale learning across semantic and temporal dimensions. Our method effectively identifies rare-topic structure and visualizes smooth topic drift over time. Experiments highlight the strength of interpretable bag-of-words models when paired with principled geometric alignment.