基于排序原型的流形感知话题建模方法
原标题：A Manifold-Aware Topic Modeling Approach via Rank-Based Prototypes
AI 摘要
研究者提出 MARETopic，一种无需训练的基于排序原型的话题建模框架。该方法将文档嵌入投影到低维流形后构建排序列表，用贪心算法选出恰好 K 篇真实语料文档作为话题原型，其邻域覆盖整个语料。在两个类别最多的基准上，其 Purity 与 NMI 超过神经和聚类类话题模型，且无需梯度更新，在两个数据集上话题连贯性领先，另一变体运行速度快 1.7 至 1.9 倍。
正文节选
A Manifold-Aware Topic Modeling Approach via Rank-Based Prototypes Abstract Recent topic models leverage pretrained embeddings, but neural architectures produce latent representations without grounding in specific texts, and clustering-based pipelines assign representative documents only post hoc, relying on absolute distances distorted by hubness and anisotropy in high-dimensional spaces. We introduce MARETopic, a training-free framework that casts topic discovery as rank-based prototype select