Timezone: »

HyperMiner: Topic Taxonomy Mining with Hyperbolic Embedding
Yi.shi Xu · Dongsheng Wang · Bo Chen · Ruiying Lu · Zhibin Duan · Mingyuan Zhou


Embedded topic models are able to learn interpretable topics even with large and heavy-tailed vocabularies. However, they generally hold the Euclidean embedding space assumption, leading to a basic limitation in capturing hierarchical relations. To this end, we present a novel framework that introduces hyperbolic embeddings to represent words and topics. With the tree-likeness property of hyperbolic space, the underlying semantic hierarchy among words and topics can be better exploited to mine more interpretable topics. Furthermore, due to the superiority of hyperbolic geometry in representing hierarchical data, tree-structure knowledge can also be naturally injected to guide the learning of a topic hierarchy. Therefore, we further develop a regularization term based on the idea of contrastive learning to inject prior structural knowledge efficiently. Experiments on both topic taxonomy discovery and document representation demonstrate that the proposed framework achieves improved performance against existing embedded topic models.

Author Information

Yi.shi Xu (Xidian University)
Dongsheng Wang (Xidian University)
Bo Chen (Xidian University)
Ruiying Lu (Xidian University)
Zhibin Duan (Xidian University)
Mingyuan Zhou (University of Texas at Austin)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors