Nonparametric Forest-Structured Neural Topic Modeling

Zhihong Zhang, Xuewen Zhang, Yanghui Rao


Abstract
Neural topic models have been widely used in discovering the latent semantics from a corpus. Recently, there are several researches on hierarchical neural topic models since the relationships among topics are valuable for data analysis and exploration. However, the existing hierarchical neural topic models are limited to generate a single topic tree. In this study, we present a nonparametric forest-structured neural topic model by firstly applying the self-attention mechanism to capture parent-child topic relationships, and then build a sparse directed acyclic graph to form a topic forest. Experiments indicate that our model can automatically learn a forest-structured topic hierarchy with indefinite numbers of trees and leaves, and significantly outperforms the baseline models on topic hierarchical rationality and affinity.
Anthology ID:
2022.coling-1.228
Volume:
Proceedings of the 29th International Conference on Computational Linguistics
Month:
October
Year:
2022
Address:
Gyeongju, Republic of Korea
Venue:
COLING
SIG:
Publisher:
International Committee on Computational Linguistics
Note:
Pages:
2585–2597
Language:
URL:
https://aclanthology.org/2022.coling-1.228
DOI:
Bibkey:
Cite (ACL):
Zhihong Zhang, Xuewen Zhang, and Yanghui Rao. 2022. Nonparametric Forest-Structured Neural Topic Modeling. In Proceedings of the 29th International Conference on Computational Linguistics, pages 2585–2597, Gyeongju, Republic of Korea. International Committee on Computational Linguistics.
Cite (Informal):
Nonparametric Forest-Structured Neural Topic Modeling (Zhang et al., COLING 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.coling-1.228.pdf
Code
 hostnlp/ntsntm +  additional community code