Research map / Label-efficient and robust learning / Recognition and long-tail
Effective Neural Topic Modeling with Embedding Clustering Regularization
Xiaobao Wu, Xinshuai Dong, Thong Thanh Nguyen, Anh Tuan Luu
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add embedding clustering regularization (ECR) to topic model (ablation) | ECRTM · 20NG, IMDB, Yahoo, AG News K=50 | topic diversity | Improved |
| add ECR to topic model (ablation) | ECRTM · document clustering (ablation) | Purity/NMI | Improved |
| replace Deep KMeans clustering regularization with ECR (optimal transport, sparse, size-constrained) (ablation) | ECRTM · topic diversity | TD | Improved |
| add entropy minimization to DKM for sparse assignments (DKM+Entropy) (ablation) | DKM regularization · IMDB topics | TD | No clear effect |
| use ECRTM instead of NSTM | Neural topic model · Yahoo Answer K=50 | C_V: 0.390 -> 0.405 | Improved |
| use ECRTM instead of NSTM | Neural topic model · Yahoo Answer K=50 | TD: 0.658 -> 0.985 | Improved |
| use ECRTM instead of NSTM | Neural topic model · AG News K=100 | C_V: 0.421 -> 0.416 | Hurt |
| use ECRTM instead of runner-up for document clustering | Topic model · 20NG K=50 | Purity: 0.367 -> 0.560 | Improved |
| remove high-frequency words by lowering max-df for baselines | LDA, ETM, NSTM · 20NG | TD | Improved |
| use ECRTM regardless of preprocessing (max-df) | ECRTM · 20NG max-df 0.5 to 0.1 | topic quality and clustering | Improved |
| use ECRTM instead of LDA, DVAE, ETM | Topic model · 20NG | perplexity (lower better): 2044.6 -> 1896.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HyperMiner — HyperMiner: Topic Taxonomy Mining with Hyperbolic Embedding (NeurIPS 2022) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Context-guided Embedding Adaptation for Effective Topic Modeling in Low-Resource Regimes (NeurIPS 2023 · similarity 0.87)
- Representing Mixtures of Word Embeddings with Mixtures of Topic Embeddings (ICLR 2022 · similarity 0.87)
- Neural Topic Model via Optimal Transport (ICLR 2021 · similarity 0.87)
- Inter and Intra Topic Structure Learning with Word Embeddings (ICML 2018 · similarity 0.86)
- TopicNet: Semantic Graph-Guided Topic Discovery (NeurIPS 2021 · similarity 0.86)
- FASTopic: Pretrained Transformer is a Fast, Adaptive, Stable, and Transferable Topic Model (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
