Research map / Label-efficient and robust learning / Recognition and long-tail
Knowledge-Aware Bayesian Deep Topic Model
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 topic-tree graph likelihood and GCN aggregation (Sawtooth -> TopicKG), 7 layers (ablation) | SawETM · 20NG document classification | Micro F1: 53.71 -> 56.32 | Helped |
| add topic-tree graph likelihood and GCN aggregation (Sawtooth -> TopicKG), 3 layers (ablation) | SawETM · 20NG document classification | Micro F1: 52.41 -> 55.73 | Helped |
| add topic-tree graph likelihood (Sawtooth -> TopicKG), 7 layers (ablation) | SawETM · RCV2 document classification | Micro F1: 82.46 -> 82.48 | No clear effect |
| add adaptive graph structure (TopicKG -> TopicKGA), 7 layers (ablation) | TopicKG · 20NG document classification | Micro F1: 56.32 -> 60.04 | Helped |
| add adaptive graph structure (TopicKG -> TopicKGA), 7 layers (ablation) | TopicKG · R8 document classification | Micro F1: 94.04 -> 94.10 | Helped |
| add adaptive graph structure (TopicKG -> TopicKGA), 7 layers (ablation) | TopicKG · RCV2 document classification | Micro F1: 82.48 -> 83.08 | Helped |
| add adaptive graph structure (TopicKG -> TopicKGA), 3 layers (ablation) | TopicKG · 20NG document classification | Micro F1: 55.73 -> 58.63 | Helped |
| TopicKGA (7 layers) versus BERT-based CombinedTM | CombinedTM · RCV2 document classification | Micro F1: 82.91 -> 83.08 | Helped |
| increase hierarchy depth from 3 to 7 layers (ablation) | TopicKGA · 20NG document classification | Micro F1: 58.63 -> 60.04 | Helped |
| increase hierarchy depth from 3 to 7 layers (ablation) | TopicKG · RCV2 document classification | Micro F1: 84.75 -> 82.48 | Hurt |
| set knowledge confidence beta to 0 (ignore topic tree) (ablation) | TopicKG · 20NG, R8, RCV2 | Micro F1 and WE | Hurt |
Closest papers by meaning
- Bayesian Deep Embedding Topic Meta-Learner (ICML 2022 · similarity 0.91)
- Bayesian Progressive Deep Topic Model with Knowledge Informed Textual Data Coarsening Process (ICML 2023 · similarity 0.90)
- Dirichlet belief networks for topic structure learning (NeurIPS 2018 · similarity 0.90)
- TopicNet: Semantic Graph-Guided Topic Discovery (NeurIPS 2021 · similarity 0.90)
- Representing Mixtures of Word Embeddings with Mixtures of Topic Embeddings (ICLR 2022 · similarity 0.89)
- On the Affinity, Rationality, and Diversity of Hierarchical Topic Modeling (AAAI 2024 · similarity 0.87)
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.