Research map / Label-efficient and robust learning / Recognition and long-tail
Discourse Relations Detection via a Mixed Generative-Discriminative Framework
Jifan Chen, Qi Zhang, Pengfei Liu, Xuanjing Huang
Read the paperCheck your idea against 137,217 papers
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 offset weights (Weighted Fisher Vector instead of Fisher Vector) (ablation) | Fisher vector offset model · PDTB Contingency | score: 51.86% -> 53.57% | Improved |
| add offset weights (WFV vs FV) (ablation) | Fisher vector offset model · PDTB Temporal | score: 18.28% -> 20.24% | Improved |
| add offset weights (WFV vs FV) (ablation) | Fisher vector offset model · PDTB Comparison | score: 29.75% -> 30.21% | Improved |
| add offset weights (WFV vs FV) (ablation) | Fisher vector offset model · PDTB Expansion | score: 80.50% -> 80.90% | Improved |
| add offset weights (WFV vs FV) (ablation) | Fisher vector offset model · Explanatory relations in reviews | F1: 77.4% -> 79.3% | Improved |
| add offset weights (WFV vs FV) (ablation) | Fisher vector offset model · Explanatory relations in reviews | accuracy: 73.4% -> 75.2% | Improved |
| word-embedding offsets with Fisher vector instead of composed segment vectors (ADD+CON) | relation classifier · PDTB Contingency | score: 40.03% -> 53.57% | Improved |
| WFV vs best prior method on Temporal | relation classifier · PDTB Temporal | score: 27.63% -> 20.24% | Hurt |
| WFV vs Ji and Eisenstein on Comparison | relation classifier · PDTB Comparison | score: 35.93% -> 30.21% | Hurt |
| change word embedding type (ablation) | Fisher vector model · PDTB (four relations) | precision/recall | No clear effect |
| increase number of GMM Gaussians (ablation) | Fisher vector model · PDTB (four relations) | precision/recall/F1 | No clear effect |
| use offsets model vs previous best on reviews | relation classifier · Explanatory relations in reviews | accuracy: 56.2% -> 75.2% | Improved |
Closest papers by meaning
- Further results on latent discourse models and word embeddings (JMLR 2021 · similarity 0.87)
- Linguistic Properties Matter for Implicit Discourse Relation Recognition: Combining Semantic Interaction, Topic Continuity and Attribution (AAAI 2018 · similarity 0.86)
- Unsupervised Learning of Discourse Structures using a Tree Autoencoder (AAAI 2021 · similarity 0.86)
- Predicting Above-Sentence Discourse Structure Using Distant Supervision from Topic Segmentation (AAAI 2022 · similarity 0.85)
- A Label Dependence-Aware Sequence Generation Model for Multi-Level Implicit Discourse Relation Recognition (AAAI 2022 · similarity 0.85)
- A Positive-Unlabeled Metric Learning Framework for Document-Level Relation Extraction with Incomplete Labeling (AAAI 2024 · similarity 0.82)
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.
