Research map / Probabilistic and scientific modeling / Molecules and proteins
Sequence Modeling via Segmentations
Chong Wang, Yining Wang, Po-Sen Huang, Abdelrahman Mohamed, Dengyong Zhou, Li Deng
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 |
|---|---|---|---|
| Use segmented output modeling instead of LDA | Text segmentation model · AP | predictive log likelihood: -9.22 -> -8.28 | Improved |
| Use segmented output modeling instead of LDA | Text segmentation model · CiteULike | predictive log likelihood: -7.82 -> -7.60 | Improved |
| Use SWAN instead of Attention RNN | TIMIT phoneme recognition · TIMIT core test | PER (%): 17.6 -> 18.1 | Hurt |
| Use SWAN instead of CTC | TIMIT character recognition · TIMIT character recognition | CER (%): 31.8% -> 30.5% | Improved |
| Allow maximum segment length 3 instead of 1 (ablation) | Text segmentation model (hidden 150) · AP | predictive log likelihood: -8.38 -> -8.28 | Improved |
| Allow maximum segment length 3 instead of 1 (ablation) | Text segmentation model (hidden 150) · CiteULike | predictive log likelihood: -8.12 -> -7.60 | Improved |
| Increase maximum segment length from 1 to 2 (ablation) | Text segmentation model (hidden 150) · AP | predictive log likelihood: -8.38 -> -8.30 | Improved |
| Increase maximum segment length from 1 to 2 (ablation) | Text segmentation model (hidden 150) · CiteULike | predictive log likelihood: -8.12 -> -7.67 | Improved |
| Increase maximum segment length from 2 to 3 (ablation) | Text segmentation model (hidden 150) · AP | predictive log likelihood: -8.30 -> -8.28 | Improved |
| Increase maximum segment length from 2 to 3 (ablation) | Text segmentation model (hidden 150) · CiteULike | predictive log likelihood: -7.67 -> -7.60 | Improved |
| Increase maximum segment length from 3 to 6 (ablation) | SWAN phoneme recognition · TIMIT | average segment length | No clear effect |
Closest papers by meaning
- Temporal Attention-Gated Model for Robust Sequence Classification (CVPR 2017 · similarity 0.85)
- Dynamic Chunking for End-to-End Hierarchical Sequence Modeling (ICLR 2026 · similarity 0.84)
- Unified Segment-to-Segment Framework for Simultaneous Sequence Generation (NeurIPS 2023 · similarity 0.84)
- Structured Reordering for Modeling Latent Alignments in Sequence Transduction (NeurIPS 2021 · similarity 0.84)
- A Semantic Space is Worth 256 Language Descriptions: Make Stronger Segmentation Models with Descriptive Properties (ECCV 2024 · similarity 0.84)
- Incremental Sequence Classification with Temporal Consistency (NeurIPS 2025 · similarity 0.83)
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.
