AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Unified Segment-to-Segment Framework for Simultaneous Sequence Generation

Shaolei Zhang, Yang Feng

NeurIPS 2023 · Image generation › Autoregressive and latent generation · approach: Neural text-to-speech synthesis · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use Seg2Seg instead of T-T on latencyStreaming ASR · LibriSpeechlatency: 610 -> 324Improved
Use Seg2Seg instead of T-T on noisy subsetStreaming ASR · LibriSpeech otherWER (%): 9.50 -> 8.73Improved
Use Seg2Seg instead of T-T on clean subsetStreaming ASR · LibriSpeech cleanWER (%): 3.40 -> 3.55No clear effect
Use Seg2Seg instead of MoChAStreaming ASR · LibriSpeech otherWER (%): 14.20 -> 8.73Improved
Add latency-reducing methods (SelfAlign) to T-T, lowering latencyT-T · LibriSpeechWER other (%): 9.50 -> 10.70Hurt
Replace fixed 280ms segmentation with adaptive latent segments (ablation)Seg2Seg · Buckeye segmentationR-value: 38.4 -> 41.2Improved
Replace fixed 280ms segmentation with adaptive latent segments (ablation)Seg2Seg · Buckeye segmentationprecision: 28.1 -> 41.1Improved
Replace fixed-length source segmentation with adaptive aggregation (ablation)Seg2Seg · SimulST MuST-C En-DeBLEU: 1 BLEUImproved
Replace wait-k/wait-k-stride-n emission with adaptive emission (ablation)Seg2Seg · SimulST MuST-C En-DeBLEUImproved
Add multi-task learning with streaming ASR and SimulMTSeg2Seg · SimulST MuST-C En-DeBLEUImproved
Increase latency weight lambda from 0.05 to 0.4Seg2Seg · SimulMT WMT15 De-EnBLEU: 32.20 -> 27.90Hurt
Increase latency weight lambda from 0.05 to 0.4Seg2Seg · SimulST MuST-C En-DeBLEU: 23.63 -> 14.47Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy