Research map / Large language models / Language, safety and interpretability
Dict-TTS: Learning to Pronounce with Prior Dictionary Knowledge for Text-to-Speech
Ziyue Jiang, Zhe Su, Zhou Zhao, Qian Yang, Yi Ren, Jinglin Liu, Zhenhui Ye
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 Dict-TTS instead of character input | Dict-TTS · Biaobei | SER-S: 30.50% -> 6.50% | Improved |
| Use Dict-TTS instead of character input | Dict-TTS · JSUT | SER-S: 65.50% -> 22.75% | Improved |
| Use Dict-TTS instead of pypinyin phoneme input | Dict-TTS · Biaobei | PER-O: 2.78% -> 2.12% | Improved |
| Use Dict-TTS instead of pyopenjtalk phoneme input | Dict-TTS · JSUT | SER-S: 4.25% -> 22.75% | Hurt |
| Use Dict-TTS instead of NLR | Dict-TTS · Biaobei | SER-S: 26.50% -> 6.50% | Improved |
| Use Dict-TTS instead of BERT embedding | Dict-TTS · Biaobei | SER-S: 38.75% -> 6.50% | Improved |
| Pre-train S2PA on large-scale ASR data (ablation) | Dict-TTS · Biaobei | SER-S: 6.50% -> 4.25% | Improved |
| Pre-train S2PA on large-scale ASR data (ablation) | Dict-TTS · Biaobei | PER-O: 2.12% -> 1.54% | Improved |
| Remove semantic embeddings (ablation) | Dict-TTS · Biaobei | CMOS-P: -0.280 | Improved |
| Remove semantic embeddings (ablation) | Dict-TTS · Biaobei | PER-O: 2.12% -> 2.15% | No clear effect |
| Replace Gumbel-Softmax sampling with weighted sum (ablation) | Dict-TTS · Biaobei | SER-S: 6.50% -> 7.75% | Hurt |
| Use Dict-TTS instead of pypinyin phoneme input | Dict-TTS · Biaobei | MOS-P: 3.89 -> 4.03 | Improved |
Closest papers by meaning
- DiTTo-TTS: Diffusion Transformers for Scalable Text-to-Speech without Domain-Specific Factors (ICLR 2025 · similarity 0.86)
- EmergentTTS-Eval: Evaluating TTS Models on Complex Prosodic, Expressiveness, and Linguistic Challenges Using Model-as-a-Judge (NeurIPS 2025 · similarity 0.86)
- ProsodyFM: Unsupervised Phrasing and Intonation Control for Intelligible Speech Synthesis (AAAI 2025 · similarity 0.84)
- NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers (ICLR 2024 · similarity 0.83)
- NTPP: Generative Speech Language Modeling for Dual-Channel Spoken Dialogue via Next-Token-Pair Prediction (ICML 2025 · similarity 0.83)
- Paralinguistics-Aware Speech-Empowered Large Language Models for Natural Conversation (NeurIPS 2024 · 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.
