Mengyue Wu
Associate Professor
Works on speech and audio processing, including audio captioning and speech-based mental health assessment.
Mengyue Wu's lab at SJTU works on speech recognition & spoken language, music, audio & sound and vision-language & multimodal. 40 papers since 2023, 27 at top venues, mostly ICASSP and Interspeech. 4 PhD students and 11 master's students. 3 alumni, most often to Ant Group and Xiaomi. Applications close .
At a glance
- papers, 2023–2026
- 10/ year
- at T1 venues
- 68%
- PhD students
- 4
- MS students
- 11
- alumni listed
- 3
Papers by yearfrom DBLP, 2023–2026
- 3 papers2026 · T1 2 · collecting
- 12 papers2025 · T1 11
- 16 papers2024 · T1 9
- 9 papers2023 · T1 5
T1 venuesOtherVenues: ICASSP, Interspeech, NAACL, AAAI, ACL
Research focuspaper titles that use each topic's words
- Music, audio & sound19 papers
- audio18 papers
- Speech recognition & spoken language8 papers
- Vision-language & multimodal7 papers
- audio processing0 papers
Paper list
From DBLP.
20263 papers
- AAAI (T1)A Human-Centric Pipeline for Aligning Large Language Models with Chinese Medical Ethics (opens in a new tab)
- ICLR (T1)From Medical Records to Diagnostic Dialogues: A Clinical-Grounded Approach and Dataset for Psychiatric Comorbidity (opens in a new tab)
- Findings of ACLSemi-Supervised Diseased Detection from Speech Dialogues with Multi-Level Data Modeling (opens in a new tab)
202512 papers
- NAACL (T1)A Cognitive Evaluation Benchmark of Image Reasoning and Description for Large Vision-Language Models (opens in a new tab)
- NAACL (T1)A Diverse and Effective Retrieval-Based Debt Collection System with Expert Knowledge (opens in a new tab)
- ICASSP (T1)AudioTime: A Temporally-aligned Audio-text Benchmark Dataset (opens in a new tab)
- AI OpenDFM: Dialogue foundation model for universal large-scale dialogue-oriented task learning (opens in a new tab)
- Interspeech (T1)Dog2vec: Self-Supervised Pre-Training for Canine Vocal Representation (opens in a new tab)
- AAAI (T1)Is Your Image a Good Storyteller? (opens in a new tab)
- NAACL (T1)MedEthicEval: Evaluating Large Language Models Based on Chinese Medical Ethics (opens in a new tab)
- ICASSP (T1)PicoAudio: Enabling Precise Temporal Controllability in Text-to-Audio Generation (opens in a new tab)
- ICASSP (T1)Smooth-Foley: Creating Continuous Sound for Video-to-Audio Generation Under Semantic Guidance (opens in a new tab)
- ACL (T1)Toward Automatic Discovery of a Canine Phonetic Alphabet (opens in a new tab)
- ACL (T1)Tracking Life's Ups and Downs: Mining Life Events from Social Media Posts for Mental Health Analysis (opens in a new tab)
- NeurIPS (T1)WritingBench: A Comprehensive Benchmark for Generative Writing (opens in a new tab)
202416 papers
- ICASSP (T1)A Detailed Audio-Text Data Simulation Pipeline Using Single-Event Sounds (opens in a new tab)
- ACM MMAuto-ACD: A Large-scale Dataset for Audio-Language Representation Learning (opens in a new tab)
- Findings of EMNLPAutomatic Reconstruction of Ancient Chinese Pronunciations (opens in a new tab)
- IEEE ACM Trans. Audio Speech Lang. Process.Beyond the Status Quo: A Contemporary Survey of Advances and Challenges in Audio Captioning (opens in a new tab)
- Interspeech (T1)DiveSound: LLM-Assisted Automatic Taxonomy Construction for Diverse Audio Generation (opens in a new tab)
- Interspeech (T1)Efficient Audio Captioning with Encoder-Level Knowledge Distillation (opens in a new tab)
- ICASSP (T1)Enhancing Audio Generation Diversity with Visual Information (opens in a new tab)
- Interspeech (T1)Enhancing Zero-shot Audio Classification using Sound Attribute Knowledge from Large Language Models (opens in a new tab)
- Interspeech (T1)FakeSound: Deepfake General Audio Detection (opens in a new tab)
- MLSPInvestigating Passive Filter Pruning for Efficient CNN-Transformer Audio Captioning (opens in a new tab)
- NAACL (T1)Mapping Long-term Causalities in Psychiatric Symptomatology and Life Events from Social Media (opens in a new tab)
- AAAI (T1)Multi-Label Supervised Contrastive Learning (opens in a new tab)
20239 papers
- ACM MMBLAT: Bootstrapping Language-Audio Pre-training based on AudioSet Tag-guided Synthetic Data (opens in a new tab)
- EMNLP (T1)Detection of Multiple Mental Disorders from Social Media with Two-Stream Psychiatric Experts (opens in a new tab)
- ICASSP (T1)Diverse and Vivid Sound Generation from Text Descriptions (opens in a new tab)
- Interspeech (T1)Enhance Temporal Relations in Audio Captioning with Sound Event Detection (opens in a new tab)
- ICASSP WorkshopInvestigating Pooling Strategies and Loss Functions for Weakly-Supervised Text-to-Audio Grounding via Contrastive Learning (opens in a new tab)
- Trans. Assoc. Comput. LinguisticsOPAL: Ontology-Aware Pretrained Language Model for End-to-End Task-Oriented Dialogue (opens in a new tab)
- Interspeech (T1)ReCLR: Reference-Enhanced Contrastive Learning of Audio Representation for Depression Detection (opens in a new tab)
- EMNLP (T1)Semantic Space Grounded Weighted Decoding for Multi-Attribute Controllable Dialogue Generation (opens in a new tab)
- Findings of ACLTranscribing Vocal Communications of Domestic Shiba lnu Dogs (opens in a new tab)
Where alumni went3 listed, 67% to industry
Industry 2Academia 1
Most often to
- Ant Group
- Xiaomi
Apply to SJTUFall 2027
Computer Science PhD (international applicants)
- Round
- Self-funded
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- GBUCL
Ifat Yasin
- Speech recognition & spoken language
- CNSJTU
Jie Zhu
- Speech recognition & spoken language
- CNTsinghua
Weiqiang Zhang
- Speech recognition & spoken language
- CNSJTU
Yanmin Qian
- Speech recognition & spoken language
- CNSJTU
Wangyou Zhang
- Speech recognition & spoken language
- CNZhejiang
Li Lu
- Speech synthesis & voice
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- DBLP
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
SJTUcloses