This organization hosts my PhD research projects on sarcasm and irony understanding in human-machine interaction.
My work investigates how computers can move beyond literal language processing to recognize intended meaning from context, prosody, multimodal signals, subjectivity, and cultural interpretation.
Xiyuan Gao · University of Groningen · 2021–2026
- Multimodal computing
- Computational pragmatics
- Affective computing
- Cross-lingual and multilingual transfer
- Linguistically informed AI
- Human-centered human--machine interaction
- For datasets: begin with MCSD 1.0
- For speech-based sarcasm modeling: see Speech-based Sarcasm Detection
- For multimodal text+audio modeling with synthesized data augmentation: see AMuSeD
- For an end-to-end sarcasm recognition tool using acoustic, textual, emotional, and sentiment cues: see SarcEmotiq
- For LLM/MLLM-based sarcasm understanding: see Multimodal LLM Sarcasm Understanding
The first large-scale dataset for sarcasm detection in Mandarin Chinese, combining audio, text, and visual modalities. Built to push sarcasm research beyond English.
- Repo: MCSD 1.0 wiki
- 📦 Dataset: MCSD 1.0 DataverseNL
-
- 📄 Paper: Interspeech 2025
- 🌐 Languages: Chinese Mandarin
Can a machine hear sarcasm? SarcEmotiq detects sarcasm in English audio by combining what is said, how it sounds, and how it feels. It works out of the box on MUStARD++ and supports retraining on your own data.
- 🔗 Repo: SarcEmotiq
- 🚀 Live Demo: Coming soon
- 📄 Paper: JASA 2022
- 🌐 Languages: English
- 🔗 Repo:
coming soon - 🌐 Languages: Chinese
- 🔗 Repo:
coming soon - 🖼️ Slides: 'coming soon'
- 🌐 Languages: Chinese, English
- 🔗 Repo:
coming soon - 🖼️ Paper: Transactions on Affective Computing 2026
- 🌐 Languages: English
- 🔗 Repo:
coming soon - 🖼️ Paper: Interspeech 2022
- 🌐 Languages: English
- E-mail: xiyuan.gao1022@gmail.com
- Google Scholar: [Xiyuan Gao] (https://scholar.google.com/citations?user=Gwl55gQAAAAJ&hl=en)
- LinkedIn: [Xiyuan Gao] (https://www.linkedin.com/in/xiyuangao/)