Dialogue summarization and NLP evaluation
I joined NUS in September 2021 after completing my Ph.D. at USC, working with Prof. Haizhou Li.
- FacEval: developed evaluation of factual consistency in dialogue summarization, releasing annotated data and an evaluation toolkit. First author, EMNLP 2022.
- EvalRank: proposed a method for evaluating word and sentence representations, testing its alignment with downstream performance across more than 60 models. First author, ACL 2022.
- InstructDS: built a dialogue summarizer tuned on synthesized query-based instruction data. First author, EMNLP 2023.
This work connected representation learning with the practical question of how to tell whether a model’s output preserves meaning.
Publications
- Zero-Shot Relation Classification Through Inference on Category Attributes — IEEE TNNLS 2024
- Relational Sentence Embedding for Flexible Semantic Matching — RepL4NLP 2023
- Speech-Aware Multi-Domain Dialogue State Generation with ASR Error Correction Modules — DSTC 2023
- Enhancing Black-Box Few-Shot Text Classification with Prompt-Based Data Augmentation — arXiv 2023
- Just Rank: Rethinking Evaluation with Word and Sentence Similarities — ACL 2022
- Analyzing and Evaluating Faithfulness in Dialogue Summarization — EMNLP 2022
- Generate, Discriminate and Contrast: A Semi-Supervised Sentence Representation Learning Framework — EMNLP 2022
- A Focused Study on Sequence Length for Dialogue Summarization — arXiv 2022
Students Supervised
- Yan Xiao — Ph.D., Donghua University, China (2023-02 – 2024-02). Relation extraction from text. (IEEE TNNLS)
Academic Services
- Reviewer: IEEE/ACM TASLP, ACL ARR, NAACL, EMNLP, ICME
- Reviewer: Nature Human Behaviour 2022
- Session Chair: IJCNN 2021
- Organizational Volunteer: IALP 2022
- Organizational Volunteer: CLSW 2023