My research interests are mainly applying Deep Learning for Natural Language Processing. I’m currently interested in Natural language Generation, including creative text generation applications, better NLG models, and better evaluation method for NLG systems. I am also interested in making large-scale pretrained language models more accessible to both academic and industry applications. To achieve this, I am actively working on Efficient Pretrained Language Modeling and Transfer Learning for NLP. This includes efficiency in terms of both the amount of data and time required for pretraining and fine-tuning pretrained langauge models, as well as the number of parameters and computational cost, which affects the memory requirement and latency for application of pretrained language models.

I am also interested in or working in the following subjects:
Commonsense reasoning and knowledge-based reasoning.
Interpretability, explainability for NLP models, bias and fairness in NLP models, and their interactions.
Low resource NLP and Multilingual Pretrained Language Models.