Research Interests
My research focuses on natural language processing that utilizes computers to understand the text as human beings. In particular, I am interested in general functions and models that can be easily applied to most cases to improve results, instead of aiming for concrete applications, such as text summarization and knowledge graph.

For instance, I am interested in the text degeneration problem, i.e., the appearance of bland and repetitive texts when using likelihood as a decoding objective for deep neural networks (Holtzman et al. 2020).

Additionally, I attempt to solve the overfitting problem caused by the scarcity of labeled data for training deep neural networks.

I am currently working on the representation degeneration problem that the overall similarity between token embeddings is increasing during training and causes the expressiveness of these embeddings to decrease (Gao et al. 2019).