Charlie Welch

I am a research group leader working on coversation modeling, personalization, controllable generation with large language models and modelling subjective differences across populations for classification tasks, for answering questions such as what is humorous, sarcastic, offensive, or moral. For more information regarding the topics that I am looking to supervise masters theses, check below the areas of interest.

Some of my work is focused on predicting conversational behavior (mainly what people will say) in longitudinal dialog using insights from conversations. Moreover, I am more broadly interested in dialog systems and personalization. I have been looking into ways to create personalized language models when different volumes of data and different types of metadata are available. In the past, I have also worked on parsing, and sentiment analysis. Recently, I’ve been looking for ways to apply natural language processing to help reduce anxiety surrounding COVID-19, mitigate bias in language models, and how to customize language models for mental health and misinformation applications.

Areas of Interest

- Conversation Modeling
- Personalization
- Perspectivist NLP i.e. Modeling subjective differences
- Controllable Generation