Charlie F. Welch

prof_pic.jpg

04C21 Fachbereich 12

Hans-Meerwein-Straβe 6

35043 Marburg Germany

I am a Hessian.AI Junior Research Group Leader and Alexander von Humboldt Postdoctoral Research Fellow at the University of Marburg. I am collaborating with the CAISA lab led by Prof. Lucie Flek. I received my PhD at the University of Michigan, advised by Prof. Rada Mihalcea, where I worked with the LIT lab.

My research focuses on the frontiers of customizable language generation and models of perception. I am interested in how attributes of text, such as toxicity and empathy, interact and can be used for mutual improvement. I will be investigating how these techniques can be used to study stylistic attributes as well as populations. By better modeling both people and properties of language, I hope to both make language technology more useful and to better understand how diverse groups of people view the world. With a heightened understanding of who we are solving problems for, we can more sustainably train models, reducing environmental costs and supporting a more inclusive research community.

news

Mar 31, 2023 Our follow-up paper on EI was accepted to JMIR. Expressive Interviewing Agents to Support Health-Related Behavior Change: A Study of COVID-19 Behaviors.
Oct 26, 2022 Two more papers were accepted to workshops at EMNLP! Understanding Interpersonal Conflict Types and their Impact on Perception Classification and Nearest Neighbor Language Models for Stylistic Controllable Generation.
Oct 13, 2022 Two of our papers were accepted to EMNLP: A Critical Reflection and Forward Perspective on Empathy and Natural Language Processing and Unifying Data Perspectivism and Personalization: An Application to Social Norms.
Jul 13, 2022 Allison Lahnala will be presenting our paper, Mitigating Toxic Degeneration with Empathetic Data: Exploring the Relationship Between Toxicity and Empathy, at NAACL in Seattle.
May 23, 2022 I will be presenting our paper, Leveraging Similar Users for Personalized Language Modeling with Limited Data, at ACL in Dublin.