Meet Professor Katrin Erk

Content

The new names and faces in the College of Humanities and Fine Arts this fall aren’t just those belonging to our students. We are proud to welcome ten new faculty members across seven different departments of the college, as well as five new folks in the UMass Amherst Writing Program. 
Over the course of the semester, we’re introducing you to these faculty members and to the work they do.

For those of us who aren’t sure what computational linguistics is, Wikipedia offers the following: “Computational linguistics is an interdisciplinary field concerned with the computational modelling of natural language, as well as the study of appropriate computational approaches to linguistic questions.” 
Given that definition, it’s no surprise that Professor Katrin Erk — who holds a joint position as a Professor in Linguistics and the Manning College of Information and Computer Sciences — is excited about the opportunities for inter-disciplinary work in the College of Humanities and Fine Arts. Linguistics, computer science, philosophy, artificial intelligence and more come into play in Erk’s field, and she has already begun to seek out collaborators within our academic community.

Question:  What is your academic background?
Professor Erk: I got my undergraduate degree in computer science, with a minor in computational linguistics, at Koblenz University in Germany in 1998, and my Ph.D., also in computer science, at Saarland University in Germany in 2002. After a postdoc in the computational linguistics center at Saarland University, I joined the linguistics department at the University of Texas at Austin in 2006. I got a joint appointment between linguistics and computer science there in 2023.  

Question:  What is your teaching experience? 
Professor Erk: I've always enjoyed helping others understand complex mathematical ideas. This led me to co-write an introductory textbook on theoretical computer science (formal languages and automata theory) with my undergrad advisor, Lutz Priese. We wrote the book while I was still an undergraduate student; it came out in 2000. At the University of Texas at Austin, I taught classes in computational linguistics and natural language processing in the linguistics and computer science departments. Among them was a course on computational semantics: How can you describe, and computationally represent, the meanings of words and sentences? How can the meanings of words be put together to form the meanings of sentences? There are very different approaches to that, either based on logic or based on distributed representations, and it's particularly fun to look at both of them. 
 
Question: What are you passionate about when it comes to this work? 
Professor Erk: I want to understand the lexicon. The lexicon is vast, and so varied — there are concrete and abstract concepts, verbs that come with an argument structure and complex temporal dynamics, words that seem to carry large cultural frames with them and cannot be understood without them. What representations do we need to encompass all of this?  

Question: Have you published, exhibited, or conducted research? Please share a bit about it, and feel free to include any relevant links. 
Professor Erk: Yes, both more computational and more theoretical work. I've long worked with so-called distributional models, automatically generated representations of word meaning that characterize a word through the contexts in which it has been observed. 
Here is an early paper where I think about what these models have to say about word meaning:  Erk 2010, "What is word meaning, really? (And how can distributional models help us describe it?)", Workshop on Geometrical Models of Natural Language Semantics.    
More recently, I've looked at how internal representations of recent AI models can be used to give us a glimpse of word meaning in context, something like a partial decomposition of word meaning: Chronis, Mahowald & Erk 2023, "A Method for Studying Semantic Construal in Grammatical Constructions with Interpretable Contextual Embedding Spaces", Proceedings of the Annual Meeting of the Association for Computational Linguistics.  
On the theoretical side, I've argued for a view of word meaning in context that is influenced by multiple interacting, probabilistic constraints, including constraints from stored world knowledge: Erk & Herbelot 2024, "How to Marry a Star: Probabilistic Constraints for Meaning in Context", Journal of Semantics  
And here is a recent paper that best summarizes my view of what the lexicon as a whole may look like: Jackendoff & Erk 2025, "Towards a Deeper Lexical Semantics", Topics in Cognitive Science 

Question: What are you most looking forward to at UMass? 
Professor Erk: I look forward to working together with fantastic and incredibly interesting researchers, both faculty and students. I have already started collaborating with some of my new colleagues, and it is wonderful. 

Question: How do you hope to engage with the HFA community? 
Professor Erk: In teaching, I find the intersection of artificial intelligence and the humanities important and fruitful. At the University of Texas, I co-taught a class with a colleague in Anthropology about Humans and AI, and it was a great experience that taught me a lot (and, hopefully, the students too).  
The work I do on the nature of the lexicon is very interdisciplinary, and I am very much looking forward to discussing it with colleagues in linguistics, philosophy, and other disciplines.