Rupak Sarkar
I am a fourth year PhD. student at the University of Maryland,
College Park in the department of Computer Science in the amazing
CLIP lab
with Prof. Philip Resnik.
My area of research is computational pragmatics and conversational AI, with a special interest in the role of
implicit content. Current NLP methods focus (too) heavily on the surface form of text, but a large part
of the communicative intent of humans is implicit.
I am interested in understanding how to model and interpret these implicit aspects of human communication, with
specific applications in modeling common ground in conversations, and understanding the role of implicit content
in text.
Mailto : rupak@umd.edu, rupaksarkar.cs@gmail.com
More Info : [CV], [Google Scholar]
Recent updates
- 2025: Our paper on common ground misalignment in goal-oriented dialog was accepted at ACL 2025 Main!(Link to [paper] and [code]). We find that while LLMs can be efficient at finding overt sources of friction, they still struggle to explain common ground misalignment when it's not explicit.
- 2025: I am heading back to Microsoft Research as a Research Intern! I'll be working jointly with MSR and the OfficeAI team on understanding user telemetry data with LLMs.
- 2025: Our paper on measuring scalar constructs in social science with LLMs was accepted at EMNLP 2025 Main! (Paper and code incoming!). We show that pairwise comparison might not always outperform discrete scoring, and that finetuning a smaller model can still be a better option than prompting for construct measurement.
Selected Publications
- Understanding Common Ground Misalignment in Goal-Oriented Dialog: A Case-Study with Ubuntu Chat Logs Rupak Sarkar, Neha Srikanth, Taylor Hudson, Claire Bonial, Rachel Rudinger, Philip Resnik ACL 2025 Main [pdf]
- Measuring scalar constructs in social science with LLMs Hauke Licht*, Rupak Sarkar*, Patrick Y. Wu, Pranav Goel, Niklas Stoehr, Elliott Ash, Alexander Miserlis Hoyle* EMNLP 2025 Main
- Conversational User-AI Intervention: A Study on Prompt Rewriting for Improved LLM Response Generation Rupak Sarkar, Bahereh Sarrafzadeh, Nirupama Chandrasekharan, Nagu Rangan, Philip Resnik, Longqi Yang, Sujay Kumar Jauhar (Under Review at ARR) [pdf]
- PairScale: Analyzing Attitude Change with Pairwise Comparisons Rupak Sarkar, Patrick Wu, Kristina Miler, Alexander Hoyle, Philip Resnik NAACL 2025 Findings [pdf]
- Pregnant Questions: The Importance of Pragmatic Awareness in Maternal Health Question Answering Neha Srikanth*, Rupak Sarkar*, Rachel Rudinger, Jordan Boyd-Graber NAACL 2024 Main [pdf]
- Natural Language Decompositions of Implicit Content Enable Better Text Representations. Alexander Hoyle*, Rupak Sarkar*, Pranav Goel, Philip Resnik EMNLP 2023 Main [pdf]
- We Don't Speak the Same Language: Interpreting Polarization Through Machine Translation. A. R. KhudaBukhsh*, Rupak Sarkar*, Mark S. Kamlet, Tom M. Mitchell
(AAAI) 2021 [slides]
- Are chess discussions racist? An Adversarial Hate Speech Data Set (Student Abstract). Rupak Sarkar, A. R. KhudaBukhsh
35th AAAI Conference on Artificial Intelligence AAAI 2021 Best Student Abstract [poster]
- Social Media Attributions in the Context of Water Crisis. Rupak Sarkar*, Sayantan Mahinder*, Hirak Sarkar, A. R. KhudaBukhsh
Empirical Methods in Natural Language Processing (EMNLP), 2020
- The Non-native Speaker Aspect: Indian English in Social Media. Rupak Sarkar, Sayantan Mahinder, A. R. KhudaBukhsh
6th Workshop on Noisy User-generated Text (W-NUT), EMNLP 2020 [slides]
Selected Media Coverage