Papers by Rupak Sarkar
Are Neural Topic Models Broken? (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing evaluation paradigms are often divorced from real-world use . recent results have challenged the validity of the prevailing model evaluation paradigm . |
| Approach: | They show that neural topic models fare worse in both respects compared to an established classical method. |
| Outcome: | The proposed method outperforms the members of the ensemble in both respects. |
Pregnant Questions: The Importance of Pragmatic Awareness in Maternal Health Question Answering (2024.naacl-long)
Copied to clipboard
Neha Srikanth, Rupak Sarkar, Heran Mane, Elizabeth Aparicio, Quynh Nguyen, Rachel Rudinger, Jordan Boyd-Graber
| Challenge: | a question-answering system must address pragmatic inferences to answer usefully, says a new study . human information needs are often inferred from the surface form, but answers must address the pragmatic needs of the question. |
| Approach: | They examine assumptions and implications made when mothers ask questions . they find that incorporating these inferences into QA pipelines produces more complete answers . |
| Outcome: | a study shows that incorporating inferences from questions helps to address harmful beliefs . human needs vary when asking questions, but a complete answer can address them . a QA pipeline can be more effective in addressing these needs, the study finds . |
Social Media Attributions in the Context of Water Crisis (2020.emnlp-main)
Copied to clipboard
| Challenge: | In this paper, we analyze social media discussions to identify attribution factors for natural disasters/collective misfortunes. |
| Approach: | They propose a task of attribution tie detection to identify factors held responsible for a water crisis in a social media document. |
| Outcome: | The proposed task can be performed on a dataset constructed from YouTube comments on 2,500 videos relevant to the 2019 Chennai water crisis. |
Understanding Common Ground Misalignment in Goal-Oriented Dialog: A Case-Study with Ubuntu Chat Logs (2025.acl-long)
Copied to clipboard
| Challenge: | a misalignment or misunderstanding can disrupt communication, leading to confusion or conflict. |
| Approach: | They study failures of grounding in Ubuntu IRC datasets to identify misalignments . they find disruptions in conversational flow are driven by a divergence in beliefs . |
| Outcome: | The findings show that misalignment in common ground can disrupt communication . the study also shows that miscommunications can lead to confusion or conflict . |
PairScale: Analyzing Attitude Change with Pairwise Comparisons (2025.findings-naacl)
Copied to clipboard
| Challenge: | a text-based framework for measuring attitudes in communities is proposed . the framework uses both implicit and explicit evidence in language to characterize attitudes . |
| Approach: | They propose a text-based framework for measuring attitudes in communities toward issues of interest using language. |
| Outcome: | The proposed framework is validated by examining attitudes on two high-profile issues in the u.s. |
Natural Language Decompositions of Implicit Content Enable Better Text Representations (2023.emnlp-main)
Copied to clipboard
| Challenge: | a new method for the analysis of text takes implicitly communicated content into account . authors: human interpretation of each individual utterance is intractable . |
| Approach: | They propose a method that takes explicit communication into account when people interpret text . they use a large language model to generate propositions that are inferentially related to the text analyzed . |
| Outcome: | The proposed method proves useful in multiple problems that involve interpretation of utterances . it uncovers high-level narratives in public commentary, which are often not expressed in surface forms . |
Measuring scalar constructs in social science with LLMs (2025.emnlp-main)
Copied to clipboard
Hauke Licht, Rupak Sarkar, Patrick Y. Wu, Pranav Goel, Niklas Stoehr, Elliott Ash, Alexander Miserlis Hoyle
| Challenge: | Valid scalar measurement of skalar constructs is a fundamental task in text analysis. |
| Approach: | They evaluate four approaches to measuring scalar constructs using large language models . pairwise comparisons produced better measurements than prompting LLMs, they say . validation of skalar measurement enables wide range of substantive applications in social science research . |
| Outcome: | The proposed methods improve on pairwise comparisons and finetuning . the proposed methods can be used in social science research . |