Research

Primary Work: Language Reasoning and Understanding through Dense Paraphrasing

Language reasoning and understanding is a fundamental task in NLP. We, as humans, interpret sentences as contextualized components of a narrative or discourse, by both filling in missing information, and reasoning about event consequences. My work focuses on Dense Paraphrasing (DP), a linguistically-motivated generic strategy for the enrichment of the expression through both its lexical semantics and its dynamic contribution to the text in the whole narrative. We apply DP to advance many downstream tasks, including question answering, coreference resolution, and event tracking.

1. Apply basic DP techniques such as text densification and expansions for knowledge discovery and logical metonymy.

- Exploration and Discovery of the COVID-19 Literature through Semantic Visualization. Tu, J., Verhagen, M., Cochran, B.H., and Pustejovsky, J. In Proceedings of the NAACL-HLT 2021 SRW.
- COVID-19 Literature Knowledge Graph Construction and Drug Repurposing Report Generation. Wang, Q., Li, M., Wang, X., …, Tu, J. et al. Proceedings of NAACL-HLT 2021: Demonstrations. (🏆 Best Demo Award)
- Interpreting Logical Metonymy through Dense Paraphrasing. Ye, B.^*, Tu, J.^*, Jezek, E., and Pustejovsky, J. In Proceedings of the Annual Meeting of the Cognitive Science Society, 44 (CogSci 2022). (*equal contribution)

2. Improve the reasoning capabilities of language models on hidden semantics through question answering and paraphrase generation.

- Competence-based Question Generation. Tu, J., Rim, K., and Pustejovsky, J. In Proceedings of the COLING 2022.
- 2022. SemEval-2022 Task 9: R2VQ – Competence-based Multimodal Question Answering. Tu, J., Holderness, E, Maru, M, Conia, S, Rim, K, Lynch, K, Brutti, R, Navigli, R, and Pustejovsky, J. In Proceedings of the SemEval-2022. Association for Computational Linguistics. (🏆 Honorable Mention Best Task)
- Dense Paraphrasing for Textual Enrichment. Tu, J.^*, Rim, K.^*, Holderness, E, Ye, B., and Pustejovsky, J. In Proceedings of the IWCS 2023. (*equal contribution)
- Linguistically Conditioned Semantic Textual Similarity. Tu, J., Xu, K., Yue, L., et al. In Proceedings of the ACL 2024.

3. Apply DP on entity and event coreference to improve downstream tasks.

- The Coreference under Transformation Labeling Dataset: Entity Tracking in Procedural Texts Using Event Models. Rim, K.^*, Tu, J.^*, Ye, B., Verhagen, M., Holderness, E, and Pustejovsky, J. In Proceedings of the Findings of ACL 2023. (*equal contribution)
- Scalar Anaphora: Annotating Degrees of Coreference in Text. Ye, B., Tu, J., Rim, K., and Pustejovsky, J. In Proceedings of The Sixth Workshop on Computational Models of Reference, Anaphora and Coreference (CRAC 2023).
- Media Attitude Detection via Framing Analysis with Events and their Relations. Zhao, J., Tu, J., Du, H., and Xue, N. In Proceedings of the EMNLP 2024.

More Work on Dense Paraphrasing

Apply DP to produce structured textual representations compatible with various data formats and knowledge representations, such as extending AMR with event structures, and textualizing non-verbal information.

- Dense Paraphrasing for Multimodal Dialogue Interpretation Tu, J., et al. (Ongoing).
- GLAMR: Augmenting AMR with GL-VerbNet Event Structure. Tu, J., Obiso, T., et al. In Proceedings of the LREC-COLING 2024.
- Common Ground Tracking in Multimodal Dialogue. Khebour, I., Lai, K., …, Tu, J., et al. In Proceedings of the LREC-COLING 2024.

Other Interesting Projects

- Reproducing Neural Ensemble Classifier for Semantic Relation Extraction in Scientific Papers. Rim, K., Tu, J., Lynch, K. and Pustejovsky, J. In Proceedings of the LREC 2020.
- Identifying Clinically And Functionally Distinct Groups Among Healthy Controls And First Episode Psychosis Patients By Clustering On EEG Patterns. Qu, X., Liukasemsarn S., Tu, J., Higgins, A., Hickey, T. and Hall, M. Frontiers in Psychiatry Neuroimaging and Stimulation.
- TMR: Evaluating NER Recall on Tough Mentions. Tu, J. and Lignos, C. In Proceedings of the Student Research Workshop at the EACL 2021.
- Evaluating Retrieval for Multi-domain Scientific Publications. Ide, N., Suderman, K, Tu, J. et al. In Proceedings of the LREC 2022.
- HaRMoNEE at SemEval-2024 Task 6: Tuning-based Approaches to Hallucination Recognition. Obiso, T., Tu, J., and Pustejovsky, J. In Proceedings of the SemEval 2024. (🏆 #1 Winning Team)