DARE – Detection and Recognition of Euphemisms
A Polish–American–Ukrainian project extending neural euphemism detection and interpretation to Polish, Ukrainian and Russian, including new annotated euphemism datasets.
Lead Investigator: Witold Kieraś
- Grant Number
- BPN/NSF/2023/1/00008
- Start Date
- 1 March 2024
- End Date
- 31 March 2026
Project factsheet
| English name: | IMPRESS-U: DARE: Detection and Recognition of Euphemisms |
| Project type: | An IMPRESS-U (International Multilateral Partnerships for Resilient Education and Science System in Ukraine) International Supplement, Polish part financed by the Polish National Agency for Academic Exchange (NAWA) |
| Grant agreement number: | BPN/NSF/2023/1/00008 |
| Duration: | 1 March 2024 – 31 March 2026 |
| Partners: | Montclair State University (USA), Ivan Franko National University of Lviv (Ukraine), Institute of Computer Science, Polish Academy of Sciences |
| Project coordinator (IPI PAN): | Witold Kieraś |
| Related NSF project: | RI: Small: DaRE: Detection and Recognition of Euphemisms (NSF award 2226006, Montclair State University, PI: Anna Feldman) |
Project summary
IMPRESS-U: DARE: Detection and Recognition of Euphemisms is a collaborative project conducted in partnership with researchers from Montclair State University and Ivan Franko National University of Lviv. It aims to extend research on the detection and interpretation of euphemisms – mild or indirect expressions used in place of harsher or more offensive ones – to Ukrainian, Polish and Russian. The initiative focuses on developing advanced algorithms for recognising euphemisms and on investigating the interpretability of neural language models in this context. The project builds on existing datasets in English, Yoruba, Spanish and Mandarin Chinese and combines linguistically motivated methods with transformer-based approaches to address the problem of detecting euphemisms in multilingual data.
IMPRESS-U is an initiative led by the U.S. National Science Foundation together with partner agencies from Poland (NAWA and the National Science Centre), Ukraine, Estonia, Latvia and Lithuania. Within the programme, NAWA funds International Supplements that extend existing NSF-funded projects with partners from Poland and Ukraine.
Results
- A new annotated dataset for euphemism detection in Polish, comprising 2,500 entries. Potentially euphemistic terms were identified in the National Corpus of Polish and the Corpus of Contemporary Polish, using both a Polish dictionary of euphemisms and concordance searches around taboo topics, and their euphemistic and non-euphemistic uses were annotated.
- A unified multilingual euphemism benchmark covering seven languages (including the new Polish and Ukrainian datasets), with zero-shot and few-shot baselines, released under an open license at https://github.com/NLPlabMSU/lrec-euph.
Publication:
- Whitney Poh, Julia Sammartino, Jasper Andrew, Witold Kieraś, Natalia Zawadzka-Paluektau, Iryna Dilai, Libby Barak, Jing Peng, Anna Feldman. 2026. The Multilingual Euphemism Benchmark: Datasets and Baselines for Pragmatic Language Understanding. In Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026), pages 7307–7319.