Slavic NLP 2027 Shared Tasks
Two shared tasks are planned for Slavic NLP 2027 and Polish Evaluation Campaign (PolEval).
1. Persuasion Techniques Detection & Classification (2nd edition)
Detection and classification of persuasion techniques at the sentence level in three Slavic languages — Bulgarian, Polish and Russian (to be extended) — in two text types: (a) parliamentary debates, and (b) disinformation-related social media posts. Uses a taxonomy of 25 persuasion techniques and builds on prior span-level annotated datasets (SemEval-2020 Task 11, SemEval-2021 Task 6, SemEval-2023 Task 3, and the Slavic NLP 2025 Shared Task). Given the popularity of the 1st edition, high participation is expected.
2. Disinformation Detection
Binary classification of texts as disinformation vs. non-disinformation, covering Polish, Slovak and Russian (to be extended), building on existing resources such as MIPD for Polish. English-language data (MultiDis, MALINT) is also included to encourage broader participation. Co-organized with NASK National Research Institute, a leading Polish institution involved in monitoring and countering disinformation.
3. PolEval
The next edition of the well-established, long-running evaluation campaign dedicated to advancing and benchmarking Polish language processing and generation. Includes a task on recognition and classification of document layout elements from page images, with further tasks to be announced. Learn more at poleval.pl.
How to Participate
- Read the task description in the Call for Participation.
- Download the training and development data.
- Submit predictions on the test set before the deadline.
- Submit a system-description paper for the SlavicNLP proceedings. More details soon