Persuasion Classification

Task Overview

Detection and classification of persuasion techniques at the sentence level in three Slavic languages: Bulgarian, Polish and Russian in two text types: (a) parliamentary debates, and (b) disinformation-related social media posts. The task uses a taxonomy of 25 persuasion techniques and builds on prior paragraph-level annotated dataset from the Slavic NLP 2025 Shared Task.

Data & Resources

This shared task is closely related to previous persuasion-technique datasets and benchmarks, but focuses on sentence-level classification in the context of multilingual parliamentary debates and social media. All the resources listed below are relevant because they cover persuasion techniques in similar domains and annotation setups; however, the current task uses sentence-level labels and a different train/test setup.

Training data. Participants may exploit a wide range of existing multilingual datasets with span-level and paragraph-level annotation of persuasion techniques, which were used in recent SemEval and CLEF challenges and cover Bulgarian, Polish, and Russian, among other languages. In addition, the trial data from the previous SlavicNLP 2025 edition is also relevant as a starting point for this task. In particular, one may consider using resources from:

• SemEval 2020 Task 11 described in (Da San Martino et al., 2020).
• SemEval 2021 Task 6 described in (Dimitrov et al., 2021).
• SemEval 2023 Task 10 (incl.: PL, RU) described in (Piskorski et al., 2023).
• CLEF 2024 Task 3 (incl.: BG, PL, RU) described in (Piskorski et al., 2024).
• SemEval 2024 Task 4 (incl.: BG) described in (Dimitrov et al., 2024).

Accessing data from these shared tasks may require registration to the respective task platforms. The list above is not exhaustive.

Data from previous SlavicNLP 2025. The TRAINING DATA in the domain of parliamentary debates in Bulgarian, Polish, Slovene, and in the domain of social media in Russian is now available; it contains approximately 80 documents.

In addition, one may consider exploiting the ParlaMint dataset for domain adaptation, as well as the multilingual parliamentary model XLM-R-parla, pre-trained on parliamentary proceedings.

Test Data. The sentence-level test data for this task will be released in December, and the evaluation setup follows the current shared-task design while focusing on sentence-level classification.

Papers to Cite in Your Paper Describing Your System

As this shared task builds upon existing and published studies and datasets, please cite the papers mentioned below in your paper describing your system.

@inproceedings{piskorski-etal-2026-corpus,
    title = "A Corpus of Persuasion Techniques in {S}lavic Languages",
    author = "Piskorski, Jakub  and
      Dimitrov, Dimitar Iliyanov  and
      Ernst, Marina  and
      Haneczok, Jacek  and
      Marcinczuk, Michal  and
      Modzelewski, Arkadiusz  and
      Yangarber, Roman",
    booktitle = "Proceedings of the Fifteenth Language Resources and Evaluation Conference",
    month = may,
    year = "2026",
    address = "Palma de Mallorca, Spain",
    publisher = "ELRA Language Resource Association",
    url = "https://aclanthology.org/2026.lrec-1.289/",
    doi = "10.63317/46rmmqosdrgt"
}

@inproceedings{piskorski-etal-2025-slavicnlp,
    title = "{S}lavic{NLP} 2025 Shared Task: Detection and Classification of Persuasion Techniques in Parliamentary Debates and Social Media",
    author = "Piskorski, Jakub  and
      Dimitrov, Dimitar  and
      Dobrani{\'c}, Filip  and
      Ernst, Marina  and
      Haneczok, Jacek  and
      Koychev, Ivan  and
      Ljube{\v{s}}i{\'c}, Nikola  and
      Marcinczuk, Michal  and
      Modzelewski, Arkadiusz  and
      Moravski, Ivo  and
      Yangarber, Roman",
    booktitle = "Proceedings of the 10th Workshop on Slavic Natural Language Processing (Slavic NLP 2025)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.bsnlp-1.27/",
    doi = "10.18653/v1/2025.bsnlp-1.27",
    pages = "254--275",
    ISBN = "978-1-959429-57-9"
}

Task Contact

Jakub Piskorski — jpiskorski@gmail.com

See the Call for Participation page for submission guidelines.