MatteoFasulo/clef2025_checkthat_task1_subjectivity

CLEF‑2025 CheckThat! Lab Task 1: Subjectivity in News Articles

Systems are challenged to distinguish whether a sentence from a news article expresses the subjective view of the author behind it or presents an objective view on the covered topic instead.

This is a binary classification tasks in which systems have to identify whether a text sequence (a sentence or a paragraph) is subjective (SUBJ) or objective (OBJ).

The task comprises three settings:

• Monolingual: train and test on data in a given language L
• Multilingual: train and test on data comprising several languages
• Zero-shot: train on several languages and test on unseen languages

Datasets statistics

• English
  • train: 830 sentences, 532 OBJ, 298 SUBJ
  • dev: 462 sentences, 222 OBJ, 240 SUBJ
  • dev-test: 484 sentences, 362 OBJ, 122 SUBJ
• Italian
  • train: 1613 sentences, 1231 OBJ, 382 SUBJ
  • dev: 667 sentences, 490 OBJ, 177 SUBJ
  • dev-test - 513 sentences, 377 OBJ, 136 SUBJ
• German
  • train: 800 sentences, 492 OBJ, 308 SUBJ
  • dev: 491 sentences, 317 OBJ, 174 SUBJ
  • dev-test - 337 sentences, 226 OBJ, 111 SUBJ
• Bulgarian
  • train: 729 sentences, 406 OBJ, 323 SUBJ
  • dev: 467 sentences, 175 OBJ, 139 SUBJ
  • dev-test - 250 sentences, 143 OBJ, 107 SUBJ
  • test: TBA
• Arabic
  • train: 2,446 sentences, 1391 OBJ, 1055 SUBJ
  • dev: 742 sentences, 266 OBJ, 201 SUBJ
  • dev-test - 748 sentences, 425 OBJ, 323 SUBJ

Input Data Format

The data will be provided as a TSV file with three columns:

sentence_id sentence label

Where:

• sentence_id: sentence id for a given sentence in a news article
  
• sentence: sentence's text
  
• label: OBJ and SUBJ

Note: For English, the training and development (validation) sets will also include a fourth column, "solved_conflict", whose boolean value reflects whether the annotators had a strong disagreement.

Examples:

b9e1635a-72aa-467f-86d6-f56ef09f62c3 Gone are the days when they led the world in recession-busting SUBJ

f99b5143-70d2-494a-a2f5-c68f10d09d0a The trend is expected to reverse as soon as next month. OBJ

Output Data Format

The output must be a TSV format with two columns: sentence_id and label.

Evaluation Metrics

This task is evaluated as a classification task using F1-macro measure. Other metrics include Precision, Recall, and F1 of the SUBJ class and the macro-averaged scores.

Scorers

The code base with the scorer script is available on the original GitLab repository - clef2025-checkthat-lab-task1.

To evaluate the output of your model which should be in the output format required, please run the script below:

python evaluate.py -g dev_truth.tsv -p dev_predicted.tsv

where dev_predicted.tsv is the output of your model on the dev set, and dev_truth.tsv is the golden label file provided by authors.

The file can be used also to validate the format of the submission, simply use the provided test file as gold data.

Baselines

The code base with the script to train the baseline model is provided in the original GitLab repository - clef2025-checkthat-lab-task1. The script can be run as follow:

python baseline.py -trp train_data.tsv -ttp dev_data.tsv

where train_data.tsv is the file to be used for training and dev_data.tsv is the file on which doing the prediction.

The baseline is a logistic regressor trained on a Sentence-BERT multilingual representation of the data.

Leaderboard

The leaderboard is available in the original GitLab repository - clef2025-checkthat-lab-task1.

Related Work

The dataset was used in AI Wizards at CheckThat! 2025: Enhancing Transformer-Based Embeddings with Sentiment for Subjectivity Detection in News Articles.

Information regarding the annotation guidelines can be found in the following papers:

Federico Ruggeri, Francesco Antici, Andrea Galassi, aikaterini Korre, Arianna Muti, Alberto Barron, On the Definition of Prescriptive Annotation Guidelines for Language-Agnostic Subjectivity Detection, in: Proceedings of Text2Story — Sixth Workshop on Narrative Extraction From Texts, CEUR-WS.org, 2023, Vol 3370, pp. 103 - 111

Francesco Antici, Andrea Galassi, Federico Ruggeri, Katerina Korre, Arianna Muti, Alessandra Bardi, Alice Fedotova, Alberto Barrón-Cedeño, A Corpus for Sentence-level Subjectivity Detection on English News Articles, in: Proceedings of Joint International Conference on Computational Linguistics, Language Resources and Evaluation (COLING-LREC), 2024

Suwaileh, Reem, Maram Hasanain, Fatema Hubail, Wajdi Zaghouani, and Firoj Alam. "ThatiAR: Subjectivity Detection in Arabic News Sentences." arXiv preprint arXiv:2406.05559 (2024).

Credits

ECIR 2025

Alam, F. et al. (2025). The CLEF-2025 CheckThat! Lab: Subjectivity, Fact-Checking, Claim Normalization, and Retrieval. In: Hauff, C., et al. Advances in Information Retrieval. ECIR 2025. Lecture Notes in Computer Science, vol 15576. Springer, Cham. https://doi.org/10.1007/978-3-031-88720-8_68

@InProceedings{10.1007/978-3-031-88720-8_68,
  author="Alam, Firoj
  and Stru{\ss}, Julia Maria
  and Chakraborty, Tanmoy
  and Dietze, Stefan
  and Hafid, Salim
  and Korre, Katerina
  and Muti, Arianna
  and Nakov, Preslav
  and Ruggeri, Federico
  and Schellhammer, Sebastian
  and Setty, Vinay
  and Sundriyal, Megha
  and Todorov, Konstantin
  and V., Venktesh",
editor="Hauff, Claudia
  and Macdonald, Craig
  and Jannach, Dietmar
  and Kazai, Gabriella
  and Nardini, Franco Maria
  and Pinelli, Fabio
  and Silvestri, Fabrizio
  and Tonellotto, Nicola",
title="The CLEF-2025 CheckThat! Lab: Subjectivity, Fact-Checking, Claim Normalization, and Retrieval",
booktitle="Advances in Information Retrieval",
year="2025",
publisher="Springer Nature Switzerland",
address="Cham",
pages="467--478",
isbn="978-3-031-88720-8",
}

CLEF 2025 LNCS

@InProceedings{clef-checkthat:2025-lncs,
  author = {
    Alam, Firoj
    and Struß, Julia Maria      
    and Chakraborty, Tanmoy
    and Dietze, Stefan
    and Hafid, Salim
    and Korre, Katerina
    and Muti, Arianna
    and Nakov, Preslav
    and Ruggeri, Federico
    and Schellhammer, Sebastian
    and Setty, Vinay
    and Sundriyal, Megha
    and Todorov, Konstantin
    and Venktesh, V
  },
  title = {Overview of the {CLEF}-2025 {CheckThat! Lab}: Subjectivity, Fact-Checking, Claim Normalization, and Retrieval},
  editor = {
    Carrillo-de-Albornoz, Jorge and
    Gonzalo, Julio and
    Plaza, Laura and
    García Seco de Herrera, Alba and
    Mothe, Josiane and
    Piroi, Florina and
    Rosso, Paolo and
    Spina, Damiano and
    Faggioli, Guglielmo and
    Ferro, Nicola
  },
  booktitle = {Experimental IR Meets Multilinguality, Multimodality, and Interaction. Proceedings of the Sixteenth International Conference of the CLEF Association (CLEF 2025)},
  year = {2025}
}

CLEF 2025 CEUR papers

@proceedings{clef2025-workingnotes,
    editor = "Faggioli, Guglielmo and
    Ferro, Nicola and
    Rosso, Paolo and
    Spina, Damiano",
    title = "Working Notes of CLEF 2025 - Conference and Labs of the Evaluation Forum",
    booktitle = "Working Notes of CLEF 2025 - Conference and Labs of the Evaluation Forum",
    series = "CLEF~2025",
    address = "Madrid, Spain",
    year = 2025
}

Task 1 overview paper

@inproceedings{clef-checkthat:2025:task1,
  title     = {Overview of the {CLEF-2025 CheckThat!} Lab Task 1 on Subjectivity in News Article},
  author    = {
    Ruggeri, Federico and
    Muti, Arianna and
    Korre, Katerina and
    Stru{\ss}, Julia Maria and
    Siegel, Melanie and
    Wiegand, Michael and
    Alam, Firoj and
    Biswas, Rafiul and
    Zaghouani, Wajdi and
    Nawrocka, Maria and
    Ivasiuk, Bogdan and
    Razvan, Gogu and
    Mihail, Andreiana
  },
  crossref  = {clef2025-workingnotes}
}