Task 1 consists in distinguishing 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.
Task 4 consists of two subtasks:
- Subtask 4a (Scientific Web Discourse Detection): Given a social media post (tweet), detect if it contains (1) a scientific claim, (2) a reference to a scientific study / publication, or (3) mentions of scientific entities, e.g. a university or scientist.
- Subtask 4b (Scientific Claim Source Retrieval): Given an implicit reference to a scientific paper, i.e., a social media post (tweet) that mentions a research publication without a URL, retrieve the mentioned paper from a pool of candidate papers.
Challenge website: https://checkthat.gitlab.io/clef2025/task4/
Challenge GitLab: https://gitlab.com/checkthat_lab/clef2025-checkthat-lab
- Package management with Poetry (install poetry, then poetry install) then poetry shell to activate the virtual environment).
- Semantic versioning using Semantic Release.
- Commit messages using Conventional Commits (make sure your commit messages are formatted correctly).
- Use commitzen to create commit messages (install commitizen, then git cz to create a commit message).
- SciTweets: https://github.com/AI-4-Sci/SciTweets/tree/main
- SciWebClaims: https://github.com/AI-4-Sci/SciWebClaims (can be used for the science content identification task)
- Data for the task 4b is extracted from https://github.com/allenai/cord19.
- Task 1: E5 + MLP classifier.
- Task 4a: Consider SetFit / Look at nvidia/domain-classifier.
- Task 4b: Consider using re-ranker models (see https://huggingface.co/blog/train-reranker).
If you use this software, please cite (bib file):
Grégoire Burel, Pasquale Lisena, Enrico Daga, Raphael Troncy, Harith Alani.
ClimateSense at CheckThat! 2025: Combining Fine-tuned Large Language Models and Conventional Machine Learning Models for Subjectivity and Scientific Web - Discourse Analysis.
In: CLEF 2025 Working Notes, Ceur-WS Sep 2025, Madrid, Spain.