qhjqhj00 semeval2021task6-st1
SemEval-2021 Task 6: Detection of Persuasion Techniques in Texts and Images — Dimitrov et al. (2021) (arXiv:2105.09284, 2021)
What this evaluates
Detects persuasion techniques in unimodal text by classifying which techniques are present in a given text snippet. It probes the model's ability to perform multilabel classification on propaganda and persuasive content.
Datasets
- SemEval-2021 Task 6 Subtask 1 — total ?; splits: test (-1)
Metrics
F1-Micro(primary) — range: [0, 1]- Micro-averaged F1 score computed over all persuasion technique labels across the dataset.
F1-Macro— range: [0, 1]- Macro-averaged F1 score computed over all persuasion technique labels.
Input / output format
Input: Raw text snippet.
Output: Multilabel classification: a set of persuasion technique labels (e.g., Loaded Language, Repetition) applicable to the text.
Scoring recipe
tp = fp = fn = 0
for p, g in zip(predictions, golds):
for label in all_labels:
if label in p and label in g: tp += 1
elif label in p and label not in g: fp += 1
elif label not in p and label in g: fn += 1
prec = tp / (tp + fp) if (tp + fp) > 0 else 0
rec = tp / (tp + fn) if (tp + fn) > 0 else 0
return 2 * prec * rec / (prec + rec) if (prec + rec) > 0 else 0
Common pitfalls
- Using image features for Subtask 1 is strictly prohibited and leads to disqualification.
- Class imbalance is significant; a majority-class baseline always predicts 'Loaded Language'.
Evidence (verbatim from paper)
The evaluation results are shown in Table 3, which also includes two baselines: (i) random, and (ii) majority class. The latter always predicts Loaded Language, as it is the most frequent technique for Subtask 1... Table 3: Results for Subtask 1. The systems are ordered by the official score: F1-micro.
Citation
@misc{dimitrov2021semevaltask6,
title={SemEval-2021 Task 6: Detection of Persuasion Techniques in Texts and Images},
author={Dimitrov et al. (2021)},
year={2021},
note={arXiv:2105.09284}
}
- arXiv: 2105.09284