zhangpeng at SemEval-2026 Task 10: PsyCoMark - Psycholinguistic Conspiracy Marker Extraction and Detection

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, quick

Summary

The zhangpeng team at SemEval-2026 Task 10 presented PsyCoMark, a system for psycholinguistic conspiracy marker extraction and conspiracy detection from English texts. Their approach addresses two subtasks: extracting conspiracy-related markers (actor, action, effect, victim, evidence) and binary classification for conspiracy content. The system relies on fine-tuning pre-trained transformer encoders, specifically multilingual DistilBERT variants and DeBERTa-v3, without using external corpora or data augmentation. Experimental results show the best models achieved a macro-F1 score of 0.1476 for Subtask 1 and a Weighted-F1 score of 0.7267 for Subtask 2, demonstrating that simple fine-tuning provides a strong baseline for both tasks.

Key takeaway

For NLP Engineers developing content moderation systems, this work demonstrates that fine-tuning pre-trained transformer encoders like DistilBERT or DeBERTa-v3 offers a robust baseline for both psycholinguistic marker extraction and conspiracy detection. You should consider this straightforward approach before investing in complex data augmentation or external corpora, especially when rapid deployment or resource constraints are factors.

Key insights

Fine-tuned transformer encoders offer a strong baseline for psycholinguistic conspiracy marker extraction and detection.

Principles

• Fine-tuning pre-trained models establishes a strong baseline.

Method

Fine-tuning pre-trained transformer encoders (DistilBERT, DeBERTa-v3) without external data for marker extraction and binary text classification.

In practice

• Apply fine-tuned transformers for entity extraction.
• Use transformer models for binary text classification.

Topics

• SemEval-2026
• Conspiracy Detection
• Psycholinguistic Markers
• Transformer Encoders
• DistilBERT
• DeBERTa-v3
• Text Classification

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• mdok-style at SemEval-2026 Task 10: Finetuning LLMs for Conspiracy Detection — Data augmentation and self-training can effectively finetune LLMs for specialized text classification tasks with limited data.
• RacismoBR: A Manually Annotated Dataset for Racist Discourse Detection in Brazilian Portuguese — Culturally grounded annotation is more critical than model architecture for effective racism detection.
• An End-to-End Hybrid Framework for Rumour Detection in Low-Resources Algerian Dialect — A hybrid framework combining transformer embeddings with classical classifiers enables effective rumour detection in low-resource Algerian dialect.
• Muted: Multilingual Targeted Offensive Speech Identification and Visualization — Muted identifies targeted offensive speech using transformer models and spaCy, requiring no fine-tuning for toxic span detection.
• A Multimodal Framework for Financial Fake News Detection for Brazilian Portuguese — Multimodal analysis, integrating text and image-extracted text, significantly improves financial fake news detection.
• A Resource for Enthymeme Detection in Controversial Political Discourse — The study of annotator disagreement in enthymeme detection improves model performance and understanding of human inference.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.