HU at SemEval-2026 Task 10: 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: Expert, medium

Summary

The "HU at SemEval-2026 Task 10: Psycholinguistic Conspiracy Marker Extraction and Detection" paper details a methodology for combating conspiracy theory content in modern media. This work addresses two distinct subtasks: first, extracting psycholinguistic markers from text using Named Entity Recognition (NER) techniques, and second, classifying Reddit comments as either conspiratorial or non-conspiratorial. The team's approach integrated diverse extraction methodologies, including traditional bio tagging schemes, the GlobalPointer framework, and the GLiNER2 architecture. Furthermore, they employed data augmentation and synthetic data generation through Large Language Models (LLMs) and evaluated various transformer-based models like DistilBERT and Covid Twitter-BERT. The final system achieved a macro F1 score of 0.26 for Subtask 1 and 0.76 for Subtask 2.

Key takeaway

For NLP Engineers developing misinformation detection systems, you should consider a multi-faceted approach combining advanced NER techniques with transformer models. Your efforts to identify psycholinguistic markers can benefit from data augmentation via LLMs, improving model robustness. Evaluate architectures like GlobalPointer or GLiNER2 for marker extraction and fine-tune models such as DistilBERT for classification to achieve competitive performance in complex content moderation tasks.

Key insights

NLP techniques, including NER and transformer models, can detect psycholinguistic markers of conspiracy theories.

Principles

• Diverse NER methods improve marker extraction.
• LLMs enhance data for complex classification tasks.
• Transformer models are effective for text classification.

Method

The approach combines bio tagging, GlobalPointer, and GLiNER2 for marker extraction. It uses LLMs for data augmentation and synthetic data generation, then evaluates transformer models like DistilBERT for classification.

In practice

• Apply NER to identify specific psycholinguistic markers.
• Use LLMs to generate synthetic training data.
• Evaluate DistilBERT or Covid Twitter-BERT for text classification.

Topics

• Named Entity Recognition
• Conspiracy Theory Detection
• Large Language Models
• Transformer Models
• SemEval
• Misinformation Detection

Best for: Research Scientist, AI Scientist, NLP Engineer, Machine Learning 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.
• The F1 of Formula One: Applicability of Pre-trained NER Models to Brazilian TV Interview Transcripts — LLMs offer superior linguistic sensitivity for NER in noisy interview transcripts compared to statistical methods.
• 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.
• Multi-View Attention Multiple-Instance Learning Enhanced by LLM Reasoning for Cognitive Distortion Detection — Combining LLMs with MIL and ELB components improves cognitive distortion detection by enhancing interpretability and reasoning.
• Sem-Detect: Semantic Level Detection of AI Generated Peer-Reviews — AI-generated peer reviews converge semantically, while human reviews, even LLM-refined, retain unique ideas.
• Detecting HIV-Related Stigma in Clinical Narratives Using Large Language Models — LLMs can effectively detect HIV-related stigma in clinical notes, with encoder models outperforming generative models in zero-shot settings.

Open in AIssential →

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