Word2winners at SemEval-2025 Task 7: Multilingual and Crosslingual Fact-Checked Claim Retrieval

2025-03-12

AI Summary

This paper addresses SemEval 2025 Task 7—multilingual fact-checking retrieval—by retrieving semantically matching, human-verified fact checks for given social media claims from the multilingual MultiClaim dataset. To overcome zero-shot cross-lingual retrieval bottlenecks, we propose a supervised fine-tuning framework integrating machine translation: leveraging mBERT or XLM-R as backbones, we introduce back-translation–enhanced cross-lingual alignment and bilingual collaborative optimization. Our method simultaneously improves retrieval accuracy in both monolingual and cross-lingual settings, achieving 92% and 85% accuracy, respectively—substantially outperforming baselines and ranking among the top systems on the official leaderboard. The core contribution lies in explicitly modeling translation fidelity within the retrieval fine-tuning process, thereby unifying semantic alignment and domain adaptation into a single optimization objective.

Technology Category

• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

This paper describes our system for SemEval 2025 Task 7: Previously Fact-Checked Claim Retrieval. The task requires retrieving relevant fact-checks for a given input claim from the extensive, multilingual MultiClaim dataset, which comprises social media posts and fact-checks in several languages. To address this challenge, we first evaluated zero-shot performance using state-of-the-art English and multilingual retrieval models and then fine-tuned the most promising systems, leveraging machine translation to enhance crosslingual retrieval. Our best model achieved an accuracy of 85% on crosslingual data and 92% on monolingual data.

Problem

• Retrieve relevant fact-checks for claims
• Use multilingual and crosslingual retrieval models
• Achieve high accuracy on crosslingual and monolingual data

Innovation

• Utilized multilingual retrieval models
• Fine-tuned with machine translation
• Achieved high crosslingual accuracy