Researchers from Team DUTH have explored multilingual humour-aware information retrieval using the CLEF 2025 JOKER Task 1 benchmark, which assesses humour retrieval in English and Portuguese. Their approach integrates multilingual XLM-RoBERTa-based dense retrieval with neural re-ranking to gauge how well general-purpose Transformer models can grasp humour-specific relevance. The study found significant cross-lingual performance differences, with Portuguese runs showing stronger results than English runs, indicating limitations in purely semantic dense representations for humour retrieval, especially when humour relies on surface-level cues not captured by multilingual encoders. AI
IMPACT Highlights limitations of current dense retrieval models for humour-specific linguistic phenomena, suggesting future research directions for more nuanced AI understanding.
RANK_REASON The cluster contains an academic paper detailing a new approach to a specific AI research problem. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
