New Greek book retrieval benchmark CUP released

Researchers have introduced CUP, a new benchmark dataset designed for evaluating Greek book retrieval systems. The dataset comprises 868 catalog records and 104 expert-annotated queries with relevance judgments. Experiments show that multilingual embeddings outperform Greek-specific models, and hybrid retrieval methods achieve the best overall performance. Analysis indicates BM25 is effective for named-entity queries, while dense and hybrid approaches improve performance on natural-language, noisy, cross-lingual, and concept-based queries. AI

IMPACT Provides a benchmark for evaluating LLM and embedding-based retrieval systems in a specific domain.

RANK_REASON The item is an academic paper detailing a new dataset and evaluation of retrieval methods. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• BM25
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• Panagiotis Papadakos
• ScienceCast
• sentence-transformers

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →