New study evaluates IR model robustness to collection growth

A new study published on arXiv investigates the robustness of Information Retrieval (IR) models when faced with growing document collections. Researchers Emmanouil Georgios Lionis and colleagues formalized the concept of IR model effectiveness not decreasing with the addition of non-relevant documents. Their experiments, which involved merging two distinct collections, revealed that neither Multi-Document-Agnostic (MDA) nor Multi-Document-Dependent (MDD) models are entirely immune to performance degradation when non-relevant documents are added. The study found that MDA models were more effective for retrieval, while both MDA and MDD rerankers showed similar effectiveness. AI

IMPACT This research could lead to more efficient and stable information retrieval systems as datasets grow.

RANK_REASON Academic paper on information retrieval models. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.IR (Information Retrieval) →

• alphaXiv
• arXiv
• BM25
• CatalyzeX
• DagsHub
• Emmanouil Georgios Lionis
• Gotit.pub
• Hugging Face
• Influence Flower
• information retrieval
• ScienceCast

• paper
• other

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