Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
July 8, 2026Journal of the Association for Information Science and TechnologyOpen Access

Incremental refinement of relevance rankings: Balancing relevance depth and scope

View Full Paper
Ask AI
Bookmark
Share

Authors

MAMüge AkbulutYTYaşar Tonta

Discussion

Loading...

Member takes

Overview

Randomized trial demonstrates improved relevance and diversity in information retrieval, suggesting a refined hybrid method.

Key Points

  • This study aims to enhance information retrieval by balancing relevance and topical diversity in results.
  • Introduced a hybrid method combining latent dirichlet allocation (LDA) with citation-based retrieval, grounded in relevance theory.
  • Applied the method to the iSearch corpus containing 435,000 physics papers across 65 queries.
  • Evaluated relevance and diversity considering degree centrality, subject categories, and maximal marginal relevance.
  • The hybrid method improved normalized discounted cumulative gain (NDCG) performance, facilitating better ranking of both central and peripheral works.
  • Interactive visualizations showed effective fine-tuning of search results in real time, revealing interdisciplinary connections.
  • Sensitivity tests confirmed robustness to hyperparameter variations and query reformulation.

Cite This Study

Akbulut et al. (2026) studied this question.

synapsesocial.com/papers/6a4dea28d2ea289ef6283ff1https://doi.org/10.1002/asi.70101
View Full Paper
Ask AI
Bookmark
Share