Behavior-Driven Semantic Re-Ranking for Personalized Web Search
Journal of Technology Research · 13.03.2026 · 4 (1) · 1-11
DOI: 10.26629/jtr.2026.01 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Technology Research, online erschienen am 13.03.2026. Verfasst von Anwar Alhenshiri und 2 weiteren Autor:innen.
Abstract (Originalsprache)
This study proposes a behaviour-driven semantic re-ranking system designed to enhance personalized web search without modifying the underlying search engine infrastructure. The approach operates as a lightweight middleware layer over Google Search and integrates BERT-based semantic embeddings, KeyBERT-driven dynamic user profiling, cosine similarity scoring, and an exponential decay mechanism to maintain adaptive interest modelling. A controlled comparative experiment involving 14 participants was conducted to evaluate performance against the baseline Google ranking. Results indicate that the proposed system reduced average search time by 15.3% (from 274 seconds to 232 seconds), decreased the number of required clicks by 21.2% (from 7.71 to 6.07), and lowered query reformulations from 43% to 21%. Furthermore, 79% of participants preferred the personalized system, with statistically significant improvements observed in efficiency and satisfaction (p < 0.05). These findings demonstrate that behaviour-driven semantic re-ranking can substantially improve search relevance, user efficiency, and overall experience while maintaining transparency and scalable deployment.
Zitieren
APA
Alhenshiri, A., Jelwal, A., & Badesh, H. (2026). Behavior-Driven Semantic Re-Ranking for Personalized Web Search. Journal of Technology Research, 4(1), 1-11. https://doi.org/10.26629/jtr.2026.01
Vancouver
Alhenshiri A, Jelwal A, Badesh H. Behavior-Driven Semantic Re-Ranking for Personalized Web Search. Journal of Technology Research. 2026;4(1):1-11. doi:10.26629/jtr.2026.01
BibTeX
@article{Alhenshiri2026,
author = {Alhenshiri, Anwar and Jelwal, Alaa and Badesh, Hoda},
title = {Behavior-Driven Semantic Re-Ranking for Personalized Web Search},
journal = {Journal of Technology Research},
year = {2026},
volume = {4},
number = {1},
pages = {1--11},
doi = {10.26629/jtr.2026.01}
}
EUVIMED