LLM-Powered Framework to Explore Summarized Aggregated Multimedia Vertical Web Search Results
International Journal of Innovations in Science and Technology · 26.04.2026 · 1
DOI: 10.33411/ijist/1777 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Innovations in Science and Technology, online erschienen am 26.04.2026. Verfasst von Muhammad Wajeeh Uz Zaman und 2 weiteren Autor:innen. Die Arbeit nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The exponential growth of multimedia content has shifted users’ information-seeking behavior from lookup-based to exploratory search. To aid exploration, search engines adopted two prominent approaches: presenting results in verticals (web images, videos, news) and integrating Generative AI (GenAI) to enable rapid comprehension. However, integrations like GenAI primarily focus on lookup search by providing basic text summaries of top results, which also hinder users’ ability to explore information through multimedia. Consequently, users make additional navigation efforts (clicking, scrolling, switching verticals), hindering information exploration. In this approach, we propose a framework that summarizes vertical search results into comprehensive documents. The framework is powered by a large language model (LLM) that extracts topics from search results and groups semantically similar multimedia results across verticals into unified topic-based summaries. This unified interaction reduces users' navigation effort and increases interest in exploration. We evaluated our approach using ROUGE across three domains (Movies, Music, and Sports) and conducted a system usability study with 31 participants, using the Bing search engine as a baseline. The proposed system achieved an average ROUGE F1 at: R-1 = 0.67 ± 0.15, R-2 = 0.26 ± 0.17, R-L = 0.60 ± 0.20. The navigation efforts were significantly reduced in terms of clicks (21.5 vs. 30.8, p < 0.01), scrolls (74.6 vs. 218.4, p < 0.001), and vertical switches (0 vs. 3.7). The average system usability score was reported at 88%, significantly higher than baseline (77%, p < 0.05). These results confirm that our framework reduces exploratory navigation efforts while maintaining high user satisfaction.
Zitieren
APA
Zaman, M. W. U., Rashid, U., & Khan, A. R. (2026). LLM-Powered Framework to Explore Summarized Aggregated Multimedia Vertical Web Search Results. International Journal of Innovations in Science and Technology, 1. https://doi.org/10.33411/ijist/1777
Vancouver
Zaman MWU, Rashid U, Khan AR. LLM-Powered Framework to Explore Summarized Aggregated Multimedia Vertical Web Search Results. International Journal of Innovations in Science and Technology. 2026:1. doi:10.33411/ijist/1777
BibTeX
@article{Zaman2026,
author = {Zaman, Muhammad Wajeeh Uz and Rashid, Umer and Khan, Abur Rehman},
title = {LLM-Powered Framework to Explore Summarized Aggregated Multimedia Vertical Web Search Results},
journal = {International Journal of Innovations in Science and Technology},
year = {2026},
pages = {1},
doi = {10.33411/ijist/1777}
}
EUVIMED