Progressive browsing-state delivery for interactive multimedia exploration

Unlike traditional retrieval, dynamic exploration of multimedia collections may require complex aggregation queries whose performance is highly sensitive to dataset size, filters and grouping applied at any time. Such queries often yield unstable response times, thus undermining interactivity. Inspired by the online aggregation approach from the database community, we investigate how progressively refined intermediate results, accompanied by progress estimates, can help users retain control over the aggregation process, even when applied to an extremely large collection. We evaluate this approach within the Multidimensional Media Model. Our study shows that moving deduplication and grouping operators from the database to the server, coupled with a batched streaming strategy, reduces time-to-first-byte by over 90% for heavy queries while maintaining steady improvements in result quality. Our examination of the effect of join order shows that the database query optimiser produces near-optimal plans, with sensitivity compressed to at most 26%. An evaluation of classical selectivity estimation for progress tracking identifies attribute correlation as the primary source of error and evaluates two alternative estimation strategies.

Author dc:creator

• Sigurður Þórarinsson 1991-

Contributors dc:contributor

• Háskólinn í Reykjavík

• Meistaraprófsritgerðir
• Tölvunarfræði
• Gagnagrunnar
• Söfnun
• Computer science
• Databases
• Aggregation

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/1946/53724

OAI identifier oai:identifier

oai:skemman.is:1946/53724