Towards Self-organizing Search Systems

This publication doesn't include Faculty of Arts. It includes Faculty of Informatics. Official publication website can be found on muni.cz.

Authors

BARTOŇ Stanislav DOHNAL Vlastislav SEDMIDUBSKÝ Jan ZEZULA Pavel

Year of publication 2010
Type Chapter of a book
MU Faculty or unit

Faculty of Informatics

Citation
Description The huge amount of images, video and music clips produced by various digital devices everyday, must be processed. Firstly, this kind of data calls for content-based search or similarity search rather than keyword-based or text-based search. Secondly, new scalable and efficient methods capable of storing and querying such data must be developed. Although many distributed approaches exist, one of the most suitable and flexible is provided by self-organizing systems. These systems exhibit high resistance to failures in dynamically changing environments. In this chapter, we propose a general three-layer model for designing and implementing a self-organizing system that aims at searching in multimedia data. This model gives a developer guidelines what component must be implemented and how they should behave. The usability of this model is illustrated on a system called Metric Social Network. The architecture of this system is based on the social-network theory that is utilized for establishing links between nodes. The system's properties are verified by organizing and searching in 10 million images.
Related projects: