mk:isis13

Summary

Search in a Hybrid P2P? System Using Visualized Network Information. Kei Ohnishi, Hiroshi Yamamoto and Mario Köppen. In Proc. The 14th Intl. Symposium on Advanced Intelligent Systems (ISIS 2013), November 2013.

Abstract

We propose a hybrid P2P? information search system that combines (1) an idea that a searcher searches not informationitself but people who has KANSEI (human sensitivity) well-fitted to the searcher, (2) a P2P? network on which we can expectto build a scalable information search system against rapid increase of the amount of information, and (3) a folksonomywhich is a system that allows users to classify information by themselves, to enable us to obtain our desired informationfrom the huge amount of information on the network quickly and reliably. In addition, we evaluate the proposed systemthrough simulations. The main characteristics of the proposed system is that it produces a two dimensional mapvisualizing all nodes'KANSEI information represented by over twodimensional numerical vectors, that is the entirenetwork, using self-organizing map, and then utilizes the map for search. In simulations for evaluation, we assume thatthere exist many nodes with a variety of KANSEI and then examine the effect of the two dimensional map, which isreduced from a higher dimensional space for representing KANSEI information, on the node search. Simulation resultsshow that the proposed system provides balanced opportunities that we obtain our desired information and that we meetother unknown information compared to a search method that conducts search directly in the higher space forrepresenting KANSEI information and a random search method. The results suggest that the proposed system is suitablenot for reliably searching only information that a searcher explicitly desires but for searching with some heuristic factorssuch that a searcher happens to meet people whose KANSEI interests the searcher.

Bibtex entry

@INPROCEEDINGS { mk:isis13,
    ABSTRACT = { We propose a hybrid P2P? information search system that combines (1) an idea that a searcher searches not informationitself but people who has KANSEI (human sensitivity) well-fitted to the searcher, (2) a P2P? network on which we can expectto build a scalable information search system against rapid increase of the amount of information, and (3) a folksonomywhich is a system that allows users to classify information by themselves, to enable us to obtain our desired informationfrom the huge amount of information on the network quickly and reliably. In addition, we evaluate the proposed systemthrough simulations. The main characteristics of the proposed system is that it produces a two dimensional mapvisualizing all nodes'KANSEI information represented by over twodimensional numerical vectors, that is the entirenetwork, using self-organizing map, and then utilizes the map for search. In simulations for evaluation, we assume thatthere exist many nodes with a variety of KANSEI and then examine the effect of the two dimensional map, which isreduced from a higher dimensional space for representing KANSEI information, on the node search. Simulation resultsshow that the proposed system provides balanced opportunities that we obtain our desired information and that we meetother unknown information compared to a search method that conducts search directly in the higher space forrepresenting KANSEI information and a random search method. The results suggest that the proposed system is suitablenot for reliably searching only information that a searcher explicitly desires but for searching with some heuristic factorssuch that a searcher happens to meet people whose KANSEI interests the searcher. },
    AUTHOR = { Kei Ohnishi and Hiroshi Yamamoto and Mario Köppen },
    BOOKTITLE = { Proc. The 14th Intl. Symposium on Advanced Intelligent Systems (ISIS 2013) },
    ADDED = { 2014-01-05 06:49:36 +0000 },
    MODIFIED = { 2014-01-05 06:52:12 +0000 },
    MONTH = { November },
    TITLE = { Search in a Hybrid P2P? System Using Visualized Network Information },
    YEAR = { 2013 },
}