DoSO: a document self-organizer

Spanakis, G; Siolas, G; Stafylopatis, A

dc.contributor.author	Spanakis, G	en
dc.contributor.author	Siolas, G	en
dc.contributor.author	Stafylopatis, A	en
dc.date.accessioned	2014-03-01T11:46:43Z
dc.date.available	2014-03-01T11:46:43Z
dc.date.issued	2012	en
dc.identifier.issn	09259902	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/38027
dc.subject	Document clustering	en
dc.subject	Document representation	en
dc.subject	SOM	en
dc.subject	Wikipedia	en
dc.title	DoSO: a document self-organizer	en
heal.type	other	en
heal.identifier.primary	10.1007/s10844-012-0204-9	en
heal.identifier.secondary	http://dx.doi.org/10.1007/s10844-012-0204-9	en
heal.publicationDate	2012	en
heal.abstract	In this paper, we propose a Document Self Organizer (DoSO), an extension of the classic Self Organizing Map (SOM) model, in order to deal more efficiently with a document clustering task. Starting from a document representation model, based on important ""concepts"" exploiting Wikipedia knowledge, that we have previously developed in order to overcome some of the shortcomings of the Bag-of-Words (BOW) model, we demonstrate how SOM's performance can be boosted by using the most important concepts of the document collection to explicitly initialize the neurons. We also show how a hierarchical approach can be utilized in the SOM model and how this can lead to a more comprehensive final clustering result with hierarchical descriptive labels attached to neurons and clusters. Experiments show that the proposed model (DoSO) yields promising results both in terms of extrinsic and SOM evaluation measures. © 2012 Springer Science+Business Media, LLC.	en
heal.journalName	Journal of Intelligent Information Systems	en
dc.identifier.doi	10.1007/s10844-012-0204-9	en
dc.identifier.spage	1	en
dc.identifier.epage	34	en