MAINTAININGTHE INFLUENCEOF CONTEXT MATRIX CONSTRUCTIONPRINCIPLESON THE ABILITYOFLINGUISTIC FEATUREEXTRACTIONBY ICA
Markus Borschbach, Martin Pyka · 2007
Blind Signal Separation (BSS)based onIndependent Component Analysis (ICA) isanemerging approach whichapplication isnotlimited tothesignal processing research, whereitsapplication principle israther straight forward. Foranincreasing amountofinformation processing fields, ICAhasmeaningful application whicharestill undiscovered. Theaimofthis paperistoinvestigate theability of linguistic feature extraction based onwordcontext preprocessing byICA.Theworkrefers toafirst brief analysis in whichICAwasapplied toanEnglish corpus. Wecontinue this analysis depending onthenumber ofcomponents and theamountofsyntactical informations that wetakeinto account. Further morewediscuss towhichextent theresults deliver either general linguistic features, orlinguistic features giving usinformation about thetext. Index Terms-Textprocessing, Textrecognition, Linguistic feature extraction, Independent component analysis