Web Information Extraction Using Generalized
Jinlin Chen · 2006
Hidden MarkovModel(HMM) isanimportant approach forinformation extraction (IE). Whenapplied toWeb IE,several problems exist withHMM basedapproaches dueto thelackofconsideration onWeb-specific features. Inthis paper wepresent aGeneralized HiddenMarkovModel(GHMM)that extends traditional HMMs by makinguseofWeb-specific information forWeb IE.Inourapproach weuseWeb content blockinstead oftermasbasic extraction unit. Besides, instead of using thetraditional sequential state transition order, wedetect thestate transition orderofGHMM basedonlayout structure of thecorresponding web page.Furthermore, we usemultiple emission features instead ofsingle emission feature. Inthis way GHMM canbetter accommodate Web IE.Experiments show promising results comparing totraditional HMM based WebIE. IndexTerms-Hidden MarkovModel, Information extraction, Layout Analysis, Web