machine learning - Modeling features of Relation Extraction in the SVMlight input format -

i working on project focuses on relation extraction corpus of wikipedia text, , plan use svm extract these relations. model this, plan use word features, pos tag features, entity features, mention features , on mentioned in following paper - https://gate.ac.uk/sale/eswc06/eswc06-relation.pdf (page 6 onwards)

now, have set pipeline feature extraction , got corpus annotated , wish use package svm-light purpose of project. according input file format of svm-light package, requisite format - .=. : : ... : #

example (from svm-light webpage) -

in classification mode, target value denotes class of example. +1 target value marks positive example, -1 negative example respectively. so, example, line

-1 1:0.43 3:0.12 9284:0.2 # abcdef

specifies negative example feature number 1 has value 0.43, feature number 3 has value 0.12, feature number 9284 has value 0.2, , other features have value 0. in addition, string abcdef stored vector, can serve way of providing additional information user defined kernels.

now, wish know how model features using values include words, pos tags , entity types , subtypes feature vector accepted svm-light package, each feature has real number value associated it. how mapping choice of features these real values done?

it of great if has worked @ similar problem before prod me in right direction.

thanks.