There　are　a　lot　benefits　to　enable　intelligent　agent　understanding　the　information　from　semantic　web.　It　enhances　the　efficiency　of　information　usage　and　at　the　same　time,　suffices　the　need　of　users.　Semantic　documents　contain　adequate　semantic　information　which　helps　understanding.　However,　discrepancy　between　ontology　which　is　an　interpreter　Of　semantic　document　prevents　the　share　of　knowledge.　In　this　paper,　we　proposed　a　uniform　representation　for　the　content,　which　include　concepts　and　relations,　of　semantic　documents　based　on　WordNet.　First,　disambiguation　is　preceded　within　the　key　words　in　a　document　for　the　purpose　of　mapping　them　to　concepts.　Then　we　present　the　whole　document　in　the　form　of　concept　graph　that　Levenshtein　Distance　could　be　applied　for　making　a　classification　of　documents.　We　have　empirical　result　that　this　methodology　makes　a　promising　raise　in　accuracy.