comparing traditional clustering method with the mutual information method

first of all we have to make 3 folders which are empty say ( data structure,,computer network,,compiler)..now make 1 directory in which store the html pages about 100 related to the folders..now extract or retrieve the html pages one by one..then convert each html pages into string..and store the releavnt data of string in an array and delete the data of string like(is,, are ,,am) ...now seperately make again three files in which we have to store words related to that files means in CN file folder we can store CN related data nad so on... take atleast 30 data in each folder....now compare the data of an array (t1)with the each files folder with the formula n(t1)intersection n(datastructure)*100/n(t1 U datastructure)..and soon...compare like this each element of datastructure ,, computernetwork,,and compiler folder with an array...from which match found store that value in empty folder that we make in starting..