Text Classification Using Sentential Frequent Itemsets

Text classification techniques mostly rely onsingle term analysis of the document data set, while more concepts,especially the specific ones, are usually conveyed by set of terms. Toachieve more accurate text classifier, more informative featureincluding frequent co-occurring words in the same sentence and theirweights are particularly important in such scenarios. In this paper, wepropose a novel approach using sentential frequent itemset, a conceptcomes from association rule mining, for text classification, whichviews a sentence rather than a document as a transaction, and uses avariable precision rough set based method to evaluate each sententialfrequent itemset's contribution to the classification. Experiments overthe Reuters and newsgroup corpus are carried out, which validate thepracticability of the proposed system.