Use semantic meaning of coreference to improve classification text representation

On large scale dataset, the effect of automatic text classification is now still far from perfect. It´s a common agreement that more sufficient text semantic meaning be adopted in text representation to deal with the challenge. This paper introduces semantic meaning of coreference in and to improve traditional BOW representation. The result of text classification experiment shows that, contrasted with traditional BOW representation, the improved model increases the discernment to positive instances. And that the classification performance of the new BOW representation model is no less good than that of stemmed BOW representation model.