Text classification with document embeddings

Abstract

Distributed representations have gained a lot of interests in natural language processing community. In this paper, we propose a method to learn document embedding with neural network architecture for text classification task. In our architecture, each document can be represented as a fine-grained representation of different meanings so that the classification can be done more accurately. The results of our experiments show that our method achieve better performances on two popular datasets.