IMPROVED FEATURE EXTRACTION ON TEXT DOCUMENTS USING NEURAL NETWORK MODEL

Abstract

In natural language processing, the text clustering plays a major role on reducing the text dimensionality. However, the lack of data models has made the clustering algorithm to face sparsity problems. The integration with deep learning has resolved the problem of scarce knowledge on text documents. However, deeper architectures learn such redundant features, which limit the efficiency of solutions. In this paper, a complete extraction of features from text document using neural network model. The neural network model utilizes feed forward mechanism and a type of unsupervised learning that denoises the corrupted input features. The reconstructed feature is used for initialing the feed forward network. This method reduces the manual labelling in the process of screening. For evaluation, series of experiments are conducted to investigate the performance of the method over the text datasets with various conventional algorithms.

Keywords

• text document
• feature extraction
• neural network