Digital Text Feature Extraction Using Singular Value Decomposition and Principal Component Analysis

The management and automated processing of digital information is an increasingly significant topic of research. Digital text information is becoming particularly pervasive and poses particular challenges to processing methods. In this paper, we present a feature extraction method using singular value decomposition and principal component analysis for effective information reconstruction. We first use singular value decomposition to modify the small eigenvalues of the test matrix. This new test matrix may contain redundant information and contribute to increased computational complexity. Thus, we use principal component analysis in a second stage of processing. This method reduces the amount of redundant information and the dimensionality of the data, and eliminates the correlation between each input variable. This greatly simplifies the computational complexity of the feature extraction.