Arabic Semantic Text Classification Based on Wavelet Spectral Analysis

We propose in this paper a new document representation in Text Mining based on signal representation and spectral processing by Wavelets Transform. Our method gives a solution of syntactic and semantic descriptor dependency problem, without deleting information. This can be done by grouping dependent descriptors in clusters with a single representative. Thereafter each class is represented by a discrete signal giving different degrees of dependence between descriptors, we then take advantage of the Multi Resolution Analysis properties of the Wavelet Transform. We show that we are able to achieve higher precision when compared to Vector Space Model representation and Latent Semantic Analysis in the context of Arabic Text Classification.