Cross-language identification using the wavelet transform and artificial neural network

With the advent of the Internet, search engines were developed for English language because English language was a lingua franca. Currently, most of popular search engines such as Google and Yahoo! are available in more than 50 languages. However, these search engines have received less attention in South Asian languages especially, Urdu language. In this paper, we propose a novel approach for feature extraction and classification of queries in cross-language search engines. This novel approach presents an automatic method for classification of English and Urdu languages identification. The classifier used is a three-layered feedforward artificial neural network and the feature vector is formed by calculating the wavelet coefficients. Three wavelet decomposition functions (filters), namely Haar, Bior 2.2 and Bior 3.1 have been used to extract the feature vector set and their performance results have been compared. The performance results of the Haar filter have given superior results than other filters.