Noisy-parallel and comparable corpora filtering methodology for the extraction of bi-lingual equivalent data at sentence level