Context While deep learning approaches have significantly pushed language applications like keyword spotting in audio recordings, natural language processing on text documents etc., these solutions for the masses can not applied directly to non-standard speech signals, e.