Large vocabulary speech recognition with multispan statistical language models | IEEE Journals & Magazine | IEEE Xplore