Studying machine translation technologies for large-data CLIR tasks: a patent prior-art search case study