Improving acoustic models with captioned multimedia speech | IEEE Conference Publication | IEEE Xplore