Document-specific context plsa language model for speech recognition | IEEE Conference Publication | IEEE Xplore