Feature-rich continuous language models for speech recognition | IEEE Conference Publication | IEEE Xplore