Incremental language models for speech recognition using finite-state transducers | IEEE Conference Publication | IEEE Xplore