Context-dependent acoustic modeling using graphemes for large vocabulary speech recognition | IEEE Conference Publication | IEEE Xplore