Alex Graves大神的论文 Thispaperpresentsaspeechrecognitionsys-temthatdirectlytranscribesaudiodatawithtext,withoutrequiringanintermediatephoneticrepre-sentation.Thesystemisba