作者: Hynek Hermansky
DOI: 10.1007/978-3-642-23538-2_1
关键词:
摘要: Unexpected words attract listener's attention. They are informationrich and getting them right is important for human communication. In the automatic recognition of speech (ASR), that not in expected lexicon machine typically substituted by some acoustically similar but nevertheless wrong words. The article discusses reasons this undesirable behavior machine, describes known examples dealing with unexpected perception their implications, proposes an alternative architecture ASR could alleviate problems acoustic inputs. Some published experimental results from using given.