论文部分内容阅读
人机之间使用自然口语通讯是一个很值得追求的目标;从30年前开始,这项工作已取得真正的进展。达到这个目标有很多复杂的因素,如:(a)说话人发音各异,(b)不易将连续语言分成片段以用于分析和与已储存的参考式样相比较,(c)很大的词汇量(如50000个字)要求机器具有高速计算率,(d)不同语言之间的频谱相似性要求机器根据在不同语言的上下文中发生的机率作出决策。因此,不管语言通讯的研究或语言识别系统的发展都是高度多科性的;一些彼此无关的学科,如生理学,声学和语音学和由数学派生的方法,如统计学,计
The use of natural language communication between human-computer is a very worthwhile goal; since 30 years ago, this work has made real progress. There are a number of complex factors that can be achieved to achieve this goal, such as: (a) a speaker’s pronunciation varies; (b) it is not easy to break up consecutive languages into fragments for analysis and comparison with stored reference patterns; (c) very large terms The amount (say 50000 words) requires the machine to have a high computational rate of speed, (d) the spectral similarity between languages requires the machine to make decisions based on the odds that occur in the context of different languages. Therefore, regardless of language communication research or the development of speech recognition systems are highly multidisciplinary; some non-related disciplines, such as physiology, acoustics and phonetics and mathematics-derived methods, such as statistics