语言获得的计算模型述评

A Review of Evaluation of Computational Models of Language Acquisition

人类语言是一个十分复杂的系统,语言研究者还没有完全了解儿童语言的获得过程,语言获得的计算模型能够较好地解释儿童是如何快速地在已有的语言中寻找范式,并将它们归纳到更大的语言结构中去。语言获得计算模型的材料都是以原始发声的形式呈现的,这样便于更好地对模型进行评价。通过总结对语言获得模型进行评估时所存在的问题,并比较两种计算方法的共同点和不同点,最后提出一个完善的计算模型的标准。未来的研究需将心理语言学、认知科学和计算机科学等不同领域结合起来,才能使语言获得模型更加有说服力。

Human language is a very complicated system and the process of child language acquisition is not yet well understood by language researchers. Computational models of language acquisition can shed inter- esting light on this process, and explain how children find patterns in the ambient language quickly and generalize them to larger structures. The materials of computational models of language acquisition were presented in the form of raw utterances; it＇s good to evaluate the models. We reviewed the evaluation process of computational models of language acquisition, identified the commonalities and differences be- tween the various existing approaches to language learning and finally proposed a specific criterion of a per- fect computational model. Works in future are conducted in different fields, including psycholinguistics, cognitive science and computer science for a well-informed model to emerge.