Head-Driven Phrase Structure Grammar: An Investigation in Linguistic Theory and Its Computational Implementation

这是一项重新支持语言学理论项目的建议，该项目已被证明是非常富有成效的，其结果具有重要的理论和实践意义。研究人员正在继续研究一种明确的语言句法和语义理论的可能性和问题，这种理论运行在单一的表示水平上，具有严格定义和限制的形式属性的规则。大多数理论家发现，为了解释用自然语言观察到的数据，有必要假设两个或更多级别的表示和各种形式类型的规则。这里提出的理论，由于其范围和力量受到更多的限制，对语言的潜在性质提出了更强有力和更有趣的主张，并在一定程度上被证明是经验上的有效，从而构成了语言理论的重大进步。为了从经验上验证理论，研究者必须满意地处理大量的经验问题，这些问题导致其他研究人员假设更复杂的理论，这就是这里提出的项目的负担。如上所述，该项目除了具有经验意义外，还具有实际意义。正在探索的受限理论允许比更多不同的理论更快和更有效的计算机处理，并且在一定程度上它可以处理自然语言的数据并产生有用的输出，它提供了用于自然语言处理的计算机系统的重大进步的前景。

This is a proposal for renewed support of a project in linguistic theory which has proven exceedingly productive, with results of both theoretical and practical importance. The researcher is continuing to investigate the possibilities and problems of an explicit theory of linguistic syntax and semantics which operates on a single level of representation, with rules with strictly defined and restricted formal properties. Most theoreticians have found it necessary to postulate two or more levels of representation, and rules of a variety of formal types, in order to account for the data observed in natural language. The theory proposed here, by being more restricted in its scope and power, makes much stronger and more interesting claims about the underlying nature of language, and to the extent that it can be shown to be valid empirically, it thus constitutes a significant advance in linguistic theory. In order to validate the theory empirically, the investigator must deal satisfactorily with the numerous empirical problems which have led other researchers to postulate more complex theories, and that is the burden of the project proposed here. The project has, as mentioned above, practical import in addition to its empirical interest. The restricted theory being explored permits much faster and more efficient computer processing than more diverse theories do, and to the extent it can handle the data of natural language and produce useful output, it offers the prospect of a major advance in computer systems for natural language processing.