Constrained Systems for Efficient Natural Language Processing

这个基于计算机的语言处理形式的研究将集中在促进语言模型实现的特征上。它不仅应该能够简洁地将这样的模型嵌入到系统中，而且所得到的系统应该具有计算效率。为了实现对形式的限制所确定的要求，使其适合于自然语言模型的表达。在早期的工作中，研究者发现了一系列约束语法形式的几个共同特征，尽管它们的符号非常不同。这些特征涉及派生过程形式的限制，并且似乎表明这些形式中的每一种都适用于有效的并行解析算法。该研究现在将使用并行计算模型的研究结果，形式化这些系统的共同特征及其处理含义。这将导致对计算语言学发展的理论中固有的理想计算特性的更深层次的理解。

This investigation of a computer based language processing formalism will focus on characteristics that facilitate the implementation of linguistic models. It should not only be possible to embed such models in the system concisely, but the resulting system should be computationally efficient. In order to achieve the form of the restrictions determined by the requirement that it be suitable for expressing natural language models. In earlier work, the investigator identified several features common to a range of constrained grammar formalisms, despite their very different notations. These features concern restrictions of the form of the derivation process, and appear to indicate that each of these formalisms is amenable to efficient parallel parsing algorithms. The research will now formalize the common features of these systems and of their processing implications, using results from research on parallel models of computation. This should lead to a deeper understanding of the desirable computational properties inherent in theories developed by computational linguistics.