The Development of Collative Semantics Including a Parallel Processing Implementation

本研究探讨计算机系统的自然语言理解。具体目标是：1)描述自然语言中单词的歧义如何从上下文和“语义关系”（连接单词可能的替代含义的潜在结构）中得到解决。2)利用语义关系的方法，开发了自动进行自然语言处理的计算机系统meta5。这种发展的一个关键特征是使用并行处理，许多计算元素同时操作来分析自然语言输入。本研究的意义在于，对于机器翻译、用户界面和基于知识的系统而言，自然语言处理系统必须克服普通单词具有多种含义的基本问题，并且必须为每种出现找到正确的含义。该任务的处理要求使得探索并行处理和其他大大提高计算能力的方法变得至关重要。

This research investigates natural language comprehension by computer systems. Specific targets are to: 1) Describe how the ambiguity of words used in natural language may be resolved from context and "semantic relations" - the underlying structures connecting the possible alternative meanings of words. 2) Develop a computer system, meta5, which automatically carries out natural language processing, using the semantic relations approach. A key feature of this development is the use of parallel processing, with many computing elements operating simultaneously to analyse the natural language input. The significance of this research is that natural language processing systems, for machine translation, user interfaces, and knowledge based systems, must overcome the fundamental problem that ordinary words have many meanings, and the right one must be found for each occurrence. The processing requirements of this task make it essential to explore parallel processing and other means of greatly increasing the computing power brought to bear on the problem.