Performance of a Natural Language Framework

摘要

An ideal intelligent program is one that can communicate with humans using a natural language. Three key operations of communication are to read, understand, and write sentences. We have developed a natural language processing framework that allows an incremental learning of a grammar, such as the English grammar. The framework contains algorithms for the three operations and also allows grammar terms to be taught using multiple components that define the purpose, structures, and rules of the terms. These components are used not only for parsing a sentence, but also for understanding its intent and for producing a sentence to reflect a thought. The grammar is defined in a manner that allows more complex terms to be introduced incrementally using a combination of simpler terms and an expansion of existing terms. The performance of this framework on all three operations improves as more grammar is introduced.