Augmenting and Efficiently Utilizing Domain Theory in Explanation-Based Natural Language Acquisition

Abstract

In this paper, a new language acquisition model is proposed to acquire parsing-related knowledge via an Explanation-Based Learning (EBL) approach. The domain theory in the model consists of two parts: a static part and a dynamic part. The static part consists of the universal linguistic principles proposed in the Generalized Phrase Structure Grammar (GPSG) formalism, while the dynamic part contains the Context-Free grammar rules as well as syntactic and thematic features of lexicons. In parsing (problem-solving), both parts work together to parse input sentences, and in learning, the dynamic part is enriched and generalized by obeying the principles in the static part To be a robust and practical system, the model also incorporates the concepts of knowledge indexing, common work sharing, and dynamic conflict resolution to maintain efficiency of the problem solving module. The effect of these problem solving strategies to the knowledge utility problem in machine learning is thus investigated based on the experiments of the language acquisition model.