Parsing-Driven Generalization for Natural Language Acquisition

Parsing is an important step in natural language processing. It involves tasks of searching for applicable grammatical rules which can transform natural language sentences into their corresponding parse trees. Therefore parsing can be viewed as problem solving. From this point of view, language acquisition can be generalized from problem solving heuristics. In this paper we show how learning methods can be incorporated into a wait-and-see parser (WASP), the problem solver. We. call this approach parsing-driven generalization since learning (acquisition of parsing rules and classification of lexicons) is basically derived from the parsing process. Three generalization methods are reported in this paper: a simple generalization mechanism, a mechanism of generalization by asking questions, and a mechanism of generalization back-propagations. The simple generalization mechanism generalizes from any two parsing rules whose action parts (right-hand sides) are the same while whose condition parts (left-hand sides) have a single difference. The mechanism of generalization by asking questions is triggered when a 'climbing-up' move on a concept hierarchical tree is attempted and is necessary in avoidance of overgeneralizations. The generalization back-propagation mechanism is to propagate a confirmed generalization of some later parsing rule back to its precedent rules in a parsing sequence and thus causes them to be generalized as well. This mechanism can save many questions to be asked. With the three generalization methods and a mechanism to maintain lexicon classification (the domain concept hierarchy), we have been able to show a plausible natural language acquisition model.