采用预测策略的Earley算法

Earley Algorithm Using Prediction Strategies

在自然语言处理中,句法分析主要有基于统计的方法和基于规则的方法。Earley算法是一种基于规则的方法,可以分析任意上下文无关文法(CFG),而不需要对文法进行修改。详细分析了Earley算法的特点。在通常的Earley算法中增加了多种预测机制,这些预测机制借鉴了LL,LR以及SLR等确定性分析算法的一些思想,并对这几种不同的预测机制及其组合在相同条件下进行了中文句法分析实验。结果显示,引入这些预测机制通常可以减少产生项目的数量,从而节省存储空间,减少运行时间。

There are two kinds of parsing algorithms in nature language processing：one based on statistics and the other based on grammar rules. Earley algorithm is based on grammar rules. It can parse any context free grammar （CFG） without changing the grammar. This paper used several predictive strategies in Eartey algorithm. These strategies come from LL,SLR and LR algorithms. Experiments were made for these strategies and their combinations. Results indicate that these strategies can usually reduce number of items and make parsing faster.