摘要： 提出了信息抽取系统中同指求解方法，主要针对汉语篇章中与代词和定指短语相关的同指求解。该方法分为规则消解和统计因子消解两个组成部分。规则消解法给出了规则的一般形式、获取途径及应用方法，还结合若干规则实例来表明规则的有效性。利用规则库中的规则对可能先行语进行分析过滤，如果能“滤出”唯一的一个先行语，则这个即为求解到的先行语。否则进入统计因子消解过程。在这部分利用统计的方法，对消解因子库中的各因子分析求和，然后选择最大的作为求解的先行语。本算法具有很好的跨领域性和易扩展性。实现了该方法并进行了测试运行，结果表明该方法是行之有效的。

关键词: 同指消解, 规则消解算法, 统计因子消解

Abstract: Focused on pronominal and definite nominal anaphora for an information extraction system, this method is divided into two parts: the rule-based and the factors-based statistical resolution algorithm. The general representation, the method to get rules and the application method pertaining to rule are discussed. Some examples are given to show the effectiveness of the rule. Using these rules of the rule-database, people can filter out the possible antecedents. If lucky enough, people can filter out the exact antecedent.Otherwise,it goes to the factors-based statistical resolution. Every candidate compute its value on the every factor and sum them, then select the one with the highest sum value as the antecedent. The method is implemented and tested. The result shows that the proposed method is promising.

中图分类号: 

• TP391.12