基于粗糙集的多变量决策树构造方法

MODELING TEMPORAL SEMANTICS INFORMATION FOR NATURAL LANGUAGE

本文利用粗糙集理论中条件属性相对于决策属性的核，解决多变量检验中属性的选择问题．另外，定义了２个等价关系相对泛化的概念，并将它用于解决多变量检验的构造问题．通过一个例子，对本文提出的多变量决策树方法与著名的单变量决策树（ＩＤ３）方法进行了比较，结果表明前者比后者更简单．同时，对几种多变量决策树方法做了初步的对比分析．

In natural language processing, temporal is an important situation. In order to understand temporal situation in the text, this paper presents a hierarchical temporal semantic structure independent of the surface description in sentences.The surface semantic structure formalizes the description of temporal situation; The deep semantic structure describes the dynamic attributes and the existing characteristic of the event in a given temporal situation. This temporal semantic representation model allows to express point in time as well as time intervals. It provides means to indicate time basic point with respect to “an absolute physical time”, as well as time basic point relatively to the speaker's time. A universal calculation method of the temporal semantics provided in this paper localizes the temporal entities in all languages into the temporal axis. This enables to interpret combinations of verb tense, temporal adverbs and connective. In the meanwhile, within this model some constructs enable to express tense, aspectual properties of verbs as well as the semantic calculation of temporal relationship among events in the text.