语义剪枝系统研究  

Experimental Research of Semantic Pruning System

作　　者：辛日华[1] 

机构地区：[1]呼伦贝尔学院计算机科学与技术学院,内蒙古海拉尔021008

出　　处：《内蒙古师范大学学报（自然科学汉文版）》2014年第1期106-109,113,共5页Journal of Inner Mongolia Normal University(Natural Science Edition)

基　　金：内蒙古自然科学基金资助项目(2013MS0102)

摘　　要：在自行研究和开发的基于知网的语义剪枝系统中,对大规模语料库中的语料多义词进行知网释义的自动提取,然后进行语义剪枝.将语义剪枝系统的实验结果与人工标注的结果相比,来解决自然语言处理过程中的词义排歧问题.其目的是通过语义剪枝系统尽可能地减少歧义词在上下文中的错误的或最不可能的义项.语义剪枝后形成一个词和其可能义项的列表,尽量将一个词真正正确的义项保留下来.使用召回率和简化率两个指标对语义剪枝算法进行评价,结果表明窗口尺寸和分析单元的选取对召回率和简化率有影响.A semantic pruning system based on ＂HowNet＂ is researched and developed. Automatic extraction and semantic pruning about polysemants of HowNet interpretations in large-scale corpus are studied. Through comparing the experimental results between semantic pruning system and manual labe- ling,The word sense disambiguation in natural language processing can be solved. The wrong items or the most unlikely items in context is as far as possible reduced through semantic pruning system. Then, A word and A possible meaning list retaining the real meaning of it are formed, finally,the results shows the influence of recall and simplified rate due to the window scale and the analysis unit selection by using Recall and simplified rate to evaluate.

关 键 词：词义排歧 语义剪枝 召回率 简化率 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]