地理实体概念及其位置关系的获取和验证  

Acquiring Geographical Entities and their Relations from Texts

作　　者：姜琳[1] 李宇[1] 卢汉[1] 曹存根[1] 

机构地区：[1]中国科学院计算技术研究所中科院智能信息处理重点实验室

出　　处：《计算机科学》2007年第12期151-156,174,共7页Computer Science

基　　金：自然科学基金的资助(60273019;60573064;60573063和60496326);国家重点基础研究发展计划2003CB317008和G1999032701的资助

摘　　要：文本知识获取(Knowledge acquisition fromtext,简称KAT)是知识工程中的一个重要研究课题。重点研究如何从大规模Web网页文本中获取地理实体概念及其位置关系知识,本文首先介绍了如何自动和半自动地获取这些地理实体概念及其位置关系的文法模式,建立文法模式库;然后基于文法模式库获取例句来抽取候选概念并进行概念验证;最后利用基于图论的方法构造位置关系图,利用地理领域特定规则进行分析验证。作为统一概念图管理下概念空间的一个重要组成部分,地理实体概念及其位置关系本身不仅是知识库的一个重要部分,还可为知识库中其它领域的知识提供支持。There are various sorts of knowledge contained in text, and knowledge acquisition from text （KAT） has become one of the important research problems in knowledge engineering. How to acquire Geographical entities and their relations from large-scale Webpages is the research topic in this paper. Firstly, grammatical patterns are acquired automatically and semi-automatically to construct the Extraction Pattern Base. Then quasi-concept are extracted and verified locally with those resulting sentences generated from extraction patterns. At last, a Geographical Graph is defined based on graph theory, with the geographical rules to verify geographical entities and their relations, As an important part of the Conceptual Space managed by the Concept Graph, geographical entities and their relations are not only the important part of the Knowledge Base, but also provide support for other domain or domain independent knowledge.

关 键 词：文本知识获取 地理实体概念获取 位置关系获取 知识验证 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]