Cluster-Merge本体构造算法

本体构造就是利用各种数据源以半自动方式新建或扩充改编已有本体以构建一个新本体。现有的本体构造方法大都以大量领域文本和背景语料库为基础抽取大量概念术语,然后从中选出领域概念构造出一个本体。Cluster-Merge算法首先对领域文档先用k-means聚类算法进行聚类,然后根据文档聚类的结果来构造本体,最后根据本体相似度进行本体合并得到最终的输出本体。通过实验可证明用Cluster-Merge算法得出的本体可以提高查全率、查准率。

An Ontology Learning Based on Documents Clustering

Ontology learning is a series of semi-automatic ontology building methods and techniques.It does this by using a variety of data sources in order to building a new ontology through constructing a new or expanding an ontology by using semi-automatic methods.The existing ontology construction methods are through choosing the concept of a field from a large number of documents of a special domain and background corpus to construct an ontology.This paper proposed a Cluster-Merge algorithm that first cluster documents of a special domain by using the k-means clustering algorithm;And then construct many ontologies based the results of clustered documents;Finally,merge these ontologies based ontology similarity to get the final output of ontology.It can be proved that these ontologies derived from the Cluster-Merge algorithm can improve recall ratio,precision ratio through experiments.