学术资源本体非等级关系抽取研究

目的/意义] 概念非等级关系抽取是本体构建的必要步骤，学术文献作为一种重要的学术资源类型，本文主要利用其结构特点来进行本体概念非等级关系的抽取。方法/过程] 首先，在本体概念抽取的基础上，对文献中概念的类型进行分类，以便于后期关系动词搭配的概念类型来排除不符合条件的三元组；其次，确定学术文献中的关系类型，并采用C-value方法抽取表示关系的动词，进行本体关系的表示；再次，评价概念对的关联性，利用互信息法对概念对进行排序并去除非相关概念对，实验表明该方法非常有效；最后评价概念对与关系动词的关联，分析影响三元组关联的因素，再采用实验确定模型挖掘三元组，实验比较现有的关联规则挖掘的方法。结果/结论] 结果表明本文提出的三元组选择模型效果明显超过现有的关联规则挖掘方法，并且在语料集扩大的情况下这种优势更为明显。

Extracting Non-taxonomic Relationships for Ontology Learning of Scientific Resources

Purpose/significance] Extracting non-taxonomic relationship is the most complex work of ontology learning. It is also an unsolved problem of ontology learning. Related research is mainly based on the domain vocabulary. This paper aims to study the scientific ontology construction, and the knowledge acquisition comes from scientific articles. This paper uses the structure of the scientific articles to extract non-taxonomic relationships. Method/process] Firstly, based on the concept extraction of the previous work of our project, the paper classifies the concept into some categories in order to exclude some concept types with verb templates. Then, in order to define the type of the relationship, it uses C-value to extract verb represent the relationship. After that, it evaluates the relationship of the concepts, and uses MI to exclude some concepts. The experiments show that it is very effective. At last, it evaluates the relationship of concepts and verb to analyze the factors influencing the relationship, and selects a model. The experiment compares the associate rule methods. Result/conclusion] It shows the model proposed by this paper is very effective and outperforms the associate rule methods, and it is especially significant with a spread of the corpus.