The approach is called OWL-Miner. It is a rule-learner that explores the space of (OWL) concept descriptions to find one that best describes the training data. 这种方法被称为OWL-Miner。它是一个规则学习者，探索(OWL)概念描述的空间，以找到最能描述训练数据的空间。
    It has access to an OWL ontology describing the application domain from which the training data is drawn, and it uses this to find very readable descriptions of potentially complex relationships among things, for example, relationships amongst the structural components of an object. 它可以访问描述从中提取训练数据的应用领域的OWL本体，并使用它来找到事物之间潜在的复杂关系的可读描述，例如，对象的结构组件之间的关系。
    OWL-miner can be used for a range of learning problems, but for this purpose we describe it as a binary classification algorithm. OWL-miner可以用于一系列的学习问题，但为此我们将其描述为一个二进制分类算法。

    The problem can be stated as follows:
    Given a knowledge base comprising 给定知识库

    • An OWL ontology of axioms introducing the classes and relationships of the problem domain, including an identified target class 引入问题域的类和关系的OWL公理本体，包括一个已识别的目标类
    • An hypothesis language Semantic Web Mining 语义网挖掘 - 图1 defined as a choice of class constuctors (such as union, intersection, negation, cardinality, etc).
    • RDF triples giving positive example instances of the target class, and _negative _example instances that are not in the target class, 给出目标类的正面样例和反面样例。
    • RDF triples that further describe the positive and negative examples by features expressed as class and property instances of the ontology; 通过表示为本体的类和属性实例的特征来进一步描述正面和负面的例子
    • A quality function to assess hypotheses (such as accuracy or f-measure) 评估假设的质量功能

    Find an OWL class expression in Semantic Web Mining 语义网挖掘 - 图2 that describes the positive examples but does not describe the negative examples to optimally satisfy a bounded quality function (such as accuracy > 95%). 在中找到一个描述正面示例但不描述负面示例的OWL类表达式，以最佳地满足有界质量函数(例如精度> 95%)。

    OWL-miner is a top-down refinement-based learner that searches for useful class descriptions by (in principle) enumerating every possible expression and testing them for quality against the training data. It starts with a very simple expression and gradually builds a tree of more complex expressions from there, checking them as it goes.
    OWL-miner是一个自上而下的基于细化的学习者，它通过(原则上)枚举每个可能的表达式并根据训练数据测试它们的质量来搜索有用的类描述。它从一个非常简单的表达式开始，并逐渐从那里构建一个更复杂的表达式树，并在运行时检查它们。