Hierarchical Graph Model for Incremental Data Organization

在这项研究中，我们提出了层次图模型，使半结构化数据的灵活结构增量。在这里，我们将重点介绍以试错的方式收集数据并为其添加属性的方法。在该模型中，层次图被定义为有向图。我们将“视点”明确地视为数据，并将半结构化数据和视点都视为图上的节点。图形由组件对象和图形对象组成。前者具有数据和“视点”的属性。后者维系着他们的关系。通过该模型，用户可以根据自己的观点对半结构化数据进行多种表示。我们引入虚拟对象是为了支持根据用户的特殊直觉或假设进行结构化。用户可以在运行时生成虚拟对象，并将其视为模型中的数据。我们还展示了一个原型系统，并展示了我们模型的功能。

In this research, we have proposed the hierarchical graph model that enables flexible structuring of semistructured data incrementally. Here we focus on the ways to collect data and add attributes to them in a trial and error manner. In this model, hierarchical grpah is defined as a directed graph. We treat "viewpoints" as data expicitly, and treat both semi-structured data and viewpoints as nodes on the graph. The graph consists of component objects and a graph object. The former possess the attributes of data and "viewpoints." The latter holds their relations. By this model, users can give multiple representations of semi-structured data based on user's viewpoints. We introduce virtual objects in order to support the structuring according to user's ad-hoc intuitions or hypotheses. Users can generate virtual objects at run-time, and treat them as data in our model. We have also shown a prototype system, and showed capabilities of our model.