Gram A Graph Data Model and Query Language.pdfVIP

Gram: A Graph Data Model and Query Language Bernd Amann INRIA F-78153 Le Chesnay Cedex, France Michel Scholl Cedric/CNAM 292 rue St Martin, F-75141 Paris Cedex 03, France Abstract We present a model for data organized as graphs. Regular expressions over the types of the node and edge labels are used to qualify connected subgraphs. An algebraic lan- guage based on these regular expressions and supporting a restricted form of recursion is introduced. A natural ap- plication of this model and its query language is hypertext querying. 1 Introduction Recent database [13, 5] research work shows a growing interest in the definition of graph models and languages to allow a natural way of handling data appearing in appli- cations such as hypertext or geographic database systems. Standard data models are often inefficient as they do not capture the inherent structure of data representing hyper- text documents [4, 7, 18] or networks (highways, rivers, . . . ) [12]. In this paper, we present a graph data model. Its appli- cation to hypertext querying is illustrated by an example of a travel agency that organizes journeys. We think of a hypertext as a directed labeled graph where the nodes are typed documents and the edges correspond to typed span-to-span 2 links between documents [15]. A journey corresponds to a sequence of stops in several cities, where hotels, restaurants and monuments are vis- ited. Figure 1 shows a schema, also structured as a graph. Nodes: Document Types Edges: Link Types CITY STOP JOURNEYHOTELRESTAURANT first in addraddr addr MONUMENT dist next Figure 1: The Travel Agency Schema A document can have one of the following types: STOP, JOURNEY, CITY, HOTEL, RESTAURANT and MONUMENT. A sample of the database graph is given in Figure 2. Note that links are typed and may contain information such as a date, an address or a distance in kilometers. Since restaurants like McDonalds can be de- scribed independently from the city where they are lo-