Abstract. Hierarchical structures are a way to organize and enrich semantically the available information on the Web. Popular examples of such structures are the product catalogs of e-market stores, which provide data (i.e. products) organized in thematic
PatMan: A Visual Database System to Manipulate Path Patterns and Data in
Hierarchical Catalogs
Angeliki Koukoutsaki, Theodore Dalamagas, Timos Sellis, Panagiotis Bouros
Knowledge and Database Systems Lab
School of Electrical and Computer Engineering National Techn. Univ. of Athens, Athens, Greece {akoukou,dalamag,timos,pbour}@dblab.ece.ntua.gr
Abstract. Hierarchical structures are a way to organize and enrich semantically the available information on the Web. Popular examples of such structures are the product catalogs of e-market stores, which provide data (i.e. products) organized in thematic hierarchies on a category/subcategory basis. Users can navigate these hierarchies to identify data of their preference. Such hierarchies group data under certain properties. Navigational paths are the knowledge artifacts to represent such groups. We consider paths in hierarchies as patterns which provide a conceptual clustering of data in groups sharing common properties. Under this perspective, we have designed and implemented PatMan, a visual database system to manage hierarchical catalogs. The system can store navigational paths and data from hierarchical catalogs, and provides a pictorial query-by-example language to manipulate path patterns and data in a uniform way.
1. Introduction
The Internet is today's greatest source of information. Huge volumes of data are posted and retrieved through the Web. Despite this vast exchange of information on the web, there is no consistent and strict organization of data [1]. Hierarchical structures are a way to organize and enrich semantically the available information on the Web. Popular examples of such structures are the product catalogs of e-market stores which provide data (i.e. products) organized in thematic hierarchies on a category/subcategory basis. Users can navigate these hierarchies to identify data of their preference. Such hierarchies group data under certain properties. Navigational paths in these structures are the knowledge artifacts to represent such groups, and act as semantic guides to reach the data of each group either through browsing or through path expression query languages. For example, the path /moto/bmw/funduros/ represents several BMW motorcycle models designed for off-road riding and can be used in a path expression to search for motorcycles with price less than 9000: /moto/bmw/funduros[price<9000].
We consider paths in hierarchies as patterns which provide a conceptual clustering of raw data in groups sharing common properties. We focus on the extraction and manipulation of these patterns, combined with the manipulation of data. See for example Figure 1, where two portal catalogs, Adorama and B&H1, are presented. These catalogs provide photo equipment organized in a hierarchy on a category/subcategory basis. Searching for lenses in the first catalog needs the path /cameras&lenses/lenses, while in the second catalog needs the path /photo/35mm systems/lenses. Such paths can be seen as alternative pattern versions for the same group of data.
1 ,
Abstract. Hierarchical structures are a way to organize and enrich semantically the available information on the Web. Popular examples of such structures are the product catalogs of e-market stores, which provide data (i.e. products) organized in thematic
On the other hand, there are cases where the owner of a catalog may need to provide integrated photo systems, with camera bodies from Adorama and the appropriate lenses from B&H. In such case, the two paths /cameras&lenses/35mm SLR and /photo/35mm systems/lenses form a kind of complex pattern representing different raw data, i.e. bodies and lenses, that should be grouped together.
Figure 1: Parts of Adorama and B & H catalogs.
Our work addresses part of the requirements defined for pattern management in [12], [14], emphasizing on the management of path-like patterns and data in hierarchical catalogs. Similar requirements have been also introduced in the inductive database framework [6], where patterns are integrated within the database environment as full-fledged objects. Under this context, a number of specialized inductive query languages have been proposed but most of these concern descriptive rules [5], [10]. Path manipulation techniques for biochemical pathways treated as knowledge artifacts, i.e. biochemical reactions, are discussed in [8] and [13]. In these works, database systems to manipulate and query biochemical pathways are presented. The problem of manipulating paths (and tree-like structures, in general) appears also in XML data management. In [4] the authors present an algebra for XML data, but it is tuple-based and not tree-based. A navigational algebra is presented in [9], but again it treats individual nodes as manipulation units. In [7], TAX algebra is defined, but as a means for selecting and reconstructing bulk XML data.