A Knowledge Based System for Content-based Retrieval of Scalable Vector Graphics Documents

Scalable Vector Graphics (SVG), the novel XML based language for describing two-dimensional graphics, is now a W3C standard and it is likely to become popular on the Internet, due to its inherent advantages over raster image formats in several domains. We present a system for semantic based retrieval by content of SVG. The system is endowed of a web crawler for documents search and a graphical interface for query by sketch. The approach adopted in the system implements a simple description logic devised for the semantic indexing and retrieval of complex objects. Its syntax allows to describe basic shapes and complex objects as compositions of basic ones, and transformations. Its extensional semantics, which is compositional, allows to define retrieval, classification, and subsumption services. An experimental evaluation is also presented, which shows results obtained in terms of precision and recall, but also points out that there are still few SVG documents available on the Web.