13.07.2015 Views

WWW/Internet - Portal do Software Público Brasileiro

WWW/Internet - Portal do Software Público Brasileiro

WWW/Internet - Portal do Software Público Brasileiro

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

IADIS International Conference <strong>WWW</strong>/<strong>Internet</strong> 20102. RELATED WORKSeveral publications address the topic of semantic image annotation. In general, there are two establishedapproaches to enrich bitmap data with semantics.On the one hand side various research endeavors favor a semi-automatic solution. Even though systemssuggest various annotations and support the annotation process itself, <strong>do</strong>main experts decide whether theautomatic tagging is eligible. For example Schreiber et al. describe in “Ontology-Based Photo Annotation”[Schreiber et al. 2001] a tool that allows for annotating and querying photographs. The annotation ontology isbased on the RDFS standard [RDFS 2004] and capable of capturing photo features (e.g. photographer),medium features (e.g. resolution) and subject-matter descriptions (e.g. setting). In order to finalize theannotation process experts are required to validate all produced annotations. A second accepted approachfocuses on automated semantic annotation systems. Those autonomous systems scale and are cost effective.Tansley et al. implemented an automated solution called MAVIS 2 [Tansley et al. 2000]. The semanticclassification combines latent semantic analysis with proprietary semantic classification algorithms.Resulting classifications are produced automatically but may include incorrect annotations.Currently, all existing systems are not able to join the cost-effectiveness of automated systems with theannotation accuracy of expert-supported systems. Our proposed solution offers cost-effectiveness as well asannotation accuracy for the software modeling <strong>do</strong>main.3. ARCHITECTURE & IMPLEMENTATION OF THE SEMANTICIMAGE ANNOTATION PLATFORMThe semantic image annotation platform derives bitmaps from the visual representation of software modelsand enriches the generated bitmap data with the semantics of the displayed content. Figure 1 represents theplatform architecture.Figure 1. Platform architectureFigure 2. Platform implementationThe architecture is split into 3 components: the developer’s workbench, the content extraction platformand the content management system (CMS).The developer’s workbench provides the tooling to create & modify software models. Dedicated editorsare in place to manipulate software models in an ergonomic way. The resulting models are persisted in arepository and adhere to a so called meta-model. This meta-model specifies the language vocabulary whichwill be instantiated by models. The content extraction platform consists of a bitmap extractor that transformsthe model visualization into a bitmap format. The semantics extractor analyses the model itself (modelelement properties and relations to other model elements are analyzed, constraints are calculated, etc). Sincethis analysis is executed on the software model and not on the respective bitmap data, the entire data modelcan be traversed and exploited in order to extract its semantics. Finally, the export component combinesbinary bitmap data with textual semantics data and transmits it to the CMS. The CMS stores bitmap andsemantic data. Furthermore, it links bitmap instances to ontological instances. Therefore, raw bitmap data isalways associated with its semantics.407

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!