Search and Navigation in Semantically Integrated Document Collections

The paper presents a novel approach to semantic search and navigation in office-like document collections. The approach is based on a semantic document model that we have developed to enable unique identification, semantic annotation, and semantic linking of document units of officelike documents. In order to semantically annotate document units and to link semantically related document units, we first conceptualize document units’ semantics and represent them by vectors of ontological concepts and their corresponding weight vectors. In the semantic search, we represent a user query by a query’s concept vector, which is generated in the same way as document units’ concept vectors, and then determine the search results by measuring the similarity between the query’s and the document units’ concept weight vectors. After the search, by following the semantic links of a selected document unit, the user can navigate through the document collection and discover semantically related document units. Results of the preliminary evaluation, conducted with a prototype implementation, are promising. We present a brief analysis of these results. Keywords-semantic search, semantic linking and navigation;