Interactive Concept Mining on Personal Data -- Bootstrapping Semantic Services

摘要: Semantic services (e.g. Desktops) are still afflicted by a cold start problem: in the beginning, user's personal information sphere, i.e. files, mails, bookmarks, etc., is not represented system. Information extraction tools used to kick-start system typically create 1:1 representations of different items. Higher level concepts, for example found file names, mail subjects or content body these items, extracted. Leaving concepts out may lead underperformance, having many them making every term concept) will clutter arising knowledge graph with non-helpful relations. In this paper, we present an interactive concept mining approach proposing candidates gathered exploiting given schemata usual management applications and analysing sphere using various metrics. To heed subjective view user, graphical user interface allows easily rank give feedback on proposed candidates, thus keeping only those actually considered relevant. A prototypical implementation demonstrates major steps our approach.