Justifying Semantic Search Results By Means Of Semantic Frequency Classes

Justifying Semantic Search Results By Means Of Semantic Frequency Classes
Björn Forcher, Stefan Agne, Andreas Dengel, Thomas Roth-Berghofer
Proceedings of the IJCAI -11 Workshop on Explanation-aware Computing (ExACt 2011) International Workshop on Explanation-aware Computing (ExaCt-2011), July 16-17, Barcelona, Spain

Abstract:

The research project Emergent aims at developing concepts, technologies and business processes for emergent software. One subgoal is to evolve the semantic search engine Koios that will be used for different domains such as medicine and agriculture. Since semantic search results are not always self-explanatory, an explanation facility is integrated into the engine revealing connections between search and result concepts. The constructed explanations are depicted as semantic networks containing various domain specific concepts. To provide good explanations, concepts must be labeled with adequate terms regarding the level of expertise of their users. In this paper, we present a user experiment and, based on that, define a method that predicts the understandability of terms by means of semantic frequency classes. Regarding the mentioned domains and levels of expertise the vision is to select the most understandable label for a concept if there is more than one label available.