Summarization of Concepts for Visual Disambiguation

Controlled vocabularies that power semantic applications allow them to operate with high precision, which comes with a price of having to disambiguate between senses of terms. Fully automatic disambiguation is a largely unsolved problem and semi-automatic approaches are preferred. These approaches involve users to do the disambiguation and require an adequate user interface. However, term definitions are usually lengthy and not only occupy valuable screen space, but reading and understanding these definitions requires the user’s attention and time. As an alternative to using definitions we propose to use a summary — a “single word” disambiguation label for a concept. In this paper we present an algorithm to summarize concepts from a controlled vocabulary. We evaluate the algorithm with 51 users and show that the algorithm generates summaries that have good discriminative and associative qualities. In addition, the length of summaries are comparable to the length of the original terms, thus making the algorithm particularly useful in situations where screen estate is limited.