Seed, a natural language interface to knowledge bases

Abstract

The World Wide Web has been rapidly developing in the last decade. In recent years, the Semantic Web has gained a lot of traction. It is a vision of the Web where data is understandable by machines as well as humans. Developments in the Semantic Web made way for the creation of massive knowledge bases containing a wealth of structured information. However, allowing end-users to interact with and benefit from these knowledge bases remains a challenge. In this paper, we present Seed, an extensible knowledge-supported natural language text composition tool, which provides a user-friendly way of interacting with complex knowledge systems. It is integrable not only with public knowledge bases on the Semantic Web, but also with private knowledge bases used in personal or enterprise contexts. By means of a long-term formative user-study and a short-term user evaluation of a sizable population of test subjects, we show that Seed was successfully used in exploring, modifying and creating the content of complex knowledge bases. We show it enables end-users do so with nearly no domain knowledge while hiding the complexity of the underlying knowledge representation.