The Knowledge Awakens: Keeping Knowledge Bases Fresh with Emerging Entities.

Entity search over news, social media and the Web allows users to precisely retrieve concise information about specific people, organizations, movies and their characters, and other kinds of entities. This expressive search mode builds on two major assets: 1) a knowledge base (KB) that contains the entities of interest and 2) entity markup in the documents of interest derived by automatic disambiguation of entity names (NED) and linking names to the KB. These prerequisites are not easily available, though, in the important case when a user is interested in a newly emerging entity (EE) such as new movies, new songs, etc. Automatic methods for detecting and canonicalizing EEs are not nearly at the same level as the NED methods for prominent entities that have rich descriptions in the KB. To overcome this major limitation, we have developed an approach and prototype system that allows searching for EEs in a user-friendly manner. The approach leverages the human in the loop by prompting for user feedback on candidate entities and on characteristic keyphrases for EEs. For convenience and low burden on users, this process is supported by the automatic harvesting oftentative keyphrases. Our demo system shows this interactive process and its high usability.