LINKING ENTITIES TO A KNOWLEDGE BASE

Abstract

The explosive growth in the amount of textual information brings a need for building a structured Knowledge Base (KB) to organize the knowledge scattered among these unstructured texts. On the other hand, the available KBs such as Wikipedia and Google Knowledge Graph which contain rich knowledge about the world¿s entities have been shown to form a valuable component for many natural language precessing tasks. To populate or to utilize the KBs, we need to link the mentions of entities in text to their corresponding entries in the KB, which is called entity linking. This thesis proposes annotation acquisition, instance selection, topic model and lazy learning for entity linking. Finally, I also propose a framework for entity linking in microblog.