Autoentity: automated entity detection from massive text corpora

Abstract

Entity detection is one of the fundamental tasks in Natural Language Processing and Information Retrieval. Most existing methods rely on human annotated data and hand-crafted linguistic features, which makes it hard to apply the model to an emerging domain. In this paper, we propose a novel automated entity detection framework, called AutoEntity, that performs automated phrase mining to create entity mention candidates and enforces lexico-syntactic rules to select entity mentions from candidates. Our experiments on real-world datasets in different domains and multiple languages have demonstrated the effectiveness and robustness of the proposed method.