Abstract

In specific domains, such as e-commerce, inducing a taxonomy over a given set of ‘target’ concepts is an important problem with applications ranging from good website design to knowledge organization and recommender systems. Automatically inducing a full or ‘global’ taxonomy over a large set of concepts is a difficult problem in the AI literature, typically requiring human intervention. A more tractable version of the problem is called Local Taxonomy Construction (LTC). Rather than induce a global taxonomy, LTC attempts to induce the local neighborhood of each concept in the target concept-set. Despite having much practical importance, LTC has not been properly formalized and explored in the applied AI community. In this paper, we present such a formalism on LTC, including a set of viable, minimally supervised solutions based on pre-existing representation learning algorithms in the natural language …

Date

2022

Authors

Mayank Kejriwal, Ravi Kiran Selvam, Chien-Chun Ni, Nicolas Torzec

Book

Social Media Analysis for Event Detection

Pages

133-161

Publisher

Springer International Publishing