Showing 1–1 of 1 results for author: Torzec, N

Layered Graph Embedding for Entity Recommendation using Wikipedia in the Yahoo! Knowledge Graph

Authors: Chien-Chun Ni, Kin Sum Liu, Nicolas Torzec

Abstract: In this paper, we describe an embedding-based entity recommendation framework for Wikipedia that organizes Wikipedia into a collection of graphs layered on top of each other, learns complementary entity representations from their topology and content, and combines them with a lightweight learning-to-rank approach to recommend related entities on Wikipedia. Through offline and online evaluations, w… ▽ More In this paper, we describe an embedding-based entity recommendation framework for Wikipedia that organizes Wikipedia into a collection of graphs layered on top of each other, learns complementary entity representations from their topology and content, and combines them with a lightweight learning-to-rank approach to recommend related entities on Wikipedia. Through offline and online evaluations, we show that the resulting embeddings and recommendations perform well in terms of quality and user engagement. Balancing simplicity and quality, this framework provides default entity recommendations for English and other languages in the Yahoo! Knowledge Graph, which Wikipedia is a core subset of. △ Less

Submitted 14 April, 2020; originally announced April 2020.

Comments: 8 pages, 4 figures, 8 tables. To be appeared in Wiki Workshop 2020, Companion Proceedings of the Web Conference 2020(WWW 20 Companion), Taipei, Taiwan

ACM Class: H.3.3