NeurIPS 2020

Faithful Embeddings for Knowledge Base Queries

Meta Review

This paper presents an embedding based neural query language called EmQL, that generalizes to the unknown facts in KB and also performs logical entailments better than existing methods like Qeury2Box. Strength • The proposed method is sound and novel. • Experimental results are convincing. • The paper is clearly written. Weakness • There are underlying assumption in the proposed approach. • The presentation can be further improved.