Answering Complex Logical Queries on Knowledge Graphs via Query Computation Tree Optimization

19 Dec 2022  ·  Yushi Bai, Xin Lv, Juanzi Li, Lei Hou ·

Answering complex logical queries on incomplete knowledge graphs is a challenging task, and has been widely studied. Embedding-based methods require training on complex queries, and cannot generalize well to out-of-distribution query structures. Recent work frames this task as an end-to-end optimization problem, and it only requires a pretrained link predictor. However, due to the exponentially large combinatorial search space, the optimal solution can only be approximated, limiting the final accuracy. In this work, we propose QTO (Query Computation Tree Optimization) that can efficiently find the exact optimal solution. QTO finds the optimal solution by a forward-backward propagation on the tree-like computation graph, i.e., query computation tree. In particular, QTO utilizes the independence encoded in the query computation tree to reduce the search space, where only local computations are involved during the optimization procedure. Experiments on 3 datasets show that QTO obtains state-of-the-art performance on complex query answering, outperforming previous best results by an average of 22%. Moreover, QTO can interpret the intermediate solutions for each of the one-hop atoms in the query with over 90% accuracy. The code of our paper is at https://github.com/bys0318/QTO.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Complex Query Answering FB15k QTO MRR 1p 0.895 # 1
MRR 2p 0.674 # 1
MRR 3p 0.588 # 1
MRR 2i 0.803 # 1
MRR 3i 0.836 # 1
MRR pi 0.752 # 1
MRR ip 0.740 # 1
MRR 2u 0.767 # 1
MRR up 0.613 # 1
Complex Query Answering FB15k-237 QTO MRR 1p 0.490 # 1
MRR 2p 0.214 # 1
MRR 3p 0.212 # 1
MRR 2i 0.431 # 1
MRR 3i 0.568 # 1
MRR pi 0.381 # 1
MRR ip 0.280 # 1
MRR 2u 0.227 # 1
MRR up 0.214 # 1
Complex Query Answering NELL-995 QTO MRR 1p 0.607 # 1
MRR 2p 0.241 # 1
MRR 3p 0.216 # 1
MRR 2i 0.425 # 2
MRR 3i 0.506 # 2
MRR pi 0.313 # 1
MRR ip 0.265 # 1
MRR 2u 0.204 # 1
MRR up 0.179 # 1

Methods


No methods listed for this paper. Add relevant methods here