home..
Logic Embeddings for Complex Query Answering (2021)
huyi / December 2022
0. Abstract
task and challenges
task: answering logical queries over incomplete knowledge bases
challenges:
- imcomplete KG –> call for link prediction
- brute force answering of existential first-order logic queries is exponential in the number of existential variables
previous work
that doesn’t support negation
Recent work of query embeddings provides fast querying, but most approaches model set logic with closed regions, so lack negation.
这里的recent work是query2box?that support negation
use densities, triggering drawbacks:
- only improvise logic
- use expensive distributions
- poorly model answer uncertainty
this work
- support negation
- improve on density approaches: 1) integrates well-studied t-norm logic and directly evaluates satisfiability, 2) simplifies modeling with truth values, 3) models uncertainty with truth bounds
contribution
- competitively fast and accurate in query answering over large, incomplete knowledge graphs
- outperform on negation queries
- provide improved modeling of answer uncertainty