.png){kind=icon}
.png){kind=icon}
Return
MGIT: Multi-Granularity implicit temporal framework for knowledge graph question answering
.png){kind=icon}
PRE
AI
DOI：10.1007/s10489-026-07404-w.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Temporal Knowledge Graph Question Answering (TKGQA) aims to retrieve entities or timestamps from temporal knowledge graphs in response to temporally constrained queries; however, existing methods primarily rely on single-granularity temporal representations and lack the ability to effectively handle multi-granularity temporal reasoning, leading to suboptimal performance on complex queries. To address this issue, we propose a Multi-Granularity Implicit Temporal (MGIT) framework that enhances temporal representation and reasoning by modeling implicit temporal dependencies across different granularities. Specifically, MGIT integrates a graph attention network with multi-head attention to capture structural and implicit temporal relationships, while a combination of convolutional neural networks and gated recurrent units is employed to jointly model local temporal patterns and long-range dependencies. In addition, a self-learning encoding strategy is introduced to dynamically adapt to heterogeneous temporal granularities, and an ensemble learning mechanism is adopted to aggregate representations from different temporal segments for improved robustness. Extensive experiments on both MultiTQ dataset and CronQuestions dataset demonstrate that MGIT consistently outperforms state-of-the-art baselines, highlighting its effectiveness in capturing implicit temporal information and enhancing multi-granularity temporal reasoning.
Keywords:
Temporal knowledge graph
Multi-Granularity
Temporal reasoning
Ensemble learning
Journal
IF:
3.5
Papers:
7.5K
Citations:
1.7W