Entity replacement strategy for temporal knowledge graph query relaxation.

Area of Science:

• Artificial Intelligence
• Data Science
• Knowledge Representation

Background:

• Temporal knowledge graph (TKG) queries are vital for retrieving time-constrained information.
• Current TKG query methods often fail with queries yielding no results, limiting their practical application.
• Enhancing query flexibility, especially for empty result scenarios, is an open research challenge.

Purpose of the Study:

• To propose a novel query relaxation method, Entity Replacement for Temporal Knowledge Graph Query Relaxation (ER-TQR), specifically designed for TKG queries that return empty results.
• To enhance the flexibility and robustness of temporal knowledge graph querying systems.
• To improve the accuracy and relevance of results for temporal knowledge graph queries.

Main Methods:

• Developed ER-TQR, a query relaxation technique employing an entity replacement strategy.
• Utilized Bidirectional Encoder Representations from Transformers (BERT) for enhanced semantic query embedding.
• Integrated a Bidirectional Gated Recurrent Unit (Bi-GRU) model to identify and replace problematic entities based on semantic and temporal alignment.

Main Results:

• ER-TQR demonstrated superior performance in handling temporal knowledge graph queries with empty results.
• The method successfully replaced incompatible entities with semantically and temporally relevant candidates, preserving query intent.
• Experimental results showed ER-TQR outperforming existing query relaxation methods across 4 out of 5 evaluation metrics on diverse datasets.

Conclusions:

• ER-TQR offers a significant advancement in temporal knowledge graph query relaxation, particularly for empty result sets.
• The proposed entity replacement strategy, combined with BERT and Bi-GRU embeddings, effectively enhances query flexibility and accuracy.
• This research paves the way for more robust and user-friendly temporal knowledge graph querying systems.