Temporal Knowledge Graph Link Prediction Using Synergized Large Language Models and Temporal Knowledge Graphs

EasyChair Preprint 13243, version 2

Abstract

Although large language models and temporal knowledge graphs each have significant advantages in the field of artificial intelligence, they also face certain challenges. However, through collaboration, large language models and temporal knowledge graphs can complement each other, addressing their respective shortcomings. This collaborative approach aims to harness the potential feasibility and practical effectiveness of large language models as external knowledge bases for temporal knowledge graph reasoning tasks.In our research, we have meticulously designed a synergized model that leverages the knowledge from the graph as prompts. The answers generated by the large language model undergo careful processing before being seamlessly incorporated into the training dataset. The ultimate goal is to significantly enhance the reasoning capabilities of temporal knowledge graphs. Experimental results underscore the positive impact of this synergized model on the completion tasks of temporal knowledge graphs, showcasing its potential to address gaps in knowledge and improve overall performance. While its influence on prediction tasks is relatively weak, the collaborative synergy demonstrates promising avenues for further exploration and development in the realm of AI research.

Keyphrases: Completion task, Synergetic pattern, Temporal Knowledge Graphs, large language models, prediction task

@booklet{EasyChair:13243,
  author    = {Yao Chen and Yuming Shen},
  title     = {Temporal Knowledge Graph Link Prediction Using Synergized Large Language Models and Temporal Knowledge Graphs},
  howpublished = {EasyChair Preprint 13243},
  year      = {EasyChair, 2024}}