As the research of knowledge graph(KG)is deepened and widely used,knowledge graph com-pletion(KGC)has attracted more and more attentions from researchers,especially in scenarios of in-telligent search,social networks ...As the research of knowledge graph(KG)is deepened and widely used,knowledge graph com-pletion(KGC)has attracted more and more attentions from researchers,especially in scenarios of in-telligent search,social networks and deep question and answer(Q&A).Current research mainly fo-cuses on the completion of static knowledge graphs,and the temporal information in temporal knowl-edge graphs(TKGs)is ignored.However,the temporal information is definitely very helpful for the completion.Note that existing researches on temporal knowledge graph completion are difficult to process temporal information and to integrate entities,relations and time well.In this work,a rotation and scaling(RotatS)model is proposed,which learns rotation and scaling transformations from head entity embedding to tail entity embedding in 3D spaces to capture the information of time and rela-tions in the temporal knowledge graph.The performance of the proposed RotatS model have been evaluated by comparison with several baselines under similar experimental conditions and space com-plexity on four typical knowl good graph completion datasets publicly available online.The study shows that RotatS can achieve good results in terms of prediction accuracy.展开更多
Temporal knowledge graph(TKG)reasoning has emerged as a pivotal approach in event prediction.An important yet challenging task in TKG reasoning is to predict future events by extrapolating from historical events and t...Temporal knowledge graph(TKG)reasoning has emerged as a pivotal approach in event prediction.An important yet challenging task in TKG reasoning is to predict future events by extrapolating from historical events and their correlations.Existing methods either overlook the modeling of long-term dependencies between entities or are ineffective in aggregating long-term information with recent facts.Motivated by dual process theory in cognitive sciences,we introduce TKG-LDG,an approach enhancing TKG for future entity prediction with long-term dense graph,to model event evolution in an adaptive manner.We first construct a unified dense graph from historical data to capture long-term dependencies,reflecting cumulative knowledge of entity interactions over time.This unified dense graph is compatible with any graph neural network and facilitates entity interaction learning from a long-term perspective.Then we initialize a TKG encoder from the unified dense graph to enhance short-term event interaction modeling.TKG-LDG effectively marries global context with local adaptability to recent temporal changes through its short-term recurrent encoders,in a way that mirrors human reasoning by integrating both long-term and short-term event dynamics.Extensive experiments conducted on six widely used TKG datasets demonstrate that our model outperforms strong baselines in future event prediction.展开更多
Temporal knowledge graph completion(TKGC),which merges temporal information into traditional static knowledge graph completion(SKGC),has garnered increasing attention recently.Among numerous emerging approaches,transl...Temporal knowledge graph completion(TKGC),which merges temporal information into traditional static knowledge graph completion(SKGC),has garnered increasing attention recently.Among numerous emerging approaches,translation-based embedding models constitute a prominent approach in TKGC research.However,existing translation-based methods typically incorporate timestamps into entities or relations,rather than utilizing them independently.This practice fails to fully exploit the rich semantics inherent in temporal information,thereby weakening the expressive capability of models.To address this limitation,we propose embedding timestamps,like entities and relations,in one or more dedicated semantic spaces.After projecting all embeddings into a shared space,we use the relation-timestamp pair instead of the conventional relation embedding as the translation vector between head and tail entities.Our method elevates timestamps to the same representational significance as entities and relations.Based on this strategy,we introduce two novel translation-based embedding models:TE-TransR and TE-TransT.With the independent representation of timestamps,our method not only enhances capabilities in link prediction but also facilitates a relatively underexplored task,namely time prediction.To further bolster the precision and reliability of time prediction,we introduce a granular,time unit-based timestamp setting and a relation-specific evaluation protocol.Extensive experiments demonstrate that our models achieve strong performance on link prediction benchmarks,with TE-TransR outperforming existing baselines in the time prediction task.展开更多
1 Introduction Temporal Knowledge Graphs(TKGs)provide a dynamic framework for modeling evolving events and relationships over time,with applications ranging from stock market to international politics.As to stock mark...1 Introduction Temporal Knowledge Graphs(TKGs)provide a dynamic framework for modeling evolving events and relationships over time,with applications ranging from stock market to international politics.As to stock market,TKGs can model how these relationships change over time,enabling the prediction of stock price movements,market trends,and potential risks.While graph-based methods such as Graph Neural Networks(GNNs)[1,2]have been widely adopted for TKG extrapolation,we argue that their structural focus often overshadows the critical role of historical information.Historical periodicity and temporal patterns serve as the foundation for effective temporal reasoning,particularly in forecasting future events.展开更多
基金the National Natural Science Foundation of China(No.6187022153).
文摘As the research of knowledge graph(KG)is deepened and widely used,knowledge graph com-pletion(KGC)has attracted more and more attentions from researchers,especially in scenarios of in-telligent search,social networks and deep question and answer(Q&A).Current research mainly fo-cuses on the completion of static knowledge graphs,and the temporal information in temporal knowl-edge graphs(TKGs)is ignored.However,the temporal information is definitely very helpful for the completion.Note that existing researches on temporal knowledge graph completion are difficult to process temporal information and to integrate entities,relations and time well.In this work,a rotation and scaling(RotatS)model is proposed,which learns rotation and scaling transformations from head entity embedding to tail entity embedding in 3D spaces to capture the information of time and rela-tions in the temporal knowledge graph.The performance of the proposed RotatS model have been evaluated by comparison with several baselines under similar experimental conditions and space com-plexity on four typical knowl good graph completion datasets publicly available online.The study shows that RotatS can achieve good results in terms of prediction accuracy.
基金supported by the National Natural Science Foundation of China(Nos.62176043 and 62072077)the Intelligent Terminal Key Laboratory of Sichuan Province(No.SCITLAB-30002).
文摘Temporal knowledge graph(TKG)reasoning has emerged as a pivotal approach in event prediction.An important yet challenging task in TKG reasoning is to predict future events by extrapolating from historical events and their correlations.Existing methods either overlook the modeling of long-term dependencies between entities or are ineffective in aggregating long-term information with recent facts.Motivated by dual process theory in cognitive sciences,we introduce TKG-LDG,an approach enhancing TKG for future entity prediction with long-term dense graph,to model event evolution in an adaptive manner.We first construct a unified dense graph from historical data to capture long-term dependencies,reflecting cumulative knowledge of entity interactions over time.This unified dense graph is compatible with any graph neural network and facilitates entity interaction learning from a long-term perspective.Then we initialize a TKG encoder from the unified dense graph to enhance short-term event interaction modeling.TKG-LDG effectively marries global context with local adaptability to recent temporal changes through its short-term recurrent encoders,in a way that mirrors human reasoning by integrating both long-term and short-term event dynamics.Extensive experiments conducted on six widely used TKG datasets demonstrate that our model outperforms strong baselines in future event prediction.
基金supported by the National Natural Science Foundation of China under Grant No.72293575.
文摘Temporal knowledge graph completion(TKGC),which merges temporal information into traditional static knowledge graph completion(SKGC),has garnered increasing attention recently.Among numerous emerging approaches,translation-based embedding models constitute a prominent approach in TKGC research.However,existing translation-based methods typically incorporate timestamps into entities or relations,rather than utilizing them independently.This practice fails to fully exploit the rich semantics inherent in temporal information,thereby weakening the expressive capability of models.To address this limitation,we propose embedding timestamps,like entities and relations,in one or more dedicated semantic spaces.After projecting all embeddings into a shared space,we use the relation-timestamp pair instead of the conventional relation embedding as the translation vector between head and tail entities.Our method elevates timestamps to the same representational significance as entities and relations.Based on this strategy,we introduce two novel translation-based embedding models:TE-TransR and TE-TransT.With the independent representation of timestamps,our method not only enhances capabilities in link prediction but also facilitates a relatively underexplored task,namely time prediction.To further bolster the precision and reliability of time prediction,we introduce a granular,time unit-based timestamp setting and a relation-specific evaluation protocol.Extensive experiments demonstrate that our models achieve strong performance on link prediction benchmarks,with TE-TransR outperforming existing baselines in the time prediction task.
基金supported by the National Natural Science Foundation of China(Grant Nos.62020106005,42050105,62061146002)Shanghai Pilot Program for Basic Research-Shanghai Jiao Tong University.
文摘1 Introduction Temporal Knowledge Graphs(TKGs)provide a dynamic framework for modeling evolving events and relationships over time,with applications ranging from stock market to international politics.As to stock market,TKGs can model how these relationships change over time,enabling the prediction of stock price movements,market trends,and potential risks.While graph-based methods such as Graph Neural Networks(GNNs)[1,2]have been widely adopted for TKG extrapolation,we argue that their structural focus often overshadows the critical role of historical information.Historical periodicity and temporal patterns serve as the foundation for effective temporal reasoning,particularly in forecasting future events.
