Temporal Fact Reasoning over Hyper-Relational Knowledge Graphs

TL;DR

A new type of data structure named hyper-relational TKG (HTKG) is proposed, coupled with a timestamp explicitly indicating its time validity, and every fact in an HTKG is coupled with a timestamp explicitly indicating its time validity.

Abstract

Stemming from traditional knowledge graphs (KGs), hyper-relational KGs (HKGs) provide additional key-value pairs (i.e., qualifiers) for each KG fact that help to better restrict the fact validity. In recent years, there has been an increasing interest in studying graph reasoning over HKGs. Meanwhile, as discussed in recent works that focus on temporal KGs (TKGs), world knowledge is ever-evolving, making it important to reason over temporal facts in KGs. Previous mainstream benchmark HKGs do not explicitly specify temporal information for each HKG fact. Therefore, almost all existing HKG reasoning approaches do not devise any module specifically for temporal reasoning. To better study temporal fact reasoning over HKGs, we propose a new type of data structure named hyper-relational TKG (HTKG). Every fact in an HTKG is coupled with a timestamp explicitly indicating its time validity. We develop two new benchmark HTKG datasets, i.e., Wiki-hy and YAGO-hy, and propose an HTKG reasoning model that efficiently models hyper-relational temporal facts. To support future research on this topic, we open-source our datasets and model.

Figures (4)

• Figure 1: Examples of HKG (A) and HTKG (B) facts. Contents inside dashed line squares denote qualifiers. We also provide another example of HTKG fact showcasing diverse sets of qualifiers in App. \ref{['app: more htkg example']}.
• Figure 2: Model structure of HypeTKG$^\psi$. HypeTKG$^\psi$ first uses QATGE to encode all the entities. It then uses QMD to compute score regarding every candidate entity $e_c \in \mathcal{E}$. Temporal information is considered in both QATGE and QMD for temporal reasoning. The structure of HypeTKG can be derived by excluding the components concerning TI facts. View with Sec. \ref{['sec: model']} for better understanding. $e"_1, ..., e"_{n_\text{TI}^s}$ and $r"_1, ..., r"_{n_\text{TI}^s}$ are the entities and relations from a number of $n_\text{TI}^s$ TI neighbors of query subject $s$, respectively.
• Figure 3: HypeTKG performance with a varying ratio of used qualifiers.
• Figure 4: Additional example of HTKG fact.

Theorems & Definitions (2)

• Definition 1: Hyper-Relational TKG
• Definition 2: Hyper-Relational TKG LP