ForecastTKGQuestions: A Benchmark for Temporal Question Answering and Forecasting over Temporal Knowledge Graphs

Zifeng Ding, Zongyue Li, Ruoxia Qi, Jingpei Wu, Bailan He, Yunpu Ma, Zhao Meng, Shuo Chen, Ruotong Liao, Zhen Han, Volker Tresp

Abstract: Question answering over temporal knowledge graphs (TKGQA) has recently found increasing interest. Previous related works aim to develop QA systems that answer temporal questions based on the facts from a fixed time period, where a temporal knowledge graph (TKG) spanning this period can be fully used for inference. In real-world scenarios, however, it is common that given knowledge until the current instance, we wish the TKGQA systems to answer the questions asking about future. As humans constantly plan the future, building forecasting TKGQA systems is important. In this paper, we propose a novel task: forecasting TKGQA, and propose a coupled large-scale TKGQA benchmark dataset, i.e., ForecastTKGQuestions. It includes three types of forecasting questions, i.e., entity prediction, yes-unknown, and fact reasoning questions. For every question, a timestamp is annotated and QA models only have access to TKG information prior to it for answer inference. We find that previous TKGQA methods perform poorly on forecasting questions, and they are unable to answer yes-unknown and fact reasoning questions. To this end, we propose ForecastTKGQA, a TKGQA model that employs a TKG forecasting module for future inference. Experiments show that it performs well in forecasting TKGQA.