PredictQA: A Novel Approach to Predictive Temporal Knowledge Graph Question Answering

Recently, Temporal Knowledge Graph Question Answering (TKGQA) has been introduced and studied to reason about dynamic factual knowledge. To promote research in TKGQA, several datasets (eg., CRONQUESTIONS and MULTITQ) have been constructed, and various models have been proposed based on these datasets. However, current research often overlooks the impact of past facts on the accuracy of answers to predictive questions. To overcome this limitation, this study proposes PredictQA, a TKGQA method based on the T<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>NTComplEx embedding model. The T<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>NTComplEx model integrates multiple embeddings of entities, relations, and time, effectively capturing the dynamic relationships in the knowledge graph over time. Experimental results show that the TKGQA system based on the T<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>NTComplEx model has significant advantages in addressing predictive questions. Our approach not only improves the accuracy and robustness of the question answering(QA) system but also provides new insights and tools for future temporal knowledge graph research.