“…More recently, increasingly many neural network‐based approaches have been shown to be effective for the KGQA task as well. These approaches employ neural architectures ranging from simple neural embedding‐based models (Bordes, Chopra, & Weston, 2014), over attention‐based recurrent models (Cheng, Reddy, Saraswat, & Lapata, 2019), to memory‐augmented neural controller architectures (C. Liang, Berant, Le, Forbus, & Lao, 2017; Neelakantan, Le, Abadi, McCallum, & Amodei, 2017; F. Yang, Z. Yang, & Cohen, 2017).…”