Question Answering over Knowledge Graphs with Neural Machine Translation and Entity Linking

The goal of Question Answering over Knowledge Graphs (KGQA) is to find for natural language questions over a knowledge graph. Recent KGQA adopt a neural machine translation (NMT) approach, where the natural question is translated into a structured query language. However, NMT from the out-of-vocabulary problem, where terms in a question may not been seen during training, impeding their translation. This issue is problematic for the millions of entities that large knowledge describe. We rather propose a KGQA approach that delegates the of entities to entity linking (EL) systems. NMT is then used to a query template with placeholders that are filled by entities in an EL phase. Slot filling is used to decide which entity fills placeholder. Experiments for QA over Wikidata show that our approach pure NMT: while there remains a strong dependence on having seen query templates during training, errors relating to entities are reduced.