Self-attention-based Question Answering over Knowledge Graphs

This paper proposes a model for Simple Question Answering over Knowledge Graphs. Our model includes a Question Named Entity Recognition (NER) Model and an Attribute Classification Model. First, we utilize the BERT model, BiLSTM, and CNN in the Question NER Model. The BERT model is used to represents the question. The BilSTM and CNN are employed to extract question features. Second, we introduce the answer information and self-attention mechanism in the Attribute Classification Model. The self-attention mechanism is used to assign the weight of the attribute and the answer automatically. The experimental results show that our approach can obtain competitive results compared with state-of-the-art approaches.