Entity Linking Based on ERNIE for Chinese Knowledge Graph Question Answer
Yuan Zong, Xiaoze Gong, Yongli Wang · 2022
The existing entity link models have low ability to understand the context of short questions, and there are problems such as entity boundary recognition errors and entity semantic understanding errors, and there are few models that can handle Chinese. To solve the above problems, we propose a problem entity recognition method based on the ERNIE-BiLSTM-CRF model, which divides the entity link task into two parts: problem entity recognition and entity disambiguation. 2) The ERNIE model is used to calculate the similarity between the problem entity context information and the candidate entities, and the entity popularity is combined as the feature of entity disambiguation. Experiments demonstrate the method has a good entity linking effect for Chinese knowledge map question answering.