RoBERTa Word Embedding Based Power Grid Dispatching Entity Recognition
Xiao Li, Wenteng Liang, Yifeng Li, Yuxuan Zhao, Zhizheng Zhang, Hengguang Yang · 2020
The recognition of entities is the basis of knowledge graph construction. Deep neural network models is currently the most effective and efficient way of entity recognition. Among them, a pop method is to apply the long short-term memory networks to model the reliance on sequences and use Conditional Random Field to model the dependencies of sequence outputs. However, due to the dense distribution of entities in current power grid corpus, problems such as low recognition accuracy and incorrect division of solid boundaries gradually arose in previous models. To address the issues, this paper propose an approach where the power grid corpus data is preprocessed based on the idea of RoBERTa and then train the word embedding model. The experimental results show that the results of word embedding model have practical potential