Uniting Multi-Scale Local Feature Awareness and the Self-Attention Mechanism for Named Entity Recognition

In recent years, a huge amount of text information requires processing to support the diagnosis and treatment diabetes in medical field; therefore, named entity recognition (DNER) is giving rise popularity this research topic within particular field. Although mainstream methods for Chinese can effectively capture global context information, they ignore potential local sentences, hence cannot extract features through an efficient framework. To overcome these challenges, paper constructs corpus proposes RMBC (RoBERTa Multi-scale CNN BiGRU Self-attention CRF) model. This model that unites multi-scale feature awareness self-attention mechanism. first utilizes RoBERTa-wwm encode characters; then, it designs context-wise module, which captures containing locally important by fusing multi-window attention with residual convolution at adds mechanism address restriction bidirectional gated recurrent unit (BiGRU) capturing long-distance dependencies obtain semantic information. Finally, conditional random fields (CRF) are relied on learn dependency between adjacent tags optimal tag sequence. The experimental results our constructed private dataset, termed DNER, along two benchmark datasets, demonstrate effectiveness paper.