关键词: Named entity recognition, Chinese medical text, Lexicon adapter, Self-attention mechanism, Bidirectional Long Short-Term Memory (BiLSTM)

Abstract: Abstract: To address the difficulty of word boundary recognition stemming from nested entities in Chinese medical texts, as well as significant semantic information loss in existing Lattice-LSTM structures with integrated lexical features, an adaptive lexical information enhancement model for Chinese medical named entity recognition is proposed. The proposed model utilizes BiLSTM networks to encode contextual information of character sequences and capture long-distance dependencies. Potential word information of each character is modeled using the Char-words pair sequence , and the self-attention mechanism is utilized to realize internal interactions between different words. Finally, a Lexicon adapter, based on the bilinear-attention mechanism, integrates lexical information into each character in the text sequence. This enhances semantic information effectively while fully utilizing the rich boundary information of words and suppressing words with low correlation. Experimental results demonstrate that the average F1 value of the proposed model increases by 1.94% to 2.38% compared to the character-based baseline model, and its performance is further optimized when combined with BERT.

Key words: 命名实体识别, 中文医学文本, 词汇适配器, 自注意力机制, 双向长短期记忆网络