基于细粒度词表示的命名实体识别研究

Named Entity Identification Based on Fine-Grained Word Representation

命名实体识别(NER)是自然语言处理中的一项基础任务,其性能的优劣极大地影响着关系抽取、语义角色标注等后续任务。传统的统计模型特征设计难度大、领域适应性差,一些神经网络模型则忽略了词本身所具有的形态学信息。针对上述问题,该文构建了一种基于细粒度词表示的端到端模型(Finger-BiLSTM-CRF)来进行命名实体识别任务。该文首先提出一种基于注意力机制的字符级词表示模型Finger来融合形态学信息和单词的字符信息,然后将Finger与BiLSTM-CRF模型联合进行实体识别,最终该方法以端到端、无任何特征工程的方式在CoNLL 2003数据集上取得了F1为91.09%的结果。实验表明,该文设计的Finger模型显著提升NER系统的召回率,从而使得模型的识别能力显著提升。

Named entity recognition is a fundamental task in NLP,with a substantial influence to relation extraction,sematic roll annotation etc.This paper proposes a new end-to-end neural network model(Finger-BiLSTM-CRF)based on a fine-grained word representation for named entity recognition task.First,we design Finger,a characterlevel word representation model based on the attention mechanism,for the integration of morphological information with information from each character of current token.Secondly,we combine Finger with BiLSTM-CRF for named entity recognition task.Finally,the model trained in an end-to-end process achieves a F1 score of 91.09% on test dataset from CoNLL 2003.