基于深度学习的渔业领域命名实体识别

为了解决基于分词的渔业领域命名实体识别效果受分词准确度影响这一问题,采用一种基于深度学习的渔业领域命名实体识别方法。该方法使用神经网络训练得到字向量作为模型输入,避免了分词不准确对渔业领域命名实体识别效果造成的影响;针对渔业领域命名实体长度较长这一特点,使用LSTM单元保持较长时间记忆信息,并将标记信息融入到CRF模型中构建Character+LSTM+CRF实体识别模型。为验证方法的有效性,在渔业领域语料集上进行多组实验,结果表明,本研究中提出的Character+LSTM+CRF方法具有较好的效果,与LSTM模型相比较,在准确率、召回率、F值上分别提升了3.39%、2.99%、3.19%,对于渔业领域实体识别具有较好的效果。

Recognition of nominated fishery domain entity based on deep learning architectures

A deep learning based fishery domain entity recognition model was proposed to deal with the problem in entity recognition in fishery domain caused by accuracy of participle. The neural networks was used to learn the character embedding in order to avoid the influence of the inaccuracy participle on fishery domain entity recogni-tion,and the LSTM model was used to keep along memory information based on the long fishery domain entity ter-minology. The context labeling information was incorporated into the CRF model to construct the entity recognition model. Some experiments of entity recognition in fishery domain by the model indicated that the Character+LSTM+CRF model proposed here had good effect on the entity recognition of fishery domain,with increase by 3.39% in ac-curacy,by 2.99% in recall,and by 3.19% in F-score on fishery dictionary and national and local standard docu-ments in the field of fisheries compared to the LSTM model.