Qualitative algebraic equations are the basis of qualitative simulation,which are used to express the dynamic behavior of steady-state continuous processes.When the values and operation of qualitative variables are re...Qualitative algebraic equations are the basis of qualitative simulation,which are used to express the dynamic behavior of steady-state continuous processes.When the values and operation of qualitative variables are redefined,qualitative algebraic equations can be transformed into signed direct graphs,which are frequently used to predict the trend of dynamic changes.However,it is difficult to use traditional qualitative algebra methods based on artificial trial and error to solve a complex problem for dynamic trends.An important aspect of modern qualitative algebra is to model and characterize complex systems with the corresponding computer-aided automatic reasoning.In this study,a qualitative affection equation based on multiple conditions is proposed,which enables the signed di-rect graphs to describe complex systems better and improves the fault diagnosis resolution.The application to an industrial case shows that the method performs well.展开更多
为有效解决构建电力运检知识图谱的关键步骤之一的电力运检命名实体识别问题,通过构建一种基于Stacking多模型融合的隐马尔可夫-条件随机场-双向长短期记忆网络(hidden Markov-conditional random fields-bi-directional long short-ter...为有效解决构建电力运检知识图谱的关键步骤之一的电力运检命名实体识别问题,通过构建一种基于Stacking多模型融合的隐马尔可夫-条件随机场-双向长短期记忆网络(hidden Markov-conditional random fields-bi-directional long short-term,HCB)模型方法研究了电力运检命名实体识别问题。HCB模型分为两层,第一层使用隐马尔可夫模型(hidden Markov model,HMM)、条件随机场(conditional random fields,CRF)和双向长短期记忆网络(bi-directional long short-term memory,Bi-LSTM)模型进行训练预测,再将预测结果输入第二层的CRF模型进行训练,经过双层模型训练预测得出最后的命名实体。结果表明:在电力运检命名实体识别问题上HCB模型的精确率、召回率及F1值等指标明显优于单模型以及其他的融合模型。可见HCB模型能有效解决电力运检命名实体识别问题。展开更多
【目的】针对当前茶叶领域语料数据库不完善、多源异构数据聚合能力差、知识共享困难等问题,提出一种基于BERT-WWM-BiLSTM-AttTea-CRF模型的茶叶知识图谱构建方法。【方法】以基于全词掩码的BERT-WWM(Whole Word Masking)层替换预训练...【目的】针对当前茶叶领域语料数据库不完善、多源异构数据聚合能力差、知识共享困难等问题,提出一种基于BERT-WWM-BiLSTM-AttTea-CRF模型的茶叶知识图谱构建方法。【方法】以基于全词掩码的BERT-WWM(Whole Word Masking)层替换预训练模型中的随机掩码BERT层,并根据茶叶领域语料数据的全局文本特征,设计可实现茶叶关键实体权重分配的注意力机制层以提高文本提取的准确率,最后通过条件随机场对序列中的各个实体进行分类提取,从而完成茶叶中文实体识别的整个流程。【结果】BERT-WWM-BiLSTM-AttTea-CRF模型能有效识别茶叶知识文本数据中的实体,对茶叶非结构化数据的实体抽取效果优于RoBERTa_BiLSTM_CRF、ALBERT_BiLSTM_CRF等主流模型,识别的准确率、召回率、F1值分别为92.03%、90.36%、91.19%。经改进后的模型对茶叶品种数据和茶叶病害数据的识别率有明显提升,其F1值分别达到94.32%和94.05%。【结论】本研究所构建的茶叶知识图谱具有数据覆盖面广、聚合能力强、体系完整等优势,对农业特定领域的知识图谱构建和农业中文命名实体的提取研究具有重要意义。展开更多
关系抽取是自然语言处理中一项基础的上游任务.句子的结构信息在某种意义上蕴含了实体及其关系信息,有助于提高关系抽取的准确率,然而使用现有自然语言处理(Natural Language Processing,NLP)语言工具进行句法分析时会引入一定的错误传...关系抽取是自然语言处理中一项基础的上游任务.句子的结构信息在某种意义上蕴含了实体及其关系信息,有助于提高关系抽取的准确率,然而使用现有自然语言处理(Natural Language Processing,NLP)语言工具进行句法分析时会引入一定的错误传播问题,且现有的基于图结构的关系抽取模型在一定程度上忽略了句子的时序信息.通过结合双向长短时记忆网络(Bi-directional Long Short-Term Memory,Bi LSTM)捕获句子序列的上下文关系,同时使用传统条件随机场(Conditional Random Field,CRF)的关系标注结果矫正NLP工具的错误传播问题,提出了一种用于关系抽取的双层时空图卷积神经网络(Bilayer Spatiotemporal Graph Convolution Neural Network,Bi SpGCN)模型.该模型在中文糖尿病数据集和中文人物关系数据集上的实验结果表明,相较于传统的多头注意力引导的图卷积神经网络(Attention Guided Graph Convolutional Networks for Relation Extraction,AGGCN)模型,BiSpGCN模型能够充分利用句子的有效信息,具有更好的关系抽取性能.展开更多
基金Supported by the National High Technology Research and Development Program of China(2009AA04Z133)
文摘Qualitative algebraic equations are the basis of qualitative simulation,which are used to express the dynamic behavior of steady-state continuous processes.When the values and operation of qualitative variables are redefined,qualitative algebraic equations can be transformed into signed direct graphs,which are frequently used to predict the trend of dynamic changes.However,it is difficult to use traditional qualitative algebra methods based on artificial trial and error to solve a complex problem for dynamic trends.An important aspect of modern qualitative algebra is to model and characterize complex systems with the corresponding computer-aided automatic reasoning.In this study,a qualitative affection equation based on multiple conditions is proposed,which enables the signed di-rect graphs to describe complex systems better and improves the fault diagnosis resolution.The application to an industrial case shows that the method performs well.
文摘为有效解决构建电力运检知识图谱的关键步骤之一的电力运检命名实体识别问题,通过构建一种基于Stacking多模型融合的隐马尔可夫-条件随机场-双向长短期记忆网络(hidden Markov-conditional random fields-bi-directional long short-term,HCB)模型方法研究了电力运检命名实体识别问题。HCB模型分为两层,第一层使用隐马尔可夫模型(hidden Markov model,HMM)、条件随机场(conditional random fields,CRF)和双向长短期记忆网络(bi-directional long short-term memory,Bi-LSTM)模型进行训练预测,再将预测结果输入第二层的CRF模型进行训练,经过双层模型训练预测得出最后的命名实体。结果表明:在电力运检命名实体识别问题上HCB模型的精确率、召回率及F1值等指标明显优于单模型以及其他的融合模型。可见HCB模型能有效解决电力运检命名实体识别问题。
文摘【目的】针对当前茶叶领域语料数据库不完善、多源异构数据聚合能力差、知识共享困难等问题,提出一种基于BERT-WWM-BiLSTM-AttTea-CRF模型的茶叶知识图谱构建方法。【方法】以基于全词掩码的BERT-WWM(Whole Word Masking)层替换预训练模型中的随机掩码BERT层,并根据茶叶领域语料数据的全局文本特征,设计可实现茶叶关键实体权重分配的注意力机制层以提高文本提取的准确率,最后通过条件随机场对序列中的各个实体进行分类提取,从而完成茶叶中文实体识别的整个流程。【结果】BERT-WWM-BiLSTM-AttTea-CRF模型能有效识别茶叶知识文本数据中的实体,对茶叶非结构化数据的实体抽取效果优于RoBERTa_BiLSTM_CRF、ALBERT_BiLSTM_CRF等主流模型,识别的准确率、召回率、F1值分别为92.03%、90.36%、91.19%。经改进后的模型对茶叶品种数据和茶叶病害数据的识别率有明显提升,其F1值分别达到94.32%和94.05%。【结论】本研究所构建的茶叶知识图谱具有数据覆盖面广、聚合能力强、体系完整等优势,对农业特定领域的知识图谱构建和农业中文命名实体的提取研究具有重要意义。
