针对传统方法在石油钻井领域由于检索词不标准、语义模糊导致检索结果偏差较大的问题,提出一种基于BERT(Bidirectional Encoder Representation from Transformers)孪生网络模型的注意力池化方法以提高文献相似度评估的准确率。首先使...针对传统方法在石油钻井领域由于检索词不标准、语义模糊导致检索结果偏差较大的问题,提出一种基于BERT(Bidirectional Encoder Representation from Transformers)孪生网络模型的注意力池化方法以提高文献相似度评估的准确率。首先使用爬虫技术采集石油钻井文献并清洗整理,然后利用5类石油钻井文献数据集评估指标对样本进行打分标注,最后结合钻井文献数据集特征,提出基于孪生BERT网络的注意力池化方法,对多特征样本进行整体语义表达。实验结果表明,相较于常规的池化方法,该模型能提升石油钻井文献相似度度量的效果,并具有一定的泛化性能。展开更多
探讨了PubMed,EMBASE.COM,Web of Science,中国知网,万方和维普6个文献数据库相关文献功能及其实现机制,对比分析了国内外文献数据库在相关文献提供方面的特点。结果显示,国外文献数据库检索系统的相关文献功能比较单一,具有较强的针对...探讨了PubMed,EMBASE.COM,Web of Science,中国知网,万方和维普6个文献数据库相关文献功能及其实现机制,对比分析了国内外文献数据库在相关文献提供方面的特点。结果显示,国外文献数据库检索系统的相关文献功能比较单一,具有较强的针对性和实用性;而国内文献数据库的相关检索功能相对比较完善,不仅能提供基于外部特征的相关文献,还能提供相关(关键词或分类号)文献的简单内容。在相关文献的实用性方面,中国知网优于万方数据库,万方数据库优于维普数据库。展开更多
科技文献检索时以专业角度给出合理的相关度排序是一项非常重要工作,传统PageRank算法采用了平均分配相似性权重的方式,但其会产生文献排序结果不合理的问题。为此,提出一种将深度学习方法与PageRank相结合的算法,提高文献相关度排序的...科技文献检索时以专业角度给出合理的相关度排序是一项非常重要工作,传统PageRank算法采用了平均分配相似性权重的方式,但其会产生文献排序结果不合理的问题。为此,提出一种将深度学习方法与PageRank相结合的算法,提高文献相关度排序的可靠性。首先,使用具有注意力池化的孪生BERT(Bidirectional Encoder Representation from Transformers)深度学习网络计算文献与引文的相似度;然后,对文献与其所包含引文间的相似度进行规范化处理;最后,将标准化后的相似度作为分配权重对引文网络计算排序。实验结果表明,相较于传统的PageRank算法,该方法检索结果的相关度提升6%以上,因此更适合应用于科技文献的引文网络分析。展开更多
Heathcliff and Chillingworth are the two typical satan type heroes among literature works. To some extent, their revenge is similar, but there are also differences. However, there are few comparisons between the two h...Heathcliff and Chillingworth are the two typical satan type heroes among literature works. To some extent, their revenge is similar, but there are also differences. However, there are few comparisons between the two heroes from the perspective of romanticism. From the analysis of the author＇s living background and their literary genre of the two works, the paper will analyse the reasons why Heathcliff and Chillingworth want to revenge, and reflect their similarities and differences in terms of their character, means, ending, and so on. The paper will analyse the two heroes by using the methods of literature data collection, text comparison and induction, and come to a conclusion that the romanticism carried by the two authors is influential in shaping the two satan type heroes. Thus to gain more details about the similarities of the two cruel revenger, and the differences of their individuality and different ways of revenge.展开更多
文摘针对传统方法在石油钻井领域由于检索词不标准、语义模糊导致检索结果偏差较大的问题,提出一种基于BERT(Bidirectional Encoder Representation from Transformers)孪生网络模型的注意力池化方法以提高文献相似度评估的准确率。首先使用爬虫技术采集石油钻井文献并清洗整理,然后利用5类石油钻井文献数据集评估指标对样本进行打分标注,最后结合钻井文献数据集特征,提出基于孪生BERT网络的注意力池化方法,对多特征样本进行整体语义表达。实验结果表明,相较于常规的池化方法,该模型能提升石油钻井文献相似度度量的效果,并具有一定的泛化性能。
文摘探讨了PubMed,EMBASE.COM,Web of Science,中国知网,万方和维普6个文献数据库相关文献功能及其实现机制,对比分析了国内外文献数据库在相关文献提供方面的特点。结果显示,国外文献数据库检索系统的相关文献功能比较单一,具有较强的针对性和实用性;而国内文献数据库的相关检索功能相对比较完善,不仅能提供基于外部特征的相关文献,还能提供相关(关键词或分类号)文献的简单内容。在相关文献的实用性方面,中国知网优于万方数据库,万方数据库优于维普数据库。
文摘科技文献检索时以专业角度给出合理的相关度排序是一项非常重要工作,传统PageRank算法采用了平均分配相似性权重的方式,但其会产生文献排序结果不合理的问题。为此,提出一种将深度学习方法与PageRank相结合的算法,提高文献相关度排序的可靠性。首先,使用具有注意力池化的孪生BERT(Bidirectional Encoder Representation from Transformers)深度学习网络计算文献与引文的相似度;然后,对文献与其所包含引文间的相似度进行规范化处理;最后,将标准化后的相似度作为分配权重对引文网络计算排序。实验结果表明,相较于传统的PageRank算法,该方法检索结果的相关度提升6%以上,因此更适合应用于科技文献的引文网络分析。
文摘Heathcliff and Chillingworth are the two typical satan type heroes among literature works. To some extent, their revenge is similar, but there are also differences. However, there are few comparisons between the two heroes from the perspective of romanticism. From the analysis of the author＇s living background and their literary genre of the two works, the paper will analyse the reasons why Heathcliff and Chillingworth want to revenge, and reflect their similarities and differences in terms of their character, means, ending, and so on. The paper will analyse the two heroes by using the methods of literature data collection, text comparison and induction, and come to a conclusion that the romanticism carried by the two authors is influential in shaping the two satan type heroes. Thus to gain more details about the similarities of the two cruel revenger, and the differences of their individuality and different ways of revenge.
