互联网上的攻击性言论严重扰乱了正常网络秩序,破坏了健康交流的网络环境。现有的检测技术更关注文本中的鲜明特征,难以发现更隐晦的攻击方式。针对上述问题,提出融合反讽机制的攻击性言论检测模型BSWD(Bidirectional Encoder Represent...互联网上的攻击性言论严重扰乱了正常网络秩序,破坏了健康交流的网络环境。现有的检测技术更关注文本中的鲜明特征,难以发现更隐晦的攻击方式。针对上述问题,提出融合反讽机制的攻击性言论检测模型BSWD(Bidirectional Encoder Representation from Transformers-based Sarcasm and Word Detection)。首先,提出基于反讽机制的模型Sarcasm-BERT,以检测言论中的语义冲突;其次,提出细粒度词汇攻击性特征提取模型WordsDetect,检测言论中的攻击性词汇;最后,融合两种模型得到BSWD。实验结果表明,与BERT(Bidirectional Encoder Representation from Transformers)、HateBERT模型相比,所提模型的准确率、精确率、召回率和F1分数指标大部分能提升2%,显著提高了检测性能,更能发现隐含的攻击性言论;同时,与SKS(Sentiment Knowledge Sharing)、BiCHAT(Bidirectional long shortterm memory with deep Convolution neural network and Hierarchical ATtention)模型相比,具有更强的泛化能力和鲁棒性。以上结果验证了BSWD检测隐晦攻击性言论的有效性。展开更多
Soil DNA extraction,such as microbial community analysis and gene drift detection,is an important basis for multiple analyses in different fields.Nevertheless,the soil DNA extraction methods for field detection are st...Soil DNA extraction,such as microbial community analysis and gene drift detection,is an important basis for multiple analyses in different fields.Nevertheless,the soil DNA extraction methods for field detection are still lacking.This study established a rapid soil DNA extraction(RSDE)method that can be used in field detection.In this method,we first utilized the optimized lysate to isolate DNA from soil and then used a filtration membrane and a DNA adsorption membrane to purify the DNA via the column method.Moreover,we used the pressure from the syringe instead of the conventional centrifugal force of the centrifuge to assist the sample filtration,resulting in very low requirements for this method,with an extraction time of less than 20 min.Furthermore,we demonstrated that the RSDE method was applicable for DNA extraction from different types of soils,with the demand for soil samples as low as 0.1 g and that the amount of obtained DNA was,to some extent,greater than that obtained by a commercial kit.Further analysis revealed that this extracted genomic DNA can be used directly for polymerase chain reaction(PCR)analysis,including ordinary PCR,real-time fluorescent quantitative PCR,and recombinase polymerase amplification(RPA)-CRISPR/Cas12a visual assays.In addition,we demonstrated that this method can be used to extract DNA from residual plant roots in addition to soil microbes,which lays a foundation for the comprehensive analysis of soil plants and microorganisms.In summary,the RSDE method proposed in this study may have wide application prospects.展开更多
This work introduces a novel method for measuring thin film thickness,employing a multi-wavelength method that significantly reduces the need for broad-spectrum data.Unlike traditional techniques that require sev⁃eral...This work introduces a novel method for measuring thin film thickness,employing a multi-wavelength method that significantly reduces the need for broad-spectrum data.Unlike traditional techniques that require sev⁃eral hundred spectral data points,the multi-wavelength method achieves precise thickness measurements with data from only 10 wavelengths.This innovation not only simplifies the process of spectral measurement analysis but also enables accurate real-time thickness measurement on industrial coating production lines.The method effectively reconstructs and fits the visible spectrum(400-800 nm)using a minimal amount of data,while maintaining measurement error within 7.1%.This advancement lays the foundation for more practical and efficient thin film thickness determination techniques in various industrial applications.展开更多
文摘互联网上的攻击性言论严重扰乱了正常网络秩序,破坏了健康交流的网络环境。现有的检测技术更关注文本中的鲜明特征,难以发现更隐晦的攻击方式。针对上述问题,提出融合反讽机制的攻击性言论检测模型BSWD(Bidirectional Encoder Representation from Transformers-based Sarcasm and Word Detection)。首先,提出基于反讽机制的模型Sarcasm-BERT,以检测言论中的语义冲突;其次,提出细粒度词汇攻击性特征提取模型WordsDetect,检测言论中的攻击性词汇;最后,融合两种模型得到BSWD。实验结果表明,与BERT(Bidirectional Encoder Representation from Transformers)、HateBERT模型相比,所提模型的准确率、精确率、召回率和F1分数指标大部分能提升2%,显著提高了检测性能,更能发现隐含的攻击性言论;同时,与SKS(Sentiment Knowledge Sharing)、BiCHAT(Bidirectional long shortterm memory with deep Convolution neural network and Hierarchical ATtention)模型相比,具有更强的泛化能力和鲁棒性。以上结果验证了BSWD检测隐晦攻击性言论的有效性。
基金the Experimental Technology Research Project of Zhejiang University(SYB202138)National Natural Science Foundation of China(32000195).
文摘Soil DNA extraction,such as microbial community analysis and gene drift detection,is an important basis for multiple analyses in different fields.Nevertheless,the soil DNA extraction methods for field detection are still lacking.This study established a rapid soil DNA extraction(RSDE)method that can be used in field detection.In this method,we first utilized the optimized lysate to isolate DNA from soil and then used a filtration membrane and a DNA adsorption membrane to purify the DNA via the column method.Moreover,we used the pressure from the syringe instead of the conventional centrifugal force of the centrifuge to assist the sample filtration,resulting in very low requirements for this method,with an extraction time of less than 20 min.Furthermore,we demonstrated that the RSDE method was applicable for DNA extraction from different types of soils,with the demand for soil samples as low as 0.1 g and that the amount of obtained DNA was,to some extent,greater than that obtained by a commercial kit.Further analysis revealed that this extracted genomic DNA can be used directly for polymerase chain reaction(PCR)analysis,including ordinary PCR,real-time fluorescent quantitative PCR,and recombinase polymerase amplification(RPA)-CRISPR/Cas12a visual assays.In addition,we demonstrated that this method can be used to extract DNA from residual plant roots in addition to soil microbes,which lays a foundation for the comprehensive analysis of soil plants and microorganisms.In summary,the RSDE method proposed in this study may have wide application prospects.
基金Supported by National Key R&D Program of China(2021YFA0715500)National Natural Science Foundation of China(NSFC)(12227901)+2 种基金Strategic Priority Research Program(B)of the Chinese Academy of Sciences(XDB0580000)Shanghai Municipal Science and Technology Major Project(2019SHZDZX01)Chinese Academy of Sciences President's International Fellowship Initiative(2021PT0007).
文摘This work introduces a novel method for measuring thin film thickness,employing a multi-wavelength method that significantly reduces the need for broad-spectrum data.Unlike traditional techniques that require sev⁃eral hundred spectral data points,the multi-wavelength method achieves precise thickness measurements with data from only 10 wavelengths.This innovation not only simplifies the process of spectral measurement analysis but also enables accurate real-time thickness measurement on industrial coating production lines.The method effectively reconstructs and fits the visible spectrum(400-800 nm)using a minimal amount of data,while maintaining measurement error within 7.1%.This advancement lays the foundation for more practical and efficient thin film thickness determination techniques in various industrial applications.
