Some influential factors of anther culture were studied preliminarily by conducting anther culture of the restorers of new cytoplasmic male sterile （NER）. Several results were obtain from this experiment and they we...Some influential factors of anther culture were studied preliminarily by conducting anther culture of the restorers of new cytoplasmic male sterile （NER）. Several results were obtain from this experiment and they were listed as follow：① MS cultrure medium with such hormones as 2,4-D 2 mg/L,6-BA 0.5 mg/L, NAA 0.5 mg/L was the best suitable for callus induction of NER. ②The difference of induction rate was significantly different between different plant age groups. From the 110th day to 141th day,the induction rate was increased with the increase of age and the difference of induction rate reached 0.01 significant difference level. The induction rate reached the highest value in the 141th day then it declined gradually. ③The combined use of 2, 4-D and 6-BA with proper increase of 2,4-D was good for inducing callus. ④The green plantlet induction rate of NER was increased when the concentration of 6-BA increased from 2 mg/L to 4 mg/L. Adding ZT from 0.5 mg/L to 2 mg/L. 6-BA would led 2.47% increase of green plantlet olantlet induction rate.展开更多
由于社交媒体平台上所发布的非结构化信息存在数据不一致、重要程度不同等问题,使自动准确抽取所需信息并标注受灾级别成为一个有挑战性的工作。因此,结合形式概念分析(FCA)、词共现关系和上下文语义信息构建了水灾事件知识体系。利用...由于社交媒体平台上所发布的非结构化信息存在数据不一致、重要程度不同等问题,使自动准确抽取所需信息并标注受灾级别成为一个有挑战性的工作。因此,结合形式概念分析(FCA)、词共现关系和上下文语义信息构建了水灾事件知识体系。利用所构建的知识体系,基于TencentPretrain框架对大规模语言预训练模型(LLM)进行指令微调,构建了ChatFlowFlood信息抽取模型,可以在少量人工标记情况下,准确自动抽取被困情况、紧缺物资等信息;在信息抽取模型的基础上,通过模糊层次分析法(FAHP)和CRITIC法(CRiteria Importance Through Intercriteria Correlation)主客观结合评定求助信息的救援优先级,帮助决策者理解灾情紧急程度。实验结果表明,在中文社交媒体数据上,与ChatFlow-7B模型相比,ChatFlowFlood模型的FBERT指标提升了73.09%。展开更多
[Objective]The aim was to establish optimized system for NER isolated microspore culture of Brassica napus L.. [Method]Twenty varieties of NER of Brassica napus were grown under uncontrolled temperature and light cond...[Objective]The aim was to establish optimized system for NER isolated microspore culture of Brassica napus L.. [Method]Twenty varieties of NER of Brassica napus were grown under uncontrolled temperature and light conditions,and their isolated microspore from anthers were used as explants in vitro culture. The influencing factors of microspore culture were preliminarily studied. [Result]The difference of genotypes was important influencing factors to embryoid yield. The embryoid yield increased by supplementing with 6-BA and NAA,culturing in solid-liquid double layer medium with activated charcoal; The difference was not significant of embryoid yield between culturing in medium supplemented with colchicines and the CK. The rates of cotyledonous embryoids directly developed into normal plantlets increased through enriching with 0.1-0.2 mg/L NAA and being treated on slim illumination two days before being inducted into normal plantlets. [Conclusion]The technical system of microspore culture of restorer of new cytoplasmic male sterile （NER） was established,which and lays a foundation for accelerating genotype purification of NER introgressive line.展开更多
针对当前大多数命名实体识别(NER)模型只使用字符级信息编码且缺乏对文本层次信息提取的问题,提出一种融合多粒度语言知识与层级信息的中文NER(CNER)模型(CMH)。首先,使用经过多粒度语言知识预训练的模型编码文本,使模型能够同时捕获文...针对当前大多数命名实体识别(NER)模型只使用字符级信息编码且缺乏对文本层次信息提取的问题,提出一种融合多粒度语言知识与层级信息的中文NER(CNER)模型(CMH)。首先,使用经过多粒度语言知识预训练的模型编码文本,使模型能够同时捕获文本的细粒度和粗粒度语言信息,从而更好地表征语料;其次,使用ON-LSTM(Ordered Neurons Long Short-Term Memory network)模型提取层级信息,利用文本本身的层级结构信息增强编码间的时序关系;最后,在模型的解码端结合文本的分词信息,并将实体识别问题转化为表格填充问题,以更好地解决实体重叠问题并获得更准确的实体识别结果。同时,为解决当前模型在不同领域中的迁移能力较差的问题,提出通用实体识别的理念,通过筛选多领域的通用实体类型,构建一套提升模型在多领域中的泛化能力的通用NER数据集MDNER(Multi-Domain NER dataset)。为验证所提模型的效果,在数据集Resume、Weibo、MSRA上进行实验,与MECT(Multi-metadata Embedding based Cross-Transformer)模型相比,F1值分别提高了0.94、4.95和1.58个百分点。为了验证所提模型在多领域中的实体识别效果,在MDNER上进行实验,F1值达到了95.29%。实验结果表明,多粒度语言知识预训练、文本层级结构信息提取和高效指针解码器对模型的性能提升至关重要。展开更多
针对命名实体识别(NER)任务中相关模型通常仅对字符及相关词汇进行建模,未充分利用汉字特有的字形结构信息和实体类型信息的问题,提出一种融合先验知识和字形特征的命名实体识别模型。首先,采用结合高斯注意力机制的Transformer对输入...针对命名实体识别(NER)任务中相关模型通常仅对字符及相关词汇进行建模,未充分利用汉字特有的字形结构信息和实体类型信息的问题,提出一种融合先验知识和字形特征的命名实体识别模型。首先,采用结合高斯注意力机制的Transformer对输入序列进行编码,并从中文维基百科中获取实体类型的中文释义,采用双向门控循环单元(BiGRU)编码实体类型信息作为先验知识,利用注意力机制将它与字符表示进行组合;其次,采用双向长短时记忆(BiLSTM)网络编码输入序列的远距离依赖关系,通过字形编码表获得繁体的仓颉码和简体的现代五笔码,采用卷积神经网络(CNN)提取字形特征表示,并根据不同权重组合繁体与简体字形特征,利用门控机制将它与经过BiLSTM编码后的字符表示进行组合;最后,使用条件随机场(CRF)解码,得到命名实体标注序列。在偏口语化的数据集Weibo、小型数据集Boson和大型数据集PeopleDaily上的实验结果表明,与基线模型MECT(Multi-metadata Embedding based Cross-Transformer)相比,所提模型的F1值别提高了2.47、1.20和0.98个百分点,验证了模型的有效性。展开更多
基金Supported by the National 863 Project of Tenth-five Year Plan(2001AA2411042004AA241104)+1 种基金Key Breeding Project of Sichuan Province and(200107001-16-01)Key Quality Project of Sichuan Province(200107001-1-7-4)~~
文摘Some influential factors of anther culture were studied preliminarily by conducting anther culture of the restorers of new cytoplasmic male sterile （NER）. Several results were obtain from this experiment and they were listed as follow：① MS cultrure medium with such hormones as 2,4-D 2 mg/L,6-BA 0.5 mg/L, NAA 0.5 mg/L was the best suitable for callus induction of NER. ②The difference of induction rate was significantly different between different plant age groups. From the 110th day to 141th day,the induction rate was increased with the increase of age and the difference of induction rate reached 0.01 significant difference level. The induction rate reached the highest value in the 141th day then it declined gradually. ③The combined use of 2, 4-D and 6-BA with proper increase of 2,4-D was good for inducing callus. ④The green plantlet induction rate of NER was increased when the concentration of 6-BA increased from 2 mg/L to 4 mg/L. Adding ZT from 0.5 mg/L to 2 mg/L. 6-BA would led 2.47% increase of green plantlet olantlet induction rate.
文摘由于社交媒体平台上所发布的非结构化信息存在数据不一致、重要程度不同等问题,使自动准确抽取所需信息并标注受灾级别成为一个有挑战性的工作。因此,结合形式概念分析(FCA)、词共现关系和上下文语义信息构建了水灾事件知识体系。利用所构建的知识体系,基于TencentPretrain框架对大规模语言预训练模型(LLM)进行指令微调,构建了ChatFlowFlood信息抽取模型,可以在少量人工标记情况下,准确自动抽取被困情况、紧缺物资等信息;在信息抽取模型的基础上,通过模糊层次分析法(FAHP)和CRITIC法(CRiteria Importance Through Intercriteria Correlation)主客观结合评定求助信息的救援优先级,帮助决策者理解灾情紧急程度。实验结果表明,在中文社交媒体数据上,与ChatFlow-7B模型相比,ChatFlowFlood模型的FBERT指标提升了73.09%。
基金Supported by the Key Breeding Project of Sichuan Province during National Eleventh-five Year Plan (2006YZGG25 )Key Quality Project of Sichuan Province (2006YZGG223)Foundation for Young Scientists of Sichuan Provincial Financial Breeding Project in 2008~~
文摘[Objective]The aim was to establish optimized system for NER isolated microspore culture of Brassica napus L.. [Method]Twenty varieties of NER of Brassica napus were grown under uncontrolled temperature and light conditions,and their isolated microspore from anthers were used as explants in vitro culture. The influencing factors of microspore culture were preliminarily studied. [Result]The difference of genotypes was important influencing factors to embryoid yield. The embryoid yield increased by supplementing with 6-BA and NAA,culturing in solid-liquid double layer medium with activated charcoal; The difference was not significant of embryoid yield between culturing in medium supplemented with colchicines and the CK. The rates of cotyledonous embryoids directly developed into normal plantlets increased through enriching with 0.1-0.2 mg/L NAA and being treated on slim illumination two days before being inducted into normal plantlets. [Conclusion]The technical system of microspore culture of restorer of new cytoplasmic male sterile （NER） was established,which and lays a foundation for accelerating genotype purification of NER introgressive line.
文摘针对中文网络安全领域缺乏公开数据集和有效的命名实体识别(Named Entity Recognition,NER)方法,提出一种融合汉字多源信息的网络安全NER方法。通过构建数据集中所有字符的偏旁和字频向量表,增强了中文字向量的特征表达能力,嵌入到改进的词汇融合模型中进行字向量与词向量的融合,输入到条件随机场(Conditional Random Fields,CRF)进行解码。实验结果表明,该方法在保持较快解码速度和占用较低计算机资源的情况下,在网络安全数据集上,其准确率、召回率和F1值分别为0.8649、0.8402和0.8523,均优于现有模型,能够为后续网络安全知识图谱的构建提供支撑。
文摘针对当前大多数命名实体识别(NER)模型只使用字符级信息编码且缺乏对文本层次信息提取的问题,提出一种融合多粒度语言知识与层级信息的中文NER(CNER)模型(CMH)。首先,使用经过多粒度语言知识预训练的模型编码文本,使模型能够同时捕获文本的细粒度和粗粒度语言信息,从而更好地表征语料;其次,使用ON-LSTM(Ordered Neurons Long Short-Term Memory network)模型提取层级信息,利用文本本身的层级结构信息增强编码间的时序关系;最后,在模型的解码端结合文本的分词信息,并将实体识别问题转化为表格填充问题,以更好地解决实体重叠问题并获得更准确的实体识别结果。同时,为解决当前模型在不同领域中的迁移能力较差的问题,提出通用实体识别的理念,通过筛选多领域的通用实体类型,构建一套提升模型在多领域中的泛化能力的通用NER数据集MDNER(Multi-Domain NER dataset)。为验证所提模型的效果,在数据集Resume、Weibo、MSRA上进行实验,与MECT(Multi-metadata Embedding based Cross-Transformer)模型相比,F1值分别提高了0.94、4.95和1.58个百分点。为了验证所提模型在多领域中的实体识别效果,在MDNER上进行实验,F1值达到了95.29%。实验结果表明,多粒度语言知识预训练、文本层级结构信息提取和高效指针解码器对模型的性能提升至关重要。
文摘针对命名实体识别(NER)任务中相关模型通常仅对字符及相关词汇进行建模,未充分利用汉字特有的字形结构信息和实体类型信息的问题,提出一种融合先验知识和字形特征的命名实体识别模型。首先,采用结合高斯注意力机制的Transformer对输入序列进行编码,并从中文维基百科中获取实体类型的中文释义,采用双向门控循环单元(BiGRU)编码实体类型信息作为先验知识,利用注意力机制将它与字符表示进行组合;其次,采用双向长短时记忆(BiLSTM)网络编码输入序列的远距离依赖关系,通过字形编码表获得繁体的仓颉码和简体的现代五笔码,采用卷积神经网络(CNN)提取字形特征表示,并根据不同权重组合繁体与简体字形特征,利用门控机制将它与经过BiLSTM编码后的字符表示进行组合;最后,使用条件随机场(CRF)解码,得到命名实体标注序列。在偏口语化的数据集Weibo、小型数据集Boson和大型数据集PeopleDaily上的实验结果表明,与基线模型MECT(Multi-metadata Embedding based Cross-Transformer)相比,所提模型的F1值别提高了2.47、1.20和0.98个百分点,验证了模型的有效性。
