Mango (Mangifera indica L: Anacardiaceae) is regarded as the most liked fruit in tropical and sub-tropical regions of the world. Powdery mildew of mango caused by Oidium mangiferae Bert. is one of the major plant path...Mango (Mangifera indica L: Anacardiaceae) is regarded as the most liked fruit in tropical and sub-tropical regions of the world. Powdery mildew of mango caused by Oidium mangiferae Bert. is one of the major plant pathological constraints in growing healthy mango orchards. The apparent symptoms of the disease occurred on young tissues of all parts of flowers, leaves and fruits. Severe blossom infection could result in complete damage to fruit;flower failed to open and drop off from the inflorescence. The inflorescence revealed a pattern of disease from tip to downward and showed itself by the emergence of wefts of white mycelium on the affected parts. Twenty-five mango varieties i.e. Langra, Dusehri, Ratole No.12, Fajri, Sindhri, Chaunsa Samar Bahisht, Anwar ratole, Neelam, Yakta, Tota Pari, Sensation, Saroli, Malda, Ghulab e Khas, Chaunsa Black, Chaunsa white, Anmol, Almas, Shan e Ali, Shan e Mustafa, Mahmood Khan, Armughan, Zafaran, Malda Late and Early Gold were evaluated through the observation of symptoms on young inflorescence to determine the disease incidence, disease severity index and average yield of fruit. The study was carried out in a randomized compete block design with twenty-five treatments and three replications. The mango varieties presented different performance in relation to powdery mildew incidence and could be categorized into eight groups. Maximum disease incidence was observed (33.33%, 26.66% and 26.66%, 26.66%) on Dusehri, Chaunsa Samar Bahisht, Malda and Ratole No.12 respectively and minimum disease incidence was (3.66% and 3.66%) noted on Almas and Sensation. These two varieties showed to be tolerant against the disease. It might be concluded that the presence or absence of symptoms caused by the powdery mildew had no such effect on the fruit yield of the mango cultivars.展开更多
针对现有的专家推荐算法忽略了用户评论中蕴含的情感表达对专家专长表征的影响,从而导致推荐准确度不高的问题,提出基于双向编码器表示-多头注意力机制(bidirectional encoder representations from transformers-multi-head attention,...针对现有的专家推荐算法忽略了用户评论中蕴含的情感表达对专家专长表征的影响,从而导致推荐准确度不高的问题,提出基于双向编码器表示-多头注意力机制(bidirectional encoder representations from transformers-multi-head attention,BERT-MHA)的深度语义增强专家推荐算法。该算法基于预训练BERT模型,融合MHA机制,自动调整用户评论对专家历史回答问题的情感注意力权重,获取专家动态专长表征,并与静态专长联合以实现专家特征文本的语义增强,表征专家综合专长;通过注意力机制识别用户问题特征;采用多层感知机建模专家综合专长与用户问题间的非线性交互,预测推荐专家的匹配度。利用好大夫网站(haodf.com)的数据进行了不同参数配置及不同算法的对比实验,实验结果表明该算法在准确率(accuracy,ACC)和曲线下的面积(area under curve,AUC)指标下明显优于其他算法,能有效提高在线问答社区的专家推荐准确度。展开更多
在大数据时代,推荐算法有效缓解了信息过载问题,尤其在岗位推荐领域展现出重要价值。然而,针对高校毕业生的人岗推荐面临数据冷启动和数据稀疏性挑战,需综合考量专业、实习经历和就业意向等因素。本文提出基于Transformer的双向编码器表...在大数据时代,推荐算法有效缓解了信息过载问题,尤其在岗位推荐领域展现出重要价值。然而,针对高校毕业生的人岗推荐面临数据冷启动和数据稀疏性挑战,需综合考量专业、实习经历和就业意向等因素。本文提出基于Transformer的双向编码器表示(bidirectional encoder representation from Transformers,BERT)模型的混合推荐模型,设计冷启动与热启动双路径推荐策略。冷启动路径基于BERT模型计算岗位与学生嵌入向量的相似度,解决新用户历史数据缺失的困境,热启动路径基于既有用户行为数据,采用加权平均融合策略整合岗位相似度与用户相似度评分矩阵,以提升推荐精度。用户满意度调查显示:推荐岗位数量在“3~10个”时,符合预期或引起足够兴趣的百分比超70%,验证了该系统满足毕业生就业服务需求的有效性。展开更多
文摘Mango (Mangifera indica L: Anacardiaceae) is regarded as the most liked fruit in tropical and sub-tropical regions of the world. Powdery mildew of mango caused by Oidium mangiferae Bert. is one of the major plant pathological constraints in growing healthy mango orchards. The apparent symptoms of the disease occurred on young tissues of all parts of flowers, leaves and fruits. Severe blossom infection could result in complete damage to fruit;flower failed to open and drop off from the inflorescence. The inflorescence revealed a pattern of disease from tip to downward and showed itself by the emergence of wefts of white mycelium on the affected parts. Twenty-five mango varieties i.e. Langra, Dusehri, Ratole No.12, Fajri, Sindhri, Chaunsa Samar Bahisht, Anwar ratole, Neelam, Yakta, Tota Pari, Sensation, Saroli, Malda, Ghulab e Khas, Chaunsa Black, Chaunsa white, Anmol, Almas, Shan e Ali, Shan e Mustafa, Mahmood Khan, Armughan, Zafaran, Malda Late and Early Gold were evaluated through the observation of symptoms on young inflorescence to determine the disease incidence, disease severity index and average yield of fruit. The study was carried out in a randomized compete block design with twenty-five treatments and three replications. The mango varieties presented different performance in relation to powdery mildew incidence and could be categorized into eight groups. Maximum disease incidence was observed (33.33%, 26.66% and 26.66%, 26.66%) on Dusehri, Chaunsa Samar Bahisht, Malda and Ratole No.12 respectively and minimum disease incidence was (3.66% and 3.66%) noted on Almas and Sensation. These two varieties showed to be tolerant against the disease. It might be concluded that the presence or absence of symptoms caused by the powdery mildew had no such effect on the fruit yield of the mango cultivars.
文摘针对现有的专家推荐算法忽略了用户评论中蕴含的情感表达对专家专长表征的影响,从而导致推荐准确度不高的问题,提出基于双向编码器表示-多头注意力机制(bidirectional encoder representations from transformers-multi-head attention,BERT-MHA)的深度语义增强专家推荐算法。该算法基于预训练BERT模型,融合MHA机制,自动调整用户评论对专家历史回答问题的情感注意力权重,获取专家动态专长表征,并与静态专长联合以实现专家特征文本的语义增强,表征专家综合专长;通过注意力机制识别用户问题特征;采用多层感知机建模专家综合专长与用户问题间的非线性交互,预测推荐专家的匹配度。利用好大夫网站(haodf.com)的数据进行了不同参数配置及不同算法的对比实验,实验结果表明该算法在准确率(accuracy,ACC)和曲线下的面积(area under curve,AUC)指标下明显优于其他算法,能有效提高在线问答社区的专家推荐准确度。
文摘在大数据时代,推荐算法有效缓解了信息过载问题,尤其在岗位推荐领域展现出重要价值。然而,针对高校毕业生的人岗推荐面临数据冷启动和数据稀疏性挑战,需综合考量专业、实习经历和就业意向等因素。本文提出基于Transformer的双向编码器表示(bidirectional encoder representation from Transformers,BERT)模型的混合推荐模型,设计冷启动与热启动双路径推荐策略。冷启动路径基于BERT模型计算岗位与学生嵌入向量的相似度,解决新用户历史数据缺失的困境,热启动路径基于既有用户行为数据,采用加权平均融合策略整合岗位相似度与用户相似度评分矩阵,以提升推荐精度。用户满意度调查显示:推荐岗位数量在“3~10个”时,符合预期或引起足够兴趣的百分比超70%,验证了该系统满足毕业生就业服务需求的有效性。
