复杂产品虚拟样机异构的结构信息集成研究

产品结构信息的异构性给网络化环境中的产品开发设计人员增加了复杂产品虚拟样机协同制造的难度。为了解决这个问题,以开发国家863/CIMS主题某课题为背景,论述了如何在网络化协同制造环境中建立面向对象的产品结构信息数据模型,并采用Web服务技术、软件集成技术、可扩展标记语言序列化技术以及产品数据管理技术规范,通过数据交换标准格式,实现产品结构信息的自动提取和构造产品信息集成框架,消除由于使用不同的计算机辅助设计系统所带来的信息异构性,有效地实现虚拟样机的产品信息集成和管理,为实施复杂产品虚拟样机工程中的异构产品结构信息的集成与共享提供了有效的方案,其可行性得到了实际工程的验证。

Segmental Duplications Are Common in Rice Genome

Segmental duplications on rice (Oryza sativa L.) chromosomes 8, 9, 11, and 12 were studied by examining the distributions of sequences resolved by 13 probes detecting multiple copies of DNA sequences. Four of the hybridization bands detected by a repetitive sequence probe, rTRS, were mapped to the ends of all the four chromosomes. Two or three of the bands detected by each of the other 12 probes were also mapped to different chromosomes. The bands detected by the same probe usually occurred in similar locations of different chromosomes. Loci detected by different DNA probes were often similarly arranged on different chromosomes. Chromosomes 8 and 9 showed colinearity of marker loci arrangement indicating a possible common origin. A segment on chromosome 9 was also very similar to the previously reported duplicated fragments on the ends of chromosomes 11 and 12 which were also detected in this study, indicating a likely common origin. Moreover, the various degrees of distributional similarity of the segments suggest a complex relationship among the chromosomes in the evolution of the rice genome. These results support the proposition that chromosome duplication and diversification may be a mechanism for the origin and evolution of the chromosomes in the rice genome.

融合双向语言模型的半监督属性识别方法研究

属性级情感分析是细粒度的情感分类任务,属性识别是属性情感分析任务中的重要步骤,作为典型的序列化标记问题处理。近年来,不依赖人工特征的神经网络模型在分词等序列化标记任务中表现出了很好的性能。为此,论文提出一种融合双向语言模型的半监督序列化标记网络模型(ST-BiLM),首先利用卷积神经网络(CNN)训练出单词具有形态特征的字符级向量,然后组合具有语义特征的词向量作为模型的输入,再利用ST-BiLM模型来进行实体属性识别。实验结果表明,模型在SemEval2014 Task4和Twitter不同领域的数据集上都达到了很好的结果,F1值分别为88.27%,83.15%和86.36%。