为了解决农业病虫害命名实体识别过程中存在的内在语义信息缺失、局部上下文特征易被忽略和捕获长距离依赖能力不足等问题,以农业病虫害文本为研究对象,提出一种基于部首嵌入和注意力机制的农业病虫害命名实体识别模型(Chinese agricultural diseases and pests named entity recognition with joint radical embedding and self attention, RS-ADP)。首先,该模型将部首嵌入集成到字符嵌入中作为输入,用以丰富语义信息。其中,针对部首嵌入设计了3种特征提取策略,即卷积神经网络(Convolutional neural network, CNN)、双向长短时记忆网络(Bidirectional long short term memory network, BiLSTM) 和CNN-BiLSTM;其次,采用多层不同窗口尺寸的CNNs层提取不同尺度的局部上下文信息;然后,在BiLSTM提取全局序列特征的基础上,采用自注意力机制进一步增强模型提取更长距离依赖的能力;最后,采用条件随机场(Conditional random field, CRF)联合识别实体边界和划分实体类别。在包含11个类别和24715条标注样本的农业病虫害自制语料上进行了实验。结果表明,本文模型RS-ADP在该数据集上精确率、召回率和F1值分别为94.16%、94.47%和94.32%;在具体实体类别上,RS-ADP在作物、病害、虫害等易识别实体上F1值高达95.81%、97.76%和97.23%。同时,RS-ADP在草害、病原等难以识别实体上F1值仍保持86%以上。实验结果表明,本文所提模型能够有效识别农业病虫害命名实体,其识别精度优于其他模型,且具有一定的泛化性。 相似文献
Biochar addition can expand soil organic carbon (SOC) stock and has potential ability in mitigating climate change. Also, some incubation experiments have shown that biochar can increase soil inorganic carbon (SIC) contents. However, there is no direct evidence for this from the field experiment. In order to make up the sparseness of available data resulting from the long‐term effect of biochar amendment on soil carbon fractions, here we detected the contents and stocks of the bulk SIC and SOC fractions based on a 10‐year field experiment of consecutive biochar application in Shandong Province, China. There are three biochar treatments as no‐biochar (control), and biochar application at 4.5 Mg ha?1 year?1 (B4.5) and 9.0 Mg ha?1 year?1 (B9.0), respectively. The results showed that biochar application significantly enhanced SIC content (3.2%–24.3%), >53 μm particulate organic carbon content (POC, 38.2%–166.2%) and total soil organic carbon content (15.8%–82.2%), compared with the no‐biochar control. However, <53 μm silt–clay‐associated organic carbon (SCOC) content was significantly decreased (14%–27%) under the B9.0 treatment. Our study provides the direct field evidence that SIC contributed to carbon sequestration after the biochar application, and indicates that the applied biochar was allocated mainly in POC fraction. Further, the decreased SCOC and increased microbial biomass carbon contents observed in field suggest that the biochar application might exert a positive priming effect on native soil organic carbon. 相似文献
In Hungary, maize is grown on 1 million ha and occupies more than 20% of the arable land. The rich assortment of maize cultivars of different vegetation periods and different responses to nutritional effects, water supply etc. gives the growers the possibility to choose the cultivars suiting best the site characteristics (Jolânkai et al. 1999). Among the cereals maize has the highest genetical potential. To utilize its yield and quality potential, soil types of high nutrient content and regular nutrient supply are required (Gyõrffy, 1979). Both over‐ and under‐fertilization have an unfavourable effect on the yield and quality of maize (Debreczeni, 1985). Crops can be supplied with the appropriate nutrient amounts only with the knowledge of soil characteristics in the different agro‐ecological regions (nutrient content, water supply, soil compactness, pH, nutrient supplying capacity etc.). In Hungary, a network of long‐term field fertilization trials with uniform fertilizer treatments has been maintained at nine experimental sites representing different agro‐ecological regions of the country. This experimental network gives a basis to test the nutrient responses of our main crops and calibrate their optimal nutrient supply (Kismányoky, 1991). 相似文献
In recent years, intestinal transport processes have been studied in detail regarding both, functional and structural aspects. For monosaccharides different systems have been demonstrated for apical uptake: this includes the high-affinity SGLT1 as a distinct d-glucose system and GLUT5 for fructose. Specifically in pigs a low affinity, high-capacity system for d-glucose and d-mannose with no preference for Na+ over K+ and a very low affinity system are suggested as further uptake systems. As in other species, basolateral extrusion is mediated by GLUT2. The distributions of monosaccharide transport along the gastrointestinal axis as well as the potential role of paracellular monosaccharide absorption have not yet been clarified.
Amino acids can principally be absorbed by the paracellular and transcellular pathway whereas transcellular transport can either be mediated by facilitated diffusion or secondary active Na+-coupled transport. This includes different transport systems for neutral, anionic and cationic acids. In addition, the presence of the di-/tripeptides transport system PEPT1 which depends on an inwardly directed H+-gradient has also been confirmed for the pig small intestine, its quantitative proportion is still under debate.
Short chain fatty acids (SCFA) are the major end products of microbial carbohydrate fermentation which occurs along the gastrointestinal tract with the highest production rates in the large intestines. At least two uptake mechanisms have to be assumed, i.e., non-ionic diffusion and anionic exchange via SCFA−/HCO3−-exchange. Controversial views still exist to what extent SCFA are metabolized within the epithelial tissue.
Segmental differences between small and large intestines have been demonstrated for Na+ absorption. Whereas in the small intestines the major part of Na+ absorption is mediated by coupled nutrient transport systems, aldosterone sensitive Na+ channels and Na+/H+-exchange are the dominant mechanisms in the hindgut. For Cl− paracellular transport and anionic Cl−/HCO3−-exchange are the major absorptive mechanisms. Cl− secretion is mediated by apical channels which may be activated by toxins of different origin. Different types of Cl− channels have been identified, such as Cystic Fibrosis Transmembrane Regulator (CFTR), Ca-activated Cl− channels (CLCA) and Outwardly Rectifying Cl− Channels (ORCC). Whereas CFTR has clearly been shown for jejunal and colonic epithelial and goblet cells controversy still exists on the relevance of CLCA and ORCC in pigs.
For Ca2+ there is evidence that both recently published channels TRPV5 and TRPV6 are also expressed in pig intestinal tissues, however, this has not yet been shown on protein level. From several functional approaches it was demonstrated that phosphate uptake can be mediated by both, a Na+-dependent transcellular component and paracellularly. On a molecular basis it is uncertain whether the transport protein of transcellular mechanism belongs to the NaPi-IIb cotransporter family. 相似文献