基于BERT-BILSTM-CRF模型的电力行业事故文本智能分析*

为解决电力行业事故报告文本较长、语义复杂,难以进行有效文本识别问题,提出1种以BERT作为底层的预训练模型,并设计1种双重注意力机制编码器,结合BILSTM-CRF深度挖掘事故文本语义特征,从而实现文本智能分析。首先构建电力词典,通过对BERT预训练,进行BIO标注,然后引入BILSTM-CRF模型实现对文本标签智能分类,最后将该模型与现行其他4种深度学习模型进行对比。研究结果表明:该模型智能识别精确率、召回率及F1值(查准率)均达到约97%,较其他4种模型中效果最好的模型分别提高0.02,0.03,0.02。研究结果可为电力行业事故报告文本分析提供1种新思路。     

不同pH值对磷矿废石磷素浸出特性影响的实验研究

采用EPA 1314pH酸碱预滴定实验和pH浸出实验,研究了pH对磷矿废石磷元素浸出特性的影响,以及磷矿废石的酸碱缓冲容量。实验结果表明:磷矿废石的碱缓冲容量小,pH从8.75上升到13.50消耗了0.1mmol/g的NaOH,磷矿废石酸的缓冲容量较大,pH从5.50到1.00,共消耗2.2mmol/g的H2SO4;浸提液的酸碱性是影响磷矿废石中磷元素浸出的重要参数,酸性条件下很大程度上促进了磷元素的浸出,随着H+离子浓度的增加,总磷浓度随之升高。相反在碱性条件下的反应过程中,OH-阻碍磷元素的浸出,消耗正磷酸根离子和羟基磷酸根离子,使其生成沉淀,有一定的固磷作用,为磷元素的固化提供了更多途径。     

基于演练决策文本分析的决策者分类及应急决策差异研究

为研究特大地震初期的高层决策行为，通过演练模拟决策者在可得灾情信息下进行决策部署的情形，采集参演决策者独立确认的最终决策文本为样本，对比分析样本对草案修改的方式和程度，将决策者分为独立型、合作型和依赖型。结构化处理提取样本的决策任务，分析与灾情信息及决策建议的相关性。结果表明:在高层决策者中合作型人数比例最高，群组内决策差异较小，质量相对稳定；独立型决策者偏好主观判断，对灾情的关注以及草案漏洞的审查等指标明显低于其他类型；依赖型决策者在使用合理化建议的前提下易做出高质量决策，但在决策支持失误而非漏洞的条件下，能否保证决策质量有待进一步验证。     

环境监测技术服务社会化的政策文本研究

以部分省(市)颁布的29份环境监测技术服务社会化政策为研究样本,将市场准入、监测领域、运行管理和监督制裁作为政策文本的分析框架进行计量分析。部分省(市)相关规定具有一定的创新性,使以往混乱的社会环境监测技术服务市场进一步规范化;但多数省(市)的环境监测技术服务社会化政策存在市场准入条件苛刻、监测领域有待放开、运行管理模式僵化、监督制裁手段单一等难题。最后提出放宽市场准入条件、有序开放监测领域、构建多元管理模式、完善监督制裁体系等4项建议对上述难题予以回应。     

The difficulties of systematic reviews

The need for robust evidence to support conservation actions has driven the adoption of systematic approaches to research synthesis in ecology. However, applying systematic review to complex or open questions remains challenging, and this task is becoming more difficult as the quantity of scientific literature increases. We drew on the science of linguistics for guidance as to why the process of identifying and sorting information during systematic review remains so labor intensive, and to provide potential solutions. Several linguistic properties of peer‐reviewed corpora—including nonrandom selection of review topics, small‐world properties of semantic networks, and spatiotemporal variation in word meaning—greatly increase the effort needed to complete the systematic review process. Conversely, the resolution of these semantic complexities is a common motivation for narrative reviews, but this process is rarely enacted with the rigor applied during linguistic analysis. Therefore, linguistics provides a unifying framework for understanding some key challenges of systematic review and highlights 2 useful directions for future research. First, in cases where semantic complexity generates barriers to synthesis, ecologists should consider drawing on existing methods—such as natural language processing or the construction of research thesauri and ontologies—that provide tools for mapping and resolving that complexity. These tools could help individual researchers classify research material in a more robust manner and provide valuable guidance for future researchers on that topic. Second, a linguistic perspective highlights that scientific writing is a rich resource worthy of detailed study, an observation that can sometimes be lost during the search for data during systematic review or meta‐analysis. For example, mapping semantic networks can reveal redundancy and complementarity among scientific concepts, leading to new insights and research questions. Consequently, wider adoption of linguistic approaches may facilitate improved rigor and richness in research synthesis.     

Seasonal fluctuations of DDTs and PCBs in zooplankton and fish of Lake Maggiore (Northern Italy)

Concentrations of DDTs and PCBs were determined in the zooplankton and in three different fish species (shad, whitefish and roach) collected seasonally during 2009 and 2010 in three sites in Lake Maggiore, a south-alpine lake that has been contaminated by DDT since 1996. As previously observed in 2008, even during 2009 DDTs concentrations were higher in zooplankton than in fish, probably due to the very unstable situation of the lake still influenced by local inputs. The situation changed in 2010, when all DDT compounds increased in fish to levels much higher than those measured in zooplankton. Biomagnification was statistically demonstrated for pp′DDE in all the three fish species, indicating a probable signal of recovery of the lake.Although with respect to total PCBs we observed that the contamination levels varied across time periods and across fish species, biomagnification was evident from zooplankton to fish both in 2009 and in 2010. As concern individual PCBs, biomagnification from zooplankton to all three fish species was significant for PCB 153 and PCB 138.     

Inferring chemical effects on carbon flows in aquatic food webs: Methodology and case study

The majority of ecotoxicological enclosure experiments monitor species abundances at different chemical concentrations. Here, we present a new modelling approach that estimates changes in food web flows from such data and show that population- and food web level effects are revealed that are not apparent from abundance data alone. For the case of cypermethrin in freshwater enclosures, photosynthesis and excretion (d⁻¹) of phytoplankton at 3.643 μg L⁻¹ cypermethrin were 30% lower and 100% higher than in the control, respectively. The ingestion rate of mesozooplankton (d⁻¹) was 6 times higher in the treated enclosures than in the control as food concentration increased with insecticide exposure. With increasing cypermethrin concentrations, nanoflagellates progressively relied on phytoplankton as their main food source, which rendered the food web less stable. We conclude that this tool has excellent potential to analyse the wealth of enclosure data as it only needs species abundance and general constraints.     

Do microbial processes regulate the stability of a coral atoll's enclosed pelagic ecosystem?

Complex marine ecosystems contain multiple feedback cycles that can cause unexpected responses to perturbations. To better predict these responses, complicated models are increasingly being developed to enable the study of feedback cycles. However, the sparseness of ecological data often limits the direct empirical parameterization of all model parameters. Here we use a Bayesian inverse analysis approach to synthesize empirical data and ecological theory derived from published studies of a coral atoll's enclosed pelagic ecosystem (Takapoto Atoll, French Polynesia). We then use the estimates of flux magnitudes to parameterize probabilistic compartment models with two forms of heterotrophic consumption: (1) “bottom-up” donor-controlled heterotrophic consumption and (2) “top-down” mass-action heterotrophic consumption. We explore how the flux magnitudes affect the ecosystem's stability properties of resilience, reactivity, and resistance under both assumptions for heterotrophic consumption. The models suggest that the microbial uptake of dissolved organic carbon (DOC) regulates the long term rate of return to steady state following a temporary or pulse perturbation (resilience), and the cycling of carbon between abiotic pools and heterotrophic compartments regulates the short-term response (reactivity). In the bottom-up process model, the sensitivity of steady state masses following a sustained or press perturbation (resistance) is highest for the DOC pool following a sustained change to the microbial uptake rate of DOC. Further, a change in the microbial uptake of DOC propagates through the ecosystem and affects the steady state values of zooplankton. The analysis suggests that the food web is highly dependent on the recycling between the abiotic and biotic carbon pools, particularly as mediated by the microbial consumption of DOC, and this recycling determines how the ecosystem responds to perturbations.     

Response of balanced network models to large-scale perturbation: Implications for evaluating the role of small pelagics in the Gulf of Maine

Exploring the response of an ecosystem, and subsequent tradeoffs among its biological community, to human perturbations remains a key challenge for the implementation of an ecosystem approaches to fisheries (EAF). To address this and related issues, we developed two network (or energy budget) models, Ecopath and Econetwrk, for the Gulf of Maine ecosystem. These models included 31 network “nodes” or biomass state variables across a broad range of trophic levels, with the present emphasis to particularly elucidate the role of small pelagics. After initial network balancing, various perturbation scenarios were evaluated to explore how potential changes to different fish, fisheries and lower trophic levels can affect model outputs. Categorically across all scenarios and interpretations thereof, there was minimal change at the second trophic levels and most of the “rebalancing” after a perturbation occurred via alteration of the diet matrix. Yet the model results from perturbations to a balanced energy budget fall into one of three categories. First, some model results were intuitive and in obvious agreement with established ecological and fishing theory. Second, some model results were counter-intuitive upon initial observation, seemingly contradictory to known ecological and fishing theory; but upon further examination the results were explainable given the constraints of an equilibrium energy budget. Finally, some results were counter-intuitive and difficult to reconcile with theory or further examination of equilibrium constraints. A detailed accounting of biomass flows for example scenarios explores some of the non-intuitive results more rigorously. Collectively these results imply a need to carefully track biomass flows and results of any given perturbation and to critically evaluate the conditions under which a new equilibrium is obtained for these types of models, which has implications for dynamic simulations based off of them. Given these caveats, the role of small pelagics as a prominent component of this ecosystem remains a robust conclusion. We discuss how one might use this approach in the context of further developing an EAF, recognizing that a more holistic, integrated perspective will be required as we continue to evaluate tradeoffs among marine biological communities.     

NEATS: A Network Economics Approach to Trophic Systems

The main principle of the economic approach to a trophic system we propose here lies in assuming that there is a transfer of food along a path between a prey and a predator if, for the predator, the benefits are greater than costs of predation on this path. Conversely, if the costs exceed the benefits, there are no flows. This trade-off, considered all along the food chains of an ecosystem, together with ecological processes (assimilation, somatic maintenance) results in a model coupling mass balance equations (biological constraints) and complementarity principles (Walras’ law). Here is the core of the Network Economics Approach to Trophic Systems (NEATS).