Populating Web-Scale Knowledge Graphs Using Distantly Supervised Relation Extraction and Validation

In this paper, we propose a fully automated system to extend knowledge graphs using external information from web-scale corpora. The designed system leverages a deep-learning-based technology for relation extraction that can be trained by a distantly supervised approach. In addition, the system uses a deep learning approach for knowledge base completion by utilizing the global structure information of the induced KG to further refine the confidence of the newly discovered relations. The designed system does not require any effort for adaptation to new languages and domains as it does not use any hand-labeled data, NLP analytics, and inference rules. Our experiments, performed on a popular academic benchmark, demonstrate that the suggested system boosts the performance of relation extraction by a wide margin, reporting error reductions of 50%, resulting in relative improvement of up to 100%. Furthermore, a web-scale experiment conducted to extend DBPedia with knowledge from Common Crawl shows that our system is not only scalable but also does not require any adaptation cost, while yielding a substantial accuracy gain.

A River Flood and Earthquake Risk Assessment of Railway Assets along the Belt and Road

Mitigating the disaster risk of transportation infrastructure networks along the Belt and Road is crucial to realizing the area’s high trade potential in the future. This study assessed the exposure and risk of existing and planned railway assets to river flooding and earthquakes. We found that about 9.3% of these railway assets are exposed to a one in 100 year flood event, and 22.3% are exposed to a one in 475 year earthquake event. The combined flood and earthquake risk of physical damage to railway assets, expressed by expected annual damage (EAD), is estimated at USD 1438 (between 966 and 2026) million. Floods contribute the majority of the risk (96%). China has the highest EAD for both floods and earthquakes (between USD 240 and 525 million in total). Laos and Cambodia are the countries with the highest EAD per km from flooding (USD 66,125–112,154 and USD 31,954–56,844 per km, respectively), while Italy and Myanmar have the highest EAD per km from earthquakes (USD 1000–3057 and USD 893–3019 per km, respectively). For the newly built and planned projects along the Belt and Road, the EAD is estimated at USD 271 (between 205 and 357) million. The China–Indochina Peninsula Economic Corridor and China–Pakistan Economic Corridor have the highest absolute EAD and EAD per km, with EADs reaching USD 95 and USD 67 million, and USD 18 and USD 17 thousand per km, on average, respectively. For railway segments with high risks, we found that if the required adaptation cost within 20 years to realize a 10% increase of the railway quality is below 8.4% of the replacement cost, the benefits are positive.

THE EFFECT OF DIFFERENT VOLUMES OF COMPUTER LOAD ON THE LATENT PERIODS OF A SIMPLE SENSORIMOTOR REACTION IN PRIMARY SCHOOLCHILDREN

Background. The digitalization of education makes relevant monitoring studies for assessing the impact of computer technologies on the functional status of children, their cognitive ca-pabilities and somatic support (“adaptation cost”). Aim. The paper aims to study the impact of different computer load on sensorimotor reactivity in primary schoolchildren. Materials and methods. The data obtained in Moscow schools in 2006-2011 were analyzed. Surveys were carried out twice a year (October, March-April) in 66 different educational organizations. In total, the study included data on 4205 first-fourth year schoolchildren. To evaluate the reaction time (RT) of simple sensorimotor reactions to light (L) and acoustic (A) stimuli, the computer movement meter equipment (CMM, INTOX, St. Petersburg, Russia) was used. Also, the RTA / RTL ratio was used for analysis. Results. There is a correlation between RT (both RTL and RTA) and the general (lesson-related and extracurricular) volume of computer load. It was different depending on gender and the season, and was opposite for RTL and RTA. It was established that excessive computer load (exceeding hygienic standard requirements by 3 or more times) increased the seasonal variability of RTL in spring testing, principally in the third and fourthyear girls. However, under the influence of high computer loads, the RTA / RTL ratio also changes – seasonal varia-bility is formed as a decrease of this indicator in spring testing, principally in the third and fourth-year boys. Conclusion. The data obtained indicate the ambiguity of the effect of high computer loads on sensorimotor reactivity. On the one hand, there are symptoms of fatigue in children during the academic year, which requires compensation through health promotion education. On the other hand, there is a formation of a new skill during the academic year. This period of skill formation is recommended for extension through the summer period.

THE EFFECT OF DIFFERENT VOLUMES OF COMPUTER LOAD ON THE LATENT PERIODS OF A SIMPLE SENSORIMOTOR REACTION IN PRIMARY SCHOOLCHILDREN

Background. The digitalization of education makes relevant monitoring studies for assessing the impact of computer technologies on the functional status of children, their cognitive ca-pabilities and somatic support (“adaptation cost”). Aim. The paper aims to study the impact of different computer load on sensorimotor reactivity in primary schoolchildren. Materials and methods. The data obtained in Moscow schools in 2006-2011 were analyzed. Surveys were carried out twice a year (October, March-April) in 66 different educational organizations. In total, the study included data on 4205 first-fourth year schoolchildren. To evaluate the reaction time (RT) of simple sensorimotor reactions to light (L) and acoustic (A) stimuli, the computer movement meter equipment (CMM, INTOX, St. Petersburg, Russia) was used. Also, the RTA / RTL ratio was used for analysis. Results. There is a correlation between RT (both RTL and RTA) and the general (lesson-related and extracurricular) volume of computer load. It was different depending on gender and the season, and was opposite for RTL and RTA. It was established that excessive computer load (exceeding hygienic standard requirements by 3 or more times) increased the seasonal variability of RTL in spring testing, principally in the third and fourthyear girls. However, under the influence of high computer loads, the RTA / RTL ratio also changes – seasonal varia-bility is formed as a decrease of this indicator in spring testing, principally in the third and fourth-year boys. Conclusion. The data obtained indicate the ambiguity of the effect of high computer loads on sensorimotor reactivity. On the one hand, there are symptoms of fatigue in children during the academic year, which requires compensation through health promotion education. On the other hand, there is a formation of a new skill during the academic year. This period of skill formation is recommended for extension through the summer period.

Climate change adaptation cost and residual damage to global crop production

Adaptation will be essential in many sectors, including agriculture, as a certain level of warming is anticipated even after substantial climate mitigation. However, global adaptation costs and adaptation limits in agriculture are understudied. Here, we estimate the global adaptation cost and residual damage (climate change impacts after adaptation) for maize, rice, wheat and soybean using a global gridded crop model and empirical production cost models. Producers require additional expenditures under climate change to produce the same crop yields that would be achieved without climate change, and this difference is defined as the adaptation cost. On a decadal mean basis, the undiscounted global cost of climate change (adaptation cost plus residual damage) for the crops are projected to increase with warming from 63 US$ billion (B) at 1.5°C to $80 B at 2°C and to $128 B at 3°C per year. The adaptation cost gradually increases in absolute terms, but the share decreases from 84% of the cost of climate change ($53 B) at 1.5°C to 76% ($61 B) at 2°C and to 61% ($8 B) at 3°C. The residual damage increases from 16% ($10 B) at 1.5°C to 24% ($19 B) at 2°C and to 39% ($50 B) at 3°C. Once maintaining yields becomes difficult due to the biological limits of crops or decreased profitability, producers can no longer bear adaptation costs, and residual damages increase. Our estimates offer a basis to identify the gap between global adaptation needs and the funds available for adaptation.

Collaborative Multi-Expert Active Learning for Mobile Health Monitoring: Architecture, Algorithms, and Evaluation

Mobile health monitoring plays a central role in the future of cyber physical systems (CPS) for healthcare applications. Such monitoring systems need to process user data accurately. Unlike in other human-centered CPS, in healthcare CPS, the user functions in multiple roles all at the same time: as an operator, an actuator, the physical environment and, most importantly, the target that needs to be monitored in the process. Therefore, mobile health CPS devices face highly dynamic settings generally, and accuracy of the machine learning models the devices employ may drop dramatically every time a change in setting happens. Novel learning architecture that specifically address challenges associated with dynamic environments are therefore needed. Using active learning and transfer learning as organizing principles, we propose a collaborative multiple-expert architecture and accompanying algorithms for the design of machine learning models that autonomously adapt to a new configuration, context, or user need. Specifically, our architecture and its constituent algorithms are designed to manage heterogeneous knowledge sources or experts with varying levels of confidence and type while minimizing adaptation cost. Additionally, our framework incorporates a mechanism for collaboration among experts to enrich their knowledge, which in turn decreases both cost and uncertainty of data labeling in future steps. We evaluate the efficacy of the architecture using two publicly available human activity datasets. We attain activity recognition accuracy of over 85 % (for the first dataset) and 92 % (for the second dataset) by labeling only 15 % of unlabeled data.