Analysis of the Full-Size Russian Corpus of Internet Drug Reviews with Complex NER Labeling Using Deep Learning Neural Networks and Language Models

The paper presents the full-size Russian corpus of Internet users’ reviews on medicines with complex named entity recognition (NER) labeling of pharmaceutically relevant entities. We evaluate the accuracy levels reached on this corpus by a set of advanced deep learning neural networks for extracting mentions of these entities. The corpus markup includes mentions of the following entities: medication (33,005 mentions), adverse drug reaction (1778), disease (17,403), and note (4490). Two of them—medication and disease—include a set of attributes. A part of the corpus has a coreference annotation with 1560 coreference chains in 300 documents. A multi-label model based on a language model and a set of features has been developed for recognizing entities of the presented corpus. We analyze how the choice of different model components affects the entity recognition accuracy. Those components include methods for vector representation of words, types of language models pre-trained for the Russian language, ways of text normalization, and other pre-processing methods. The sufficient size of our corpus allows us to study the effects of particularities of annotation and entity balancing. We compare our corpus to existing ones by the occurrences of entities of different types and show that balancing the corpus by the number of texts with and without adverse drug event (ADR) mentions improves the ADR recognition accuracy with no notable decline in the accuracy of detecting entities of other types. As a result, the state of the art for the pharmacological entity extraction task for the Russian language is established on a full-size labeled corpus. For the ADR entity type, the accuracy achieved is 61.1% by the F1-exact metric, which is on par with the accuracy level for other language corpora with similar characteristics and ADR representativeness. The accuracy of the coreference relation extraction evaluated on our corpus is 71%, which is higher than the results achieved on the other Russian-language corpora.

Сравнительное исследование термостойкости пелликлов на основе бериллия

We demonstrate the possibility of manufacturing Be-based ultrathin films with high transmission at wavelengths of 11.4 and 13.5 nm. For free-standing films of Be and Be-based multilayer structures (Si/Be, ZrSi2/Be, Be/BexNy, Zr/Be, Ru/Be, Mo/Be), we determine the thresholds of the absorbed power at which over a short period (tens of minutes) of vacuum annealing, initially sagging free-standing films became visibly stretched over the hole. Of the film structures tested here, the Be/BexNy structure (with beryllium nitride interlayers) showed the highest threshold for the absorbed power (1 W/cm2). However, due to the low strength of this structure, ZrSi2/Be, Mo/Be, and Be films seem to be more promising for the manufacture of a full-size pellicle. Long-term vacuum annealing of Mo/Be and Be ultrathin films showed that they could withstand 24 hours of vacuum heating at an absorbed power density of 0.2 W/cm2 (film temperature ~250°C) without noticeable changes in EUV transmission or sagging of films. With comparable transmission (~83% at 13.5 nm and ~88% at 11.4 nm), a multilayer Mo/Be structure with a thickness of 30 nm appears to be preferable, as it shows less brittleness than a monolayer Be film with a thickness of 50 nm.

Online Section

Online section presents full-size, uninterrupted figures presented on sequential pages in the printed edition.

Investigation of a full-size damper for an electrically driven centrifugal pump for oil production

The article describes a way to combat fatigue effects in the details of connecting modules of an electric driven centrifugal pump unit for oil production. A constructive solution for implementing the method in relation to complex downhole conditions in the form of a multifunctional damper using a differential piston to transfer it from the transport position when lowering into the well into the working one is shown. For a full-size damper, experimental studies of its vibration- isolating characteristics have been carried out when used in the form of substrates for supporting arms of elastomers of various densities and compositions. The preferred characteristics of elastomers and their ranking for various frequencies of forced vibrations are determined. Keywords: module, connection parts; electrically driven centrifugal pump unit; electrocentrifugal pumping unit; differential piston; damper; sbstrate; vibration velocity.

Evaluation of Antibiotic Resistance of Salmonella Serotypes and Whole-Genome Sequencing of Multiresistant Strains Isolated from Food Products in Russia

Food products may be a source of Salmonella, one of the main causal agents of food poisoning, especially after the emergence of strains resistant to antimicrobial preparations. The present work dealt with investigation of the occurrence of resistance to antimicrobial preparations among S. enterica strains isolated from food. The isolates belonged to 11 serovars, among which Infantis (28%), Enteritidis (19%), and Typhimurium (13.4%) predominated. The isolates were most commonly resistant to trimethoprim/sulfamethoxazole (n = 19, 59.38%), cefazolin (n = 15, 46.86%), tetracycline (n = 13, 40.63%), and amikacin (n = 9, 28.13%). Most of the strains (68.75%) exhibited multiple resistance to commonly used antibiotics. High-throughput sequencing was used to analyse three multidrug-resistant strains (resistant to six or more antibiotics). Two of them (SZL 30 and SZL 31) belonged to S. Infantis, while one strain belonged to S. Typhimurium (SZL 38). Analysis of the genomes of the sequenced strains revealed the genes responsible for antibiotic resistance. In the genomes of strains SZL 30 and SZL 31 the genes of antibiotic resistance were shown to be localized mostly in integrons within plasmids, while most of the antibiotic resistance genes of strain SZL 38 were localized in a chromosomal island (17,949 nt). Genomes of the Salmonella strains SZL 30, SZL 31, and SZL 38 were shown to contain full-size pathogenicity islands: SPI-1, SPI-2, SPI-4, SPI-5, SPI-9, SPI-11, SPI-13, SPI-14, and CS54. Moreover, the genome of strain SZL 38 was also found to contain the full-size pathogenicity islands SPI-3, SPI-6, SPI-12, and SPI-16. The emergence of multidrug-resistant strains of various Salmonella serovars indicates that further research on the transmission pathways for these genetic determinants and monitoring of the distribution of these microorganisms are necessary.

A Simplified FE Modeling Strategy for the Drop Process Simulation Analysis of Light and Small Drone

The numerical accuracy of drop process simulation and collision response for drones is primarily determined by the finite element modeling method and simplified method of drone airframe structure. For light and small drones exhibiting diverse shapes and configurations, mixed materials and structures, deformation and complex destruction behaviors, the way of developing a reasonable and easily achieved high-precision simplified modeling method by ensuring the calculation accuracy and saving the calculation cost has aroused increasing concern in impact dynamics simulation. In the present study, the full-size modeling and simplified modeling methods that are specific to different components of a relatively popular light and small drone were analyzed in an LS-DYNA software environment. First, a full-size high-precision model of the drone was built, and the model accuracy was verified by performing the drop tests at the component level as well as the whole machine level. Subsequently, based on the full-size high-precision model, the property characteristics of the main components of the light and small drone and their common simplification methods were classified, a series of simplified modeling methods for different components were developed, several single simplified models and combined simplified models were built, and a method to assess the calculation error of the peak impact load in the simplified models was proposed. Lastly, by comparing and analyzing the calculation accuracy of various simplified models, the high-precision simplified modeling strategy was formulated, and the suggestions were proposed for the impact dynamics simulation of the light and small drone falling. Given the analysis of the calculation scale and solution time of the simplified model, the high-precision simplified modeling method developed here is capable of noticeably reducing the modeling difficulty, the solution scale and the calculation time while ensuring the calculation accuracy. Moreover, it shows promising applications in several fields (e.g., structure design, strength analysis and impact process simulation of drone).