Search

This survey presents a comprehensive description of recent neural entity linking (EL) systems developed since 2015 as a result of the "deep learning revolution" in natural language processing. Its goal is to systemize design features of neural entity linking systems and compare their performance to the remarkable classic methods on common benchmarks. This work distills a generic architecture of a neural EL system and discusses its components, such as candidate generation, mention-context encoding, and entity ranking, summarizing prominent methods for each of them. The vast variety of modifications of this general architecture are grouped by several common themes: joint entity mention detection and disambiguation, models for global linking, domain-independent techniques including zero-shot and distant supervision methods, and cross-lingual approaches. Since many neural models take advantage of entity and mention/context embeddings to represent their meaning, this work also overviews prominent entity embedding techniques. Finally, the survey touches on applications of entity linking, focusing on the recently emerged use-case of enhancing deep pre-trained masked language models based on the Transformer architecture., Published in Semantic Web journal

Unipolar and quasi-unipolar half-cycle pulses having nonzero electric pulse area are a limit of pulse shortening in a given spectral range. In spite of the fact that existence of such pulses was considered by Jackson (1962), V.L. Ginzburg (1960-s), Bullough and Ahmad (1971) as well as Bessonov (1981), the possibility of their existence and propagation in space remained questionable for many years. Only within the past decades both the possibility of unipolar pulse existence and their propagation dynamics were shown and analyzed in detail theoretically and experimentally. So far such pulses became a subject of active research due to their potential in the ultra-fast optics and study of new regimes of light-matter interactions with subcycle resolution. Here, we show the possibility of the effective ultrafast control of the level populations in quantum well nanostructures by the half-cycle unipolar attosecond light pulses in comparison to the single-cycle ones. It is shown that the population dynamics can be determined by the electric pulse area divided to its characteristic "scale" determined by the quantum well width. The selective excitation of quantum states and the feasibility of the population inversion by the subcycle unipolar pulses is demonstrated.

The generalization of the area theorem is derived for the case of a pulse circulating inside a ring laser cavity. In contrast to the standard area theorem, which is valid for a single pass of a traveling pulse through a resonant medium, the obtained generalized area theorem takes into account the medium-assisted nonlinear self-action effects through the medium excitation left by the pulse at the previous round-trip in the cavity. The generalized area theorem was then applied to the theoretical description of the dynamics of a single-section ring-cavity laser and the steady solutions for the pulse area and for the medium parameters were found both in the limit of a lumped model and for a spatially-extended system. The derived area theorem can be used for the convenient analytical description of different coherent photonic devices, like coherently mode-locked lasers or pulse compressors, as well as for the analysis of the photon echo formation in cavity-based setups.

We propose a simple quantum system, namely, a nested quantum-well structure, which is able to generate a train of half-cycle pulses of a few-fs duration, when driven by a static electric field. We theoretically investigate the emission of such a structure and its dependence on the parameters of the quantum wells. It is shown that the production of a regular output pulse train with tunable properties and the pulse repetition frequencies of tens of THz is possible in certain parameter ranges. We expect the suggested structure can be used as an ultra-compact source of subcycle pulses in the optical range.

Here we consider coherent mode-locking (CML) regimes in single-section cavity lasers, taking place for pulse durations less than atomic population and phase relaxation times, which arise due to coherent Rabi oscillations of the atomic inversion. Typically, CML is introduced for lasers with two sections, the gain and absorber ones. Here we show that, for certain combination of the cavity length and relaxation parameters, a very stable CML in a laser, containing only gain section, may arise. The mode-locking is unconditionally self-starting and appears due to balance of intra-pulse de-excitation and slow interpulse-scale pump-induced relaxation processes. We also discuss the scaling of the system to shorter pulse durations, showing a possibility of mode-locking for few-cycle pulses. © 2021, The Author(s).

We study theoretically the possibility of few-cycle short bipolar optical pulse compression and their transformation to unipolar pulses due to coherent interaction with resonance absorbing medium. It is shown that single-cycle pulse compression occurs when each half-wave starts to behave as an independent unipolar soliton. These solitons are attracted to each other under certain conditions, that leads to the emergence of single-cycle pulse of shorter duration. Numerical simulations revealed 3-5 times reduction of the pulse duration. The substantial absence of light loss in this scheme gives the possibility to create a multistage passive system of three resonance absorbers and results in a 125-time reduction of the pulse duration. Generation of unipolar pulses occurs when two powerful extremely short bipolar pulses propagate and collide in a dense resonant medium. In this case, as shown by numerical calculations, the mutual influence of oncoming solitons leads to the fact that some part of them is destroyed and another part is not. A high power unipolar soliton and low intensity bipolar optical ringing are observed in the medium output.

The paper introduces methods of adaptation of multilingual masked language models for a specific language. Pre-trained bidirectional language models show state-of-the-art performance on a wide range of tasks including reading comprehension, natural language inference, and sentiment analysis. At the moment there are two alternative approaches to train such models: monolingual and multilingual. While language specific models show superior performance, multilingual models allow to perform a transfer from one language to another and solve tasks for different languages simultaneously. This work shows that transfer learning from a multilingual model to monolingual model results in significant growth of performance on such tasks as reading comprehension, paraphrase detection, and sentiment analysis. Furthermore, multilingual initialization of monolingual model substantially reduces training time. Pre-trained models for the Russian language are open sourced.

Designing of underground constructions, such as irrigation and drainage systems, requires engineering surveys. Such surveys include the study of the hydrological conditions of the territory, which are determined by the hydrophysical properties of soils, such as their water-retention capacity. The formation of a hysteresis loop for the soil water-retention capacity occurs because of the variability of meteorological conditions. It is almost impossible to measure all possible scanning branches that fill the gap between the main branches of the hysteresis loop. A mathematical model of the hysteretic soil water-retention capacity is proposed. The model is based on physical concepts of the structure and capillary properties of the soil pore space. Model parameters are identified by dot fitting using data on the main (boundary) hysteresis branches. Scanning branches starting with pre-calculated reversal points are evaluated. Suggested model has a quite low error to predict the scanning branches of soil water-retention capacity. The use of this model ensures reliable estimates of the hydrological conditions of the territory for underground construction. Also it gives precision irrigation rates which result in reduction of gravitational runoff of excessive moisture, preventing pollution of groundwater with agrochemicals., Проектирование объектов подземного строительства, таких как ирригационные и дренажные системы, требуют проведения инженерных изысканий. Такие изыскания включают в себя изучение гидрологических условий территории, которые определяются гидрофизическими свойствами почв, например, их водоудерживающей способностью. Изменчивостью метеорологических условий обусловлено формирование петли гистерезиса водоудерживающей способности почвы. Измерение всех возможных сканирующих ветвей, заполняющих промежуток между главными ветвями петли гистерезиса, практически невозможно. Предложена математическая модель гистерезиса водоудерживающей способности почвы. Модель основана на физических представлениях о строении и капиллярных свойствах пространства почвенных пор. Параметры модели идентифицируются путем точечной аппроксимации с использованием данных о главных (граничных) ветвях гистерезиса. Оценены сканирующие ветви, начинающиеся с предварительно рассчитанных поворотных точек. Наряду с обеспечением достоверных оценок гидрологических условий территории подземного строительства, использование данной модели позволяет более точно рассчитать нормы орошения сельскохозяйственных культур. Применение прецизионных норм орошения предотвращает гравитационный сток избыточной влаги, существенно уменьшает вымывание удобрений, мелиорантов и средств защиты растений за пределы корнеобитаемого слоя почвы и, как следствие, снижает риск загрязнения грунтовых вод агрохимикатами, что имеет важное эколого-экономическое значение.

International audience; Ticagrelor may be a more effective antiplatelet therapy than aspirin for the prevention of recurrent stroke and cardiovascular events in patients with acute cerebral ischemia.; We conducted an international double-blind, controlled trial in 674 centers in 33 countries, in which 13,199 patients with a nonsevere ischemic stroke or high-risk transient ischemic attack who had not received intravenous or intraarterial thrombolysis and were not considered to have had a cardioembolic stroke were randomly assigned within 24 hours after symptom onset, in a 1:1 ratio, to receive either ticagrelor (180 mg loading dose on day 1 followed by 90 mg twice daily for days 2 through 90) or aspirin (300 mg on day 1 followed by 100 mg daily for days 2 through 90). The primary end point was the time to the occurrence of stroke, myocardial infarction, or death within 90 days.; During the 90 days of treatment, a primary end-point event occurred in 442 of the 6589 patients (6.7%) treated with ticagrelor, versus 497 of the 6610 patients (7.5%) treated with aspirin (hazard ratio, 0.89; 95% confidence interval [CI], 0.78 to 1.01; P=0.07). Ischemic stroke occurred in 385 patients (5.8%) treated with ticagrelor and in 441 patients (6.7%) treated with aspirin (hazard ratio, 0.87; 95% CI, 0.76 to 1.00). Major bleeding occurred in 0.5% of patients treated with ticagrelor and in 0.6% of patients treated with aspirin, intracranial hemorrhage in 0.2% and 0.3%, respectively, and fatal bleeding in 0.1% and 0.1%.; In our trial involving patients with acute ischemic stroke or transient ischemic attack, ticagrelor was not found to be superior to aspirin in reducing the rate of stroke, myocardial infarction, or death at 90 days. (Funded by AstraZeneca; ClinicalTrials.gov number, NCT01994720.).

We observe experimentally a mode-locking in a continuous narrow-band tunable dye laser with molecular iodine absorber cells, which transitions have large phase relaxation time T2. We show that the mode-locking arises due to coherent interaction of light with the absorbing medium leading to Rabi oscillations, so that zero-area (0{\pi}-) pulses in the absorber are formed. Such mode-locking regime is different to most typical passive modelocking mechanisms where saturation plays the main role.

In the present work we report the possibility of passive mode-locking based on the coherent interaction of light with the amplifying and absorbing media in lasers with ring and linear cavities. We consider the realistic and practically interesting case when the absorbing and amplifying media are separated in the cavity space but not homogeneously mixed in the volume of the cavity, as was considered earlier in the literature. We perform qualitative consideration of coherent passive mode-locking based on the area theorem of McCall and Hahn and its graphical representation for the first time. We show that other, not soliton scenarios of passive mode-locking exist, and that coherent mode-locking is self-starting (lasing without an injection seeding pulse is possible). We point to the fact that the spectral width of the laser generation can be significantly larger than the spectral bandwidth of the gain medium. Numerical simulations were performed using the system of Maxwell-Bloch equations in the slowly varying envelope approximation.

This paper discusses the features of transient radiation from periodic one-dimensional resonant medium excited by ultrashort pulse. The case of circular geometry is considered for the harmonic distribution of the density of the particles along the circle. It is shown that a new frequency component arises in the spectrum of the scattered radiation in addition to the resonance frequency of medium. The new frequency appears both in the case of linear and nonlinear interaction, its value depends on the velocity of excitation pulse propagation and on the period of spatial modulation. In addition, the case when excitation moves at superluminal velocity and Cherenkov radiation arises is also studied.

In the present work we report the possibility of passive mode-locking based on the coherent interaction of light with the amplifying and absorbing media in lasers with ring and linear cavities. We consider the realistic and practically interesting case when the absorbing and amplifying media are separated in the cavity space but not homogeneously mixed in the volume of the cavity, as was considered earlier in the literature. We perform qualitative consideration of coherent passive mode-locking based on the area theorem of McCall and Hahn and its graphical representation for the first time. We show that other, not soliton scenarios of passive mode-locking exist, and that coherent mode-locking is self-starting (lasing without an injection seeding pulse is possible). We point to the fact that the spectral width of the laser generation can be significantly larger than the spectral bandwidth of the gain medium. Numerical simulations were performed using the system of Maxwell-Bloch equations in the slowly varying envelope approximation.