NeighBERT: Medical Entity Linking Using Relation-Induced Dense Retrieval.

One of the common tasks in clinical natural language processing is medical entity linking (MEL) which involves mention detection followed by linking the mention to an entity in a knowledge base. One reason that MEL has not been solved is due to a problem that occurs in language where ambiguous texts can be resolved to several named entities. This problem is exacerbated when processing the text found in electronic health records. Recent work has shown that deep learning models based on transformers outperform previous methods on linking at higher rates of performance. We introduce NeighBERT, a custom pre-training technique which extends BERT (Devlin et al [1]) by encoding how entities are related within a knowledge graph. This technique adds relational context that has been traditionally missing in original BERT, helping resolve the ambiguity found in clinical text. In our experiments, NeighBERT improves the precision, recall, and F1-score of the state of the art by 1-3 points for named entity recognition and 10-15 points for MEL on two widely known clinical datasets. Supplementary Information: The online version contains supplementary material available at 10.1007/s41666-023-00136-3.

Synthesizing evidence for the external cycling of NOx in high- to low-NOx atmospheres.

External cycling regenerating nitrogen oxides (NOx ≡ NO + NO2) from their oxidative reservoir, NOz, is proposed to reshape the temporal-spatial distribution of NOx and consequently hydroxyl radical (OH), the most important oxidant in the atmosphere. Here we verify the in situ external cycling of NOx in various environments with nitrous acid (HONO) as an intermediate based on synthesized field evidence collected onboard aircraft platform at daytime. External cycling helps to reconcile stubborn underestimation on observed ratios of HONO/NO2 and NO2/NOz by current chemical model schemes and rationalize atypical diurnal concentration profiles of HONO and NO2 lacking noontime valleys specially observed in low-NOx atmospheres. Perturbation on the budget of HONO and NOx by external cycling is also found to increase as NOx concentration decreases. Consequently, model underestimation of OH observations by up to 41% in low NOx atmospheres is attributed to the omission of external cycling in models.

AmericasNLI: Machine translation and natural language inference systems for Indigenous languages of the Americas.

Little attention has been paid to the development of human language technology for truly low-resource languages-i.e., languages with limited amounts of digitally available text data, such as Indigenous languages. However, it has been shown that pretrained multilingual models are able to perform crosslingual transfer in a zero-shot setting even for low-resource languages which are unseen during pretraining. Yet, prior work evaluating performance on unseen languages has largely been limited to shallow token-level tasks. It remains unclear if zero-shot learning of deeper semantic tasks is possible for unseen languages. To explore this question, we present AmericasNLI, a natural language inference dataset covering 10 Indigenous languages of the Americas. We conduct experiments with pretrained models, exploring zero-shot learning in combination with model adaptation. Furthermore, as AmericasNLI is a multiway parallel dataset, we use it to benchmark the performance of different machine translation models for those languages. Finally, using a standard transformer model, we explore translation-based approaches for natural language inference. We find that the zero-shot performance of pretrained models without adaptation is poor for all languages in AmericasNLI, but model adaptation via continued pretraining results in improvements. All machine translation models are rather weak, but, surprisingly, translation-based approaches to natural language inference outperform all other models on that task.

Fuzzy-Match Repair Guided by Quality Estimation.

Computer-aided translation tools based on translation memories are widely used to assist professional translators. A translation memory (TM) consists of a set of translation units (TU) made up of source- and target-language segment pairs. For the translation of a new source segment s', these tools search the TM and retrieve the TUs (s,t) whose source segments are more similar to s'. The translator then chooses a TU and edit the target segment t to turn it into an adequate translation of s'. Fuzzy-match repair (FMR) techniques can be used to automatically modify the parts of t that need to be edited. We describe a language-independent FMR method that first uses machine translation to generate, given s' and (s,t), a set of candidate fuzzy-match repaired segments, and then chooses the best one by estimating their quality. An evaluation on three different language pairs shows that the selected candidate is a good approximation to the best (oracle) candidate produced and is closer to reference translations than machine-translated segments and unrepaired fuzzy matches ( t). In addition, a single quality estimation model trained on a mix of data from all the languages performs well on any of the languages used.

Low-cost measurement techniques to characterize the influence of home heating fuel on carbon monoxide in Navajo homes.

A large fraction of the global population relies on the inefficient combustion of solid fuels for cooking and home heating, resulting in household exposure to combustion byproducts. In the southwestern United States, unhealthy air quality has been observed in some homes that use solid fuels as a primary source of heat on the Navajo Nation. In order to better understand how home heating fuel choice can influence indoor air quality in this region, we used recently developed low-cost electrochemical sensors to measure carbon monoxide (CO) air mole fractions continuously inside and outside 41 homes in two communities on the Navajo Nation. Using low-cost sensors in this study, which don't require extensive training to operate, enabled collaboration with local Diné College students and faculty in the planning and implementation of home deployments. Households used natural gas, propane, pellets, wood, and/or coal for heating. We developed quantification methods that included uncertainty estimation for Alphasense CO-B4 sensors, for measurements both inside and outside homes. CO concentrations elevated above background were observed in homes in each heating fuel group, but the highest hourly concentrations were observed in wood and coal burning homes, some of which exceeded World Health Organization Guidelines on both an hourly and eight-hourly basis. In order to probe the many factors that can influence indoor pollutant concentrations, we developed and implemented methods that employ CO emission and decay time periods observed in homes during everyday activities to estimate air exchange rates as well as CO emission rates on the basis of a given well-mixed volume of air. The air quality measurement tools and methods demonstrated in this study can be readily extended to indoor air quality studies in other communities around the world to inform how home heating and cooking practices are influencing indoor air quality during normal daily activities.

Assessing a low-cost methane sensor quantification system for use in complex rural and urban environments.

Low-cost sensors have the potential to facilitate the exploration of air quality issues on new temporal and spatial scales. Here we evaluate a low-cost sensor quantification system for methane through its use in two different deployments. The first was a one-month deployment along the Colorado Front Range and included sites near active oil and gas operations in the Denver-Julesberg basin. The second deployment was in an urban Los Angeles neighborhood, subject to complex mixtures of air pollution sources including oil operations. Given its role as a potent greenhouse gas, new low-cost methods for detecting and monitoring methane may aid in protecting human and environmental health. In this paper, we assess a number of linear calibration models used to convert raw sensor signals into ppm concentration values. We also examine different choices that can be made during calibration and data processing, and explore cross-sensitivities that impact this sensor type. The results illustrate the accuracy of the Figaro TGS 2600 sensor when methane is quantified from raw signals using the techniques described. The results also demonstrate the value of these tools for examining air quality trends and events on small spatial and temporal scales as well as their ability to characterize an area - highlighting their potential to provide preliminary data that can inform more targeted measurements or supplement existing monitoring networks.

Rapid cycling of reactive nitrogen in the marine boundary layer.

Nitrogen oxides are essential for the formation of secondary atmospheric aerosols and of atmospheric oxidants such as ozone and the hydroxyl radical, which controls the self-cleansing capacity of the atmosphere. Nitric acid, a major oxidation product of nitrogen oxides, has traditionally been considered to be a permanent sink of nitrogen oxides. However, model studies predict higher ratios of nitric acid to nitrogen oxides in the troposphere than are observed. A 'renoxification' process that recycles nitric acid into nitrogen oxides has been proposed to reconcile observations with model studies, but the mechanisms responsible for this process remain uncertain. Here we present data from an aircraft measurement campaign over the North Atlantic Ocean and find evidence for rapid recycling of nitric acid to nitrous acid and nitrogen oxides in the clean marine boundary layer via particulate nitrate photolysis. Laboratory experiments further demonstrate the photolysis of particulate nitrate collected on filters at a rate more than two orders of magnitude greater than that of gaseous nitric acid, with nitrous acid as the main product. Box model calculations based on the Master Chemical Mechanism suggest that particulate nitrate photolysis mainly sustains the observed levels of nitrous acid and nitrogen oxides at midday under typical marine boundary layer conditions. Given that oceans account for more than 70 per cent of Earth's surface, we propose that particulate nitrate photolysis could be a substantial tropospheric nitrogen oxide source. Recycling of nitrogen oxides in remote oceanic regions with minimal direct nitrogen oxide emissions could increase the formation of tropospheric oxidants and secondary atmospheric aerosols on a global scale.

Secondary organic aerosol formation and organic nitrate yield from NO3 oxidation of biogenic hydrocarbons.

The secondary organic aerosol (SOA) mass yields from NO3 oxidation of a series of biogenic volatile organic compounds (BVOCs), consisting of five monoterpenes and one sesquiterpene (α-pinene, ß-pinene, Δ-3-carene, limonene, sabinene, and ß-caryophyllene), were investigated in a series of continuous flow experiments in a 10 m(3) indoor Teflon chamber. By making in situ measurements of the nitrate radical and employing a kinetics box model, we generate time-dependent yield curves as a function of reacted BVOC. SOA yields varied dramatically among the different BVOCs, from zero for α-pinene to 38-65% for Δ-3-carene and 86% for ß-caryophyllene at mass loading of 10 µg m(-3), suggesting that model mechanisms that treat all NO3 + monoterpene reactions equally will lead to errors in predicted SOA depending on each location's mix of BVOC emissions. In most cases, organonitrate is a dominant component of the aerosol produced, but in the case of α-pinene, little organonitrate and no aerosol is formed.

Secondary Organic Aerosol Formation via 2-Methyl-3-buten-2-ol Photooxidation: Evidence of Acid-Catalyzed Reactive Uptake of Epoxides.

Secondary organic aerosol (SOA) formation from 2-methyl-3-buten-2-ol (MBO) photooxidation has recently been observed in both field and laboratory studies. Similar to the level of isoprene, the level of MBO-derived SOA increases with elevated aerosol acidity in the absence of nitric oxide; therefore, an epoxide intermediate, (3,3-dimethyloxiran-2-yl)methanol (MBO epoxide), was synthesized and tentatively proposed to explain this enhancement. In this study, the potential of the synthetic MBO epoxide to form SOA via reactive uptake was systematically examined. SOA was observed only in the presence of acidic aerosol. Major SOA constituents, 2,3-dihydroxyisopentanol and MBO-derived organosulfate isomers, were chemically characterized in both laboratory-generated SOA and in ambient fine aerosol collected from the BEACHON-RoMBAS field campaign during the summer of 2011, where MBO emissions are substantial. Our results support the idea that epoxides are potential products of MBO photooxidation leading to the formation of atmospheric SOA and suggest that reactive uptake of epoxides may explain acid enhancement of SOA observed from other biogenic hydrocarbons.

Organosulfates as tracers for secondary organic aerosol (SOA) formation from 2-methyl-3-buten-2-ol (MBO) in the atmosphere.

2-Methyl-3-buten-2-ol (MBO) is an important biogenic volatile organic compound (BVOC) emitted by pine trees and a potential precursor of atmospheric secondary organic aerosol (SOA) in forested regions. In the present study, hydroxyl radical (OH)-initiated oxidation of MBO was examined in smog chambers under varied initial nitric oxide (NO) and aerosol acidity levels. Results indicate measurable SOA from MBO under low-NO conditions. Moreover, increasing aerosol acidity was found to enhance MBO SOA. Chemical characterization of laboratory-generated MBO SOA reveals that an organosulfate species (C(5)H(12)O(6)S, MW 200) formed and was substantially enhanced with elevated aerosol acidity. Ambient fine aerosol (PM(2.5)) samples collected from the BEARPEX campaign during 2007 and 2009, as well as from the BEACHON-RoMBAS campaign during 2011, were also analyzed. The MBO-derived organosulfate characterized from laboratory-generated aerosol was observed in PM(2.5) collected from these campaigns, demonstrating that it is a molecular tracer for MBO-initiated SOA in the atmosphere. Furthermore, mass concentrations of the MBO-derived organosulfate are well correlated with MBO mixing ratio, temperature, and acidity in the field campaigns. Importantly, this compound accounted for an average of 0.25% and as high as 1% of the total organic aerosol mass during BEARPEX 2009. An epoxide intermediate generated under low-NO conditions is tentatively proposed to produce MBO SOA.

Nitrate ion photochemistry at interfaces: a new mechanism for oxidation of alpha-pinene.

The photooxidation of 0.6-0.9 ppm alpha-pinene in the presence of a deliquesced thin film of NaNO(3), and for comparison increasing concentrations of NO(2), was studied in a 100 L Teflon(R) chamber at relative humidities from 72-88% and temperatures from 296-304 K. The loss of alpha-pinene and the formation of gaseous products were followed with time using proton transfer mass spectrometry. The yields of gas phase products were smaller in the NaNO(3) experiments than in NO(2) experiments. In addition, pinonic acid, pinic acid, trans-sobrerol and other unidentified products were detected in the extracts of the wall washings only for the NaNO(3) photolysis. These data indicate enhanced loss of alpha-pinene at the NaNO(3) thin film during photolysis. Supporting the experimental results are molecular dynamics simulations which predict that alpha-pinene has an affinity for the surface of the deliquesced nitrate thin film, enhancing the opportunity for oxidation of the impinging organic gas during the nitrate photolysis. This new mechanism of oxidation of organics may be partially responsible for the correlation between nitrate and the organic component of particles observed in many field studies, and may also contribute to the missing source of SOA needed to reconcile model predictions and field measurements. In addition, photolysis of nitrate on surfaces in the boundary layer may lead to oxidation of co-adsorbed organics.

Approaches for quantifying reactive and low-volatility biogenic organic compound emissions by vegetation enclosure techniques - part B: applications.

The focus of the studies presented in the preceding companion paper (Part A: Review) and here (Part B: Applications) is on defining representative emission rates from vegetation for determining the roles of biogenic volatile organic compound (BVOC) emissions in atmospheric chemistry and aerosol processes. The review of previously published procedures for identifying and quantifying BVOC emissions has revealed a wide variety of experimental methods used by various researchers. Experimental details become increasingly critical for quantitative emission measurements of low volatility monoterpenes (MT) and sesquiterpenes (SQT). These compounds are prone to be lost inadvertently by uptake to materials in contact with the sample air or by reactions with atmospheric oxidants. These losses become more prominent with higher molecular weight compounds, potentially leading to an underestimation of their emission rates. We present MT and SQT emission rate data from numerous experiments that include 23 deciduous tree species, 14 coniferous tree species, 8 crops, and 2 shrubs. These data indicate total, normalized (30 degrees C) basal emission rates from <10 to 5600ngCg(-1)h(-1) for MT, and from <10 to 1150ngCg(-1)h(-1) for SQT compounds. Both MT and SQT emissions have exponential dependencies on temperature (i.e. rates are proportional to e(betaT)). The inter-quartile range of beta-values for MT was between 0.12 and 0.17K(-1), which is higher than the value commonly used in models (0.09K(-1)). However many of the MT emissions also exhibited light dependencies, making it difficult to separate light and temperature influences. The primary light-dependent MT was ocimene, whose emissions were up to a factor of 10 higher than light-independent MT emissions. The inner-quartile range of beta-values for SQT was between 0.15 and 0.21K(-1).

Approaches for quantifying reactive and low-volatility biogenic organic compound emissions by vegetation enclosure techniques - part A.

The high reactivity and low vapor pressure of many biogenic volatile organic compounds (BVOC) make it difficult to measure whole-canopy fluxes of BVOC species using common analytical techniques. The most appropriate approach for estimating these BVOC fluxes is to determine emission rates from dynamic vegetation enclosure measurements. After scaling leaf- and branch-level emission rates to the canopy level, these fluxes can then be used in models to determine BVOC influences on atmospheric chemistry and aerosol processes. Previously published reports from enclosure measurements show considerable variation among procedures with limited guidelines or standard protocols to follow. This article reviews this literature and describes the variety of enclosure types, materials, and analysis techniques that have been used to determine BVOC emission rates. The current review article is followed by a companion paper which details a comprehensive enclosure technique that incorporates both recommendations from the literature as well as insight gained from theoretical calculations and practical experiences. These methods have yielded new BVOC emission data for highly reactive monoterpenes (MT) and sesquiterpenes (SQT) from a variety of vegetation species.

Sesquiterpene emissions from pine trees--identifications, emission rates and flux estimates for the contiguous United States.

Biogenic volatile organic compound (BVOC) emissions were studied using vegetation enclosure experiments. Particular emphasis was given to sesquiterpene compounds (SQT), although monoterpenes (MT) were also characterized. SQT were detected in emissions from seven (out of eight) pine species that were examined. Thirteen SQT compounds were identified; the most abundant ones were beta-caryophyllene, alpha-bergamotene, beta-farnesene, and alpha-farnesene, with emission rates increasing exponentially with temperature. Regression analysis yielded exponential dependencies of both MT and SQT emissions on temperature of the form E = E0 x exp(beta(T - T0)). This resulted in SQT basal emission rates (E0 defined at T0 = 30 degrees C) ranging between <4 and 620 ng (carbon) gdw(-1) h(-1) (gdw = gram dry weight). The average value of the exponential temperature response factor beta for SQT emissions, taken from all experiments, was 0.17 degree C(-1), whereas the value for monoterpenes was 0.11 degrees C(-1). The average, total SQT emissions from pines were estimated to be 9, 16, and 29% of the MT emissions at 20, 30, and 40 degrees C respectively. The emission factors and beta-factors determined from these measurements were used to estimate pine tree MT and SQT emission distributions for the contiguous United States using MEGAN (model of emissions of gases and aerosols from nature, Guenther et al., 2006). SQT fluxes reaching 10-40 mg m(-2) for the month of July were estimated for extensive areas of most western and southern U.S. states.

Analysis of atmospheric sesquiterpenes: sampling losses and mitigation of ozone interferences.

Atmospheric standards containing parts-per-billion levels of 14 semivolatile hydrocarbon compounds, including eight sesquiterpenes (SQTs) (longipinene, alpha-copaene, isolongifolene, alpha-cedrene, trans-caryophyllene, aromadendrene, alpha-humulene, delta-cadinene), two oxidized sesquiterpenoids (cisnerolidol, trans-nerolidol), one biogenic ketone (geranylacetone) and three aromatic compounds (1,3,5-triisopropylbenzene, diphenylmethane, nonylbenzene), were collected onto four solid adsorbent materials at increasing ozone mixing ratios (0-100 ppbv 03) for analysis by thermodesorption-gas chromatography. Substantial sampling losses of up to >90% were found for the most reactive SQT, even at the lowest ozone level investigated of 20 ppbv. Loss rates from the ozone-SQT reaction were used to derive estimates of gas-phase ozone reaction rate constants for longipinene, alpha-copaene, isolongifolene, geranylacetone, aromadendrene, delta-cadinene, cis-nerolidol, and transnerolidol. Three different ozone mitigation techniques were investigated to prevent these sampling losses. These strategies included (a) placing glass fiber filters impregnated with sodium thiosulfate (Na2S2O3) into the sampling line, (b) titration of ozone in the sampling stream with nitric oxide (NO), and (c) catalytically removing ozone with a commercially available manganese dioxide (MnO2) catalyst. All three techniques reduced ozone-mixing ratios from 100 ppbv to <0.6 ppbv at sampling flow rates of 1 L min(-1). When the Na2S2O3 filters and the NO-titration techniques were applied, SQT loss rates decreased from 25-60% to 0-5% for most SQT compounds and from >90% to approximately 10-50% for the two most reactive compounds at ozone mixing ratios of up to 100 ppbv. The commercial manganese dioxide scrubber, however, caused complete analyte losses (>98%) even at 0 ppbv ozone. These results underline the need and present applicable techniques for removal of ozone in air samples for SQT analysis by solid adsorption techniques.