Yongliang Shen, Zeqi Tan, Shuhui Wu, Wenqi Zhang, Rongsheng Zhang, Yadong Xi, Weiming Lu, Yueting Zhuang. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Distantly supervised named entity recognition (DS-NER) aims to locate entity mentions and classify their types with only knowledge bases or gazetteers and unlabeled corpus.However, distant annotations are noisy and degrade the performance of NER models.In this paper, we propose a noise-robust prototype network named MProto for the DS-NER task.Different from previous prototype-based NER methods, MProto represents each entity type with multiple prototypes to characterize the intra-class variance among entity representations.To optimize the classifier, each token should be assigned an appropriate ground-truth prototype and we consider such token-prototype assignment as an optimal transport (OT) problem.Furthermore, to mitigate the noise from incomplete labeling, we propose a novel denoised optimal transport (DOT) algorithm.Specifically, we utilize the assignment result between Other class tokens and all prototypes to distinguish unlabeled entity tokens from true negatives.Experiments on several DS-NER benchmarks demonstrate that our MProto achieves state-of-the-art performance.The source code is now available on Github 1 .

Single scattering albedo (SSA) is an important parameter affecting the radiative forcing of aerosol. However, current SSA retrieval only relies on several typical aerosol models, limiting the range of SSA variation. This study proposed a new algorithm, and the comparison demonstrates the algorithm has an excellent ability to estimate SSA in pollution over land.

The air quality model can reproduce the processes of atmospheric transport, chemical reaction and removal, and analyze the spatio-temporal evolution law, internal mechanism and origin of air pollution. WRF-Chem model is the third generation of air quality model. Its biggest advantage is that the meteorological model and chemical transmission model are fully coupled in time and spatial resolution to achieve true online feedback, which is the main development direction of the future model. Therefore, this paper selected Xuzhou City, a heavy industry city with serious air pollution, as the research area, and took December when air quality was poor as the research time. WRF-Chem model was used to simulate and analyze meteorological parameters such as wind speed, temperature, relative humidity and concentration of air pollutants including PM2.5and O3in Xuzhou in December 2020. The simulation results were compared with the site data, and the simulation results under different parameter Settings of WRF-Chem mode were studied and compared.

Imported air pollution has a significant impact on urban air quality. Many urban air pollution events are not caused by local emissions, but by the transport of air pollutants from surrounding areas. Therefore, it is very necessary to prevent and control imported air pollution. However, the existing supervision of urban air quality mostly relies on ground monitoring stations, which is extremely limited in time and space. In this paper, MERRA-2 data is used to grasp urban air quality from a more macroscopic perspective, and combined with ground monitoring station data and meteorological data, the transmission route of air pollution is reconstructed. This paper takes Xuzhou City, Jiangsu Province as an example. It is proved that this method is highly feasible and can provide scientific data support for efficient prevention and control of imported air pollution.

Remote sensing inversion of global PM2.5is an important research topic. In the present study, the Aerosol Optical Depth (AOD) dataset was established by four geostationary satellites to estimate global PM2.5concentrations using improved Geographic Time-Weighted Regression model (IGTWR) models. Then a global hourly PM2.5concentration dataset was obtained in May 2020. The estimated result for PM2.5is verified at ground stations with R of 0.71 and RMSE (Root Mean Square Error) of 26.6 μg/m3. The results indicate that PM2.5has obvious spatial and temporal distribution in the world.

Based on the data of air quality monitoring stations and high-resolution remote sensing product data in Xuzhou City, the atmospheric pollutants in 2020 and 2021 were analyzed. The modified A-values method and the model simulation method were used to estimate the remaining atmospheric environmental capacity (RAEC) of PM10and PM2.5pollutants in Xuzhou, Jiangsu Province. The results show that the excessive PM10and PM2.5pollutants are the main problems of atmospheric pollution in Xuzhou, and the annual emissions still need to be reduced are 10.80×104t/a and 5.98×104t/a, respectively. Among them, the situation is the most serious in Tongshan District, which still needs to cut annual PM10and PM2.5emissions by 7.41×104t/a and 3.70×104t/a, respectively. Xinyi City has the smallest annual PM10emission reduction, which is 0.89×104t/a. Suining County needs to cut the smallest PM2.5emissions, at 0.67×104t/a. In addition, there are significant quarterly differences in RAEC, with the first quarter > the fourth quarter > the second quarter > the third quarter. Except for the third quarter, the excess atmospheric environmental capacity was the most serious in Xuzhou urban area, the other three quarters were the highest in Tongshan District.

Single scattering albedo (SSA) is an important parameter affecting the radiative forcing of aerosol. However, current SSA retrieval only relies on several typical aerosol models, limiting the range of SSA variation. This study proposed a new parameterization scheme of the aerosol model, which optimized the independent SSA value into mixed combination of three basic aerosol components, then constructed lookup table to obtain hourly SSA from the Advanced Himawari Image (AHI) sensor. The comparison is encouraging the retrieved SSA results agree well with AERONET, especially for high aerosol loading (AOD > 0.3 at 470nm), with correlation coefficient of R = 0.50, RMSE = 0.02, and approximately 93% of retrieval results falling within the 5% EE envelope. Under normal conditions, R = 0.46, RMSE = 0.04, and approximately 68% of retrieved SSA fall in within the uncertainty of$\Delta$SSA = ±0.05. The comparison demonstrates the algorithm has the excellent ability to estimate SSA in pollution over land.

The Greenhouse Gases Observing Satellite (GOSAT) is the world's first spacecraft to measure the concentrations of CO2from space and it has high-precision hyperspectral atmospheric CO2monitoring from 2009. Several atmospheric CO2products, such as ACOS, NIES, OCFP and SRFP products from full physics retrieval algorithm, both provide XCO2(the column-average dry-air mole fraction of atmospheric CO2) of GOSAT. These products have different characteristics and advantages and have different performance in different regions. In order to obtain the XCO2data set with high precision, low uncertainty and high coverage, the maximum likelihood estimation (MLE) method is used to fuse GOSAT XCO2products. The algorithm takes into account the uncertainty of each product on each pixel, and is applied from April 2009 to December 2015. The validation result between fusion XCO2and Total Carbon Column Observing Network XCO2shows R = 0.843 and RMSE = 3.248.

Dust as a kind of atmospheric pollutants, its harm are gradually being paid attention to. However, most of the existing supervision of dust relies on ground monitoring sites, which has strong spatial limitations. In this paper, high-resolution remote sensing images are used to identify regional dust sources, which can identify dust at a more macroscopic level from “plane”. At the same time, regional particulate emission inventory was established based on multi-source data, and then the dust emission is quantified. This paper takes Quanshan District, Xuzhou City, Jiangsu Province as an example, and the experimental results show that the method can reflect the actual dust situation well, and provide scientific data support for the supervision of urban dust.

Mining cities are an essential part of China's urban agglomerations, and as mining cities continue to develop, ecological and environmental pollution has become a primary problem. In the present study, the Aerosol Optical Depth (AOD) retrieval of major mining urban agglomerations in China from the Gaofen-1 satellite data. Then a new hybrid model based on CTM (chemical transport model) Transport Model 5 (TM5) and GTWR (Geographic Time-Weighted Regression model) is proposed for PM2.5 and PM10mass concentration estimation. According to the different transformation stages and urban structure of mining cities, the temporal and spatial analysis of particulate matter characteristics is carried out in mining urban agglomerations. The estimated result for PM2.5 is verified at ground stations with R2 of 0.956 and RMSE (Root Mean Square Error) of 10.377 μg/m3, Moreover, the estimated result for PM10is verified at ground stations with R2 of 0.926 and RMSE of 16.669 μg/m3, The results indicate that PM2.5 and PM10have distinct spatial and temporal distribution patterns as Chinese mining cities are undergoing different types of transformation processes.

Nested named entity recognition (nested NER) is a fundamental task in natural language processing. Various span-based methods have been proposed to detect nested entities with span representations. However, span-based methods do not consider the relationship between a span and other entities or phrases, which is helpful in the NER task. Besides, span-based methods have trouble predicting long entities due to limited span enumeration length. To mitigate these issues, we present the Propose-and-Refine Network (PnRNet), a two-stage set prediction network for nested NER. In the propose stage, we use a span-based predictor to generate some coarse entity predictions as entity proposals. In the refine stage, proposals interact with each other, and richer contextual information is incorporated into the proposal representations. The refined proposal representations are used to re-predict entity boundaries and classes. In this way, errors in coarse proposals can be eliminated, and the boundary prediction is no longer constrained by the span enumeration length limitation. Additionally, we build multi-scale sentence representations, which better model the hierarchical structure of sentences and provide richer contextual information than token-level representations. Experiments show that PnRNet achieves state-of-the-art performance on four nested NER datasets and one flat NER dataset.

The global long-term aerosol optical depth (AOD) dataset has a great significance for the study of global climate change. Advanced Very High Resolution Radiometer (AVHRR) on National Oceanic and Atmospheric Administration (NOAA) satellites can provide global observation from 1978 to present. In this paper, we have improved the algorithm for the retrieval of the AOD over land proposed by Xue et al in 2017 [1]. We obtain 0.64μm band surface reflectance by utilizing a linear relationship between the surface reflectance at the wavelength of3.75μm and 0.64μm, which has been verified in the Moderate Resolution Imaging Spectroradiometer (MODIS). Considering difference of spectral response between AVHRR's bands and MODIS's bands, we calibrate this empirical relationship by fitting the simulated surface reflectance of the relative AVHRR and MODIS bands. A radiative transfer model for Lambertian surface and the look-up table (LUT) method are applied to NOAA-7, 9,11,14, 18 over China mainland (15° - 60° N, 70° - 140° E) from 1982 to 2011. Comparison of retrieval AVHRR AOD against AErosol RObotic NETwork (AERONET) data shows good consistency with more than 60% points within uncertainty of$\pm$(0.05+0.25xAOD).

In this paper, PM2.5 and PM10 in mainland China were estimated by using the Geographically and Temporally Weighted Regression model and FY-4 AOD data. Based on the GTWR model, the PM was estimated by BLH, RH, time, space and AOD. Taking June 2, 2019 as an example, the feasibility of FY-4 data in estimating PM2.5 and PM10 mass concentrations was analyzed and confirmed.

In this paper, through the analysis and comparison of three commonly used atmospheric environmental capacity estimation methods, we find that the existing methods have many limitations in the aspects of data base and the factors considered. Taking Xuzhou City in Jiangsu Province as an example, based on the air pollution multi-source model, environmental impact assessment and primary pollutant simulation are carried out by using remote sensing data, ground monitoring station data, pollution emission inventory data and meteorological data. In addition, the iterative algorithm of multi-pollutant environmental capacity with the joint constraint of$\text{PM}_{2.5}$and$\mathrm{O}_{3}$is established to recalculate the urban atmospheric environmental capacity.