118 research outputs found
Author Identification in Free Texts
No full textInformation Extraction is a popular topic in the Natural Language Processing area. This thesis focuses on author identi cation in free text. This study divided the author identi cation task into two subtask, quotation extraction and speaker attribution. The entire system contains two parts, a rule based model for quotation extraction and a machine learning model for speaker attribution. The resource domain used in this thesis is the literary narrative. There is also a generalisation test on the news domain. The results of the experiment show that the rule based model can achieve a 0.88 F-score on quotation extraction and the best result of a machine learning model is 85.7% accuracy. The overall test on the entire system returns 77.9% accuracy on the literary source domain and 73.6% on the news domain
Author Identification in Free Texts
No full textInformation Extraction is a popular topic in the Natural Language Processing area. This thesis focuses on author identi cation in free text. This study divided the author identi cation task into two subtask, quotation extraction and speaker attribution. The entire system contains two parts, a rule based model for quotation extraction and a machine learning model for speaker attribution. The resource domain used in this thesis is the literary narrative. There is also a generalisation test on the news domain. The results of the experiment show that the rule based model can achieve a 0.88 F-score on quotation extraction and the best result of a machine learning model is 85.7% accuracy. The overall test on the entire system returns 77.9% accuracy on the literary source domain and 73.6% on the news domain
Correction: Yahui, Che et al. Validation of Aerosol Products from AATSR and MERIS/AATSR Synergy Algorithms—Part 1: Global Evaluation. Remote Sensing 2018, 10, 1414
No full textThe authors wish to make the following corrections to this paper [...
Assessment of a revised dust prediction model for Mildura, Australia
No full textDust storms are regarded as global natural hazards, adversely impacting the climate, economy, and human health. As a dry continent, Australia is extensively impacted by dust storm activity, particularly in dry seasons. Located in the Mallee region and 600 km downwind of the Lake Eyre Basin, Mildura is one of the most vulnerable regional cities to dust storms. Rapid development of agriculture in the Mildura region removed natural vegetation and increased the frequency of dust storms in the 20th century. To better understand the factors and processes that affect dust storm activity in Mildura, a seasonal predictive model for dust event days was developed in the early 1990s. This was based on an empirical relationship between seasonal rainfall in preceding autumn and summer dust event days (the most active dust season in Mildura). In this study, this model was applied for a further 24-year period (from 1990 to 2013) to test model validity for forecasting dust activity. Results show that the r2 was 0.13 and the root mean square error was 5.33 days in the ‘forecast’ mode, which indicates poorer model performance than that for the original calibration period (1960–1989). All large ‘forecast’ errors occurred in the 1990s. Winter rainfall was identified as the main climate factor for overprediction. The effect of the preceding winter rainfall on summer dust event occurrence was found to increase with the ratio of winter rainfall over autumn rainfall for the whole period of 1960–2013. An updated dust prediction model for 1960–2013 was constructed based on preceding autumn and winter rainfall. Autumn rainfall was used as the predictor when the ratio of winter and autumn rainfall was no more than 3.1; otherwise, winter rainfall was used. This was a marked improvement in model performance with an r2 value of 0.37 to that of 0.26 for the original model performance for the period as a whole (1960–2013).Full Tex
Dust activities in Australia: their severity and spatiotemporal distribution using data from multiple sources
No full textDust storms are meteorological events common in arid and semi-arid regions when strong winds entrain loose sediments and soil particles from a dry surface. As the driest inhabited continent in the world, dust storms occur regularly in Australia with adverse impacts on agriculture, public health, and the economy. The severity and frequency of occurrence of dust storms are often associated with prolonged periods of low rainfall and poor vegetation cover. Unlike water erosion, the source area of wind erosion and dust storms is diffuse, their pathways not well defined, with fine dust particles capable of traveling long distances over land and across the sea. It is therefore challenging and, at the same time, of great importance to be able to predict the occurrence and severity (or intensity) of dust storms so that the government and wider community can take appropriate actions to mitigate their adverse impacts. With the rapid development of satellite technology and general circulation models (GCMs) in recent years, remote sensing data and reanalysis have increasingly been used for temporal and spatial analysis and prediction of dust activities. There has been no systematic attempt to quantitatively compare remote sensing and GCM dust outputs, analyze the temporal and spatial distribution of dust activities using multiple data sets, and predict a time series of dust storm activity in Australia. Therefore, the overall aim of this Ph.D. thesis is to identify spatial and temporal variations of dust severity with selected high-accuracy remote sensing and GCM data and to build a predictive dust model for selected sites in eastern Australia based on datasets from multiple sources. [...]Thesis (PhD Doctorate)Doctor of Philosophy (PhD)School of Eng & Built EnvScience, Environment, Engineering and TechnologyFull Tex
Dynamic load-sharing characteristic analysis of face gear power-split gear system based on tooth contact characteristics
No full textConcrete Crack images for segmentation
No full textThis is a concrete crack dataset for segmentation. It is partially from Ozgenel FÇ. Concrete crack segmentation dataset. Mendeley Data 2019; 1: DOI: 10.17632/jwsn7tfbrp.1. and @article{liu2019deepcrack, title={DeepCrack: A Deep Hierarchical Feature Learning Architecture for Crack Segmentation}, author={Liu, Yahui and Yao, Jian and Lu, Xiaohu and Xie, Renping and Li, Li}, journal={Neurocomputing}, volume={338}, pages={139--153}, year={2019}, doi={10.1016/j.neucom.2019.01.036}
Evaluation and comparison of MERRA-2 AOD and DAOD with MODIS DeepBlue and AERONET data in Australia
No full textValidated dust and aerosol datasets provide valuable information for dust and other atmospheric research. A reanalysis product called ‘MERRA-2’ developed by NASA's Global Modeling and Assimilation Office (GMAO) in recent years provides long-term, high frequency, global coverage aerosol simulations, in terms of AOD (aerosol optical depth) and dust AOD (DAOD). The AOD assimilation system including AERONET (Aerosol Robotic Network) as constraints has led to difficulties with independent validation of MERRA-2 AOD and DAOD. The existing validation and evaluation of MERRA-2 tend to focus on central inland Australia, and comparison of DAOD with MACC (Monitoring Atmospheric Composition and Climate) is insufficient to indicate which one is more appropriate for dust research in Australia. In this study, MERRA-2 output was compared with MODIS DeepBlue (MODIS-DB) (2002–2020) and AERONET (1998–2020) AOD and DAOD in Australia. The expected error (EE) for MERRA-2 AOD was estimated to be ±(0.03+0.15τA) containing 66% of data points. MERRA-2 is less well correlated with AERONET in terms of DAOD, but with a tighter distribution of data points (78%) within the EE. The EE for MODIS-DB AOD was found to be the same and the correlation between MODIS-DB DAOD and AERONET DAOD is stronger than that between MERRA-2 DAOD and AERONET DAOD. Comparison of MERRA-2 and MODIS-DB AOD showed that MERRA-2 AOD was 22.5% higher than MODIS-DB AOD at AERONET sites on average. MERRA-2 AOD was generally lower than MODIS-DB AOD away from AERONET sites when AOD >0.1, and the spatial distribution of the difference between the two is consistent with the spatial variation in AOD. Large differences between MODIS-DB AOD and MERRA-2 occurred in areas where AOD is greater than 0.3. AOD and DAOD based on MODIS-DB could be used for areas away from AERONET sites in Australia, as they are less dependent on AERONET data compared to MERRA-2. When compared to MODIS brightness temperature difference (BTD) for two known severe storm events in Australia, MODIS-DB product is better than MERRA-2 in terms of detecting the spatial extent of dust storms and estimating the storm intensity. MERRA-2 would underestimate the magnitude and spatial extent of AOD, especially for thick dust plumes, and AERONET was likely to have misclassified dust as cloud.No Full Tex
Analysis of Dust Detection Algorithms Based on FY-4A Satellite Data
No full textDust detection is essential for environmental protection, climate change assessment, and human health issues. Based on the Fengyun-4A (FY-4A)/Advance Geostationary Radiation Imager (AGRI) images, this paper aimed to examine the performances of two classic dust detection algorithms (i.e., the brightness temperature difference (BTD) and normalized difference dust index (NDDI) thresholding algorithms) as well as two dust products (i.e., the infrared differential dust index (IDDI) and Dust Score products (DST) developed by the China Meteorological Administration). Results show that a threshold below −0.4 for BTD (11–12 µm) is appropriate for dust identification over China and that there is no fixed threshold for NDDI due to its limitations in distinguishing dust from bare ground. The IDDI and DST products presented similar results, where they are capable of detecting dust over all study areas only for daytime. A validation of these four dust detection algorithms has also been conducted with ground-based particulate matter (PM10) concentration measurements for the spring (March to May) of 2021. Results show that the average probability of correct detection (POCD) for BTD, NDDI, IDDI, and DST were 56.15%, 39.39%, 48.22%, and 46.75%, respectively. Overall, BTD performed the best on dust detection over China with its relative higher accuracy followed by IDDI and DST in the spring of 2021. A single threshold for NDDI led to a lower accuracy than those for others. Additionally, we integrated the BTD and IDDI algorithms for verification. The POFD after integration was only 56.17%, and the fusion algorithm had certain advantages over the single algorithm verification
Prediction of Total Imperviousness from Population Density and Land Use Data for Urban Areas (Case Study: South East Queensland, Australia)
No full textTotal imperviousness (residential and non-residential) increases with population growth in many regions around the world. Population density has been used to predict the total imperviousness in large areas, although population size was only closely related to residential imperviousness. In this study, population density together with land use data for 154 suburbs in Southeast Queensland (SEQ) of Australia were used to develop a new model for total imperviousness estimation. Total imperviousness was extracted through linear spectral mixing analysis (LSMA) using Landsat 8 OLI/TIRS, and then separated into residential and non-residential areas based on land use data for each suburb. Regression models were developed between population density and total imperviousness, and population density and residential imperviousness. Results show that (1) LSMA approach could retrieve imperviousness accurately (RMSE 2 = 0.77) than that between population density and total imperviousness (R2 = 0.52); (4) the new model was used to predict the total imperiousness based on population density projections to 2057 for three potential urban development areas in SEQ. This research allows accurate prediction of the total impervious area from population density and service area per capital for other regions in the world
- …
