Search Constraints
1 - 5 of 5
Number of results to display per page
View results as:
Search Results
-
- Creator:
- Budak, Ceren, Goel, Sharad, and Rao, Justin M
- Description:
- Our primary analysis is based on articles published in 2013 by the top thirteen US news outlets and two popular political blogs. To compile the set of articles published by these outlets, we first examined the complete web-browsing records for US-located users who installed the Bing Toolbar, an optional add-on application for the Internet Explorer web browser. For each of the fifteen news sites, we recorded all unique URLs that were visited by at least ten toolbar users, and we then crawled the news sites to obtain the full article title and text. This process resulted in a corpus of 803,146 articles published on the fifteen news sites over the course of a year, with each article annotated with its relative popularity. , Next, we built two binary classifiers using large-scale logistic regression. The first classifier—which we refer to as the news classifier —identifies “news” articles (i.e., articles that would typically appear in the front section of a traditional newspaper). The second classifier—the politics classifier —identifies political news from the subset of articles identified as news by the first classifier. 340,191 (42 percent) were classified as news. On the set of 340,191 news articles, 114,814 (34 percent) were classified as political. , Having identified approximately 115,000 political news articles, we next seek to categorize the articles by topic (e.g., gay rights, healthcare, etc.), and to quantify the political slant of the article. To do so, we turn to human judges recruited via Mechanical Turk to analyze the articles. For every day in 2013, we randomly selected two political articles, when available, from each of the 15 outlets we study, with sampling weights equal to the number of times the article was visited by our panel of toolbar users., Amazon Mechanical Turk Labeling task: To detect and control for possible preconceptions of an outlet’s ideological slant, workers, upon first entering the experiment, were randomly assigned to either a blinded or unblinded condition. In the blinded condition, workers were presented with only the article’s title and text, whereas in the unblinded condition, they were additionally shown the name of the outlet in which the article was published. Each article was then analyzed by two workers, one each from the sets of workers in the two conditions. For each article, each worker completed the following three tasks. First, they provided primary and secondary article classifications from a list of fifteen topics: (1) civil rights; (2) Democrat scandals; (3) drugs; (4) economy; (5) education; (6) elections; (7) environment; (8) gay rights; (9) gun-related crimes; (10) gun rights/regulation; (11) healthcare; (12) international news; (13) national security; (14) Republican scandals; and (15) other. , and Second, workers determined whether the article was descriptive news or opinion. Third, to measure ideological slant, workers were asked, “Is the article generally positive, neutral, or negative toward members of the Democratic Party?” and separately, “Is the article generally positive, neutral, or negative toward members of the Republican Party?” Choices for these last two questions were provided on a five-point scale: very positive, somewhat positive, neutral, somewhat negative, and very negative. To mitigate question-ordering effects, workers were initially randomly assigned to being asked either the Democratic or Republican party question first; the question order remained the same for any subsequent articles the worker rated. Finally, we assigned each article a partisanship score between –1 and 1, where a negative rating indicates that the article is net left-leaning and a positive rating indicates that it is net right-leaning. Specifically, for an article’s depiction of the Democratic Party, the five-point scale from very positive to very negative is encoded as –1, –0.5, 0, 0.5, 1. Analogously, for an article’s depiction of the Republican Party, the scale is encoded as 1, 0.5, 0, –.0.5, –1. The score for each article is defined as the average over these two ratings. Thus, an average score of –1, for example, indicates that the article is very positive toward Democrats and very negative toward Republicans. The result of this procedure is a large, representative sample of political news articles, with direct human judgments on partisanship and article topic.
- Keyword:
- news media, media bias, crowdsourcing, and machine learning
- Citation to related publication:
- https://academic.oup.com/poq/article-abstract/80/S1/250/2223443/?redirectedFrom=fulltext and Ceren Budak, Sharad Goel, Justin M. Rao, Fair and Balanced? Quantifying Media Bias through Crowdsourced Content Analysis, Public Opinion Quarterly, Volume 80, Issue S1, 2016, Pages 250–271, https://doi.org/10.1093/poq/nfw007
- Discipline:
- Social Sciences
-
- Creator:
- Sheppard, Anja, Sethuraman, Advaith V, Bagoren, Onur, Pinnow, Christopher, Anderson, Jamey, Havens, Timothy C, and Skinner, Katherine A
- Description:
- The AI4Shipwrecks dataset contains sidescan sonar images of shipwrecks and corresponding binary labels collected during 2022 and 2023 at the NOAA Thunder Bay National Marine Sanctuary in Alpena, MI. The data collection platform was an Iver3 Autonomous Underwater Vehicle (AUV) equipped with an EdgeTech 2205 dual-frequency ultra-high resolution sidescan sonar and 3D bathymetric system. The labels were compiled from reference labels created by experts in marine archaeology. The intended use of this dataset is to encourage development of semantic segmentation, object detection, or anomaly detection algorithms in the computer vision field. Comparisons of state-of-the-art segmentation networks on our dataset are shown in the paper. , The file structure is organized as described in the README.txt file, where images in 'images' directories are the waterfall product of sidescan sonar surveys, and images in 'labels' directories are binary representations of expert labels. Images across the 'images' and 'labels' directories are correlated by having identical filenames. In the label images, a pixel value of '0' represents the non-shipwreck/other class and '1' represents the shipwreck class for the correspondingly named image (<wreck_name>_<##>.png) in the images directory. , and The project webpage can be found at: https://umfieldrobotics.github.io/ai4shipwrecks/
- Keyword:
- machine learning, computer vision, field robotics, marine robotics, underwater robotics, sidescan sonar, semantic segmentation, and object detection
- Discipline:
- Engineering
-
- Creator:
- Swiger, Brian M., Liemohn, Michael W., and Ganushkina, Natalia Y.
- Description:
- We sampled the near-Earth plasma sheet using data from the NASA Time History of Events and Macroscale Interactions During Substorms mission. For the observations of the plasma sheet, we used corresponding interplanetary observations using the OMNI database. We used these data to develop a data-driven model that predicts plasma sheet electron flux from upstream solar wind variations. The model output data are included in this work, along with code for analyzing the model performance and producing figures used in the related publication. and Data files are included in hdf5 and Python pickle binary formats; scripts included are set up for use of Python 3 to access and process the pickle binary format data.
- Keyword:
- neural network, plasma sheet, solar wind, machine learning, keV electron flux, deep learning, and space weather
- Citation to related publication:
- Swiger, B. M., Liemohn, M. W., & Ganushkina, N. Y. (2020). Improvement of Plasma Sheet Neural Network Accuracy With Inclusion of Physical Information. Frontiers in Astronomy and Space Sciences, 7. https://doi.org/10.3389/fspas.2020.00042
- Discipline:
- Science and Engineering
-
- Creator:
- Whitaker, Steven T., Nataraj, Gopal, Nielsen, Jon-Fredrik, and Fessler, Jeffrey A.
- Description:
- File: P,jf06Sep2019,mese.7 The multi-echo spin echo (MESE) data was acquired using a 3D acquisition with an initial 90 degree excitation pulse followed by 32 refocusing (180 degree) pulses, resulting in 32 echoes with echo spacing of 10 ms. The repetition time of the sequence was 1200 ms. Each refocusing pulse was flanked by crusher gradients to impart 14 cycles of phase across the imaging volume. The initial excitation pulse had time-bandwidth product of 6, duration of 3 ms, and slab thickness of 0.9 cm, and each refocusing pulse had time-bandwidth product of 2, duration of 2 ms, and slab thickness of 2.1 cm. The scan took 36 min 11 s and covered a field of view (FOV) of 22 x 22 x 0.99 cm^3 with matrix size 200 x 200 x 9., File: P,jf06Sep2019,b1.7 The Bloch-Siegert (BS) scans were acquired using a 3D acquisition. The excitation pulse of these scans had time-bandwidth product of 8 and duration of 1 ms. The pair of scans used +/-4 kHz off-resonant Fermi pulses between excitation and readout. The BS scans took 2 min 40 s in total and covered a FOV of 22 x 22 x 0.99 cm^3 with matrix size 200 x 50 x 9., File: P,jf06Sep2019,mwf.7 The small-tip fast recovery (STFR) scans were acquired using a 3D acquisition. The first two and last two scans were pairs of spoiled gradient-recalled echo (SPGR) scans with echo time difference of 2.3 ms. (In the related paper, only the first set was used, i.e., only 11 of the 13 scans were used.) The remaining scans used scan parameters that were optimized to minimize the Cramer-Rao Lower Bound (CRLB) of estimates of myelin water fraction (MWF). The RF pulses had time-bandwidth product of 8 and duration of 1 ms. Each pair of SPGR scans took 58 s and the nine STFR scans took 3 min 36 s for a total scan time of 5 min 32 s (4 min 34 s if one pair of SPGR scans is ignored). The scans covered a field of view (FOV) of 22 x 22 x 0.99 cm^3 with matrix size 200 x 200 x 9., File: meseslice5.mat Contains the 32 echoes of the MESE image data for the middle slice of the imaging volume. Saved using Mathworks MATLAB R2019a., File: b1slice5.mat Contains the transmit field inhomogeneity map for the middle slice of the imaging volume., File: recon.jld Key "img" contains the 11 STFR images for the middle slice of the imaging volume. Key "b0map" contains a field map estimated from the two SPGR scans. Key "mask" contains a mask of the voxels for which to estimate MWF. Key "T1img" contains a T1-weighted image for anatomical reference., File: headmask.mat Contains a mask for visualizing just the brain (ignores the skull) for the middle slice of the imaging volume., File: rois.mat Contains masks for various regions of interest (ROIs), used for computing statistics. Keys "mtopleft", "mtopright", "mbottomleft", and "mbottomright" refer to the corresponding locations on the anatomical reference image (see related paper). Key "mic" refers to the internal capsules, and key "mgm" refers to a gray matter ROI., The raw data files (P-files) can be read into the Julia programming language using the Julia version of the Michigan Image Reconstruction Toolbox ( https://github.com/JeffFessler/MIRT.jl) or into MATLAB using TOPPE ( https://github.com/toppeMRI/toppe). The reconstructed slices used in the related paper are provided for convenience, and are stored in .mat files that can be loaded into Julia (using the package MAT.jl) or MATLAB, and a .jld file that can be loaded into Julia (using the package JLD.jl). The Julia code for processing the data to create MWF maps is hosted publicly on GitHub at https://github.com/StevenWhitaker/STFR-MWF., and Files: toppe-master.zip and MIRT.jl-master.zip are archived versions of the TOPPE and Michigan Image Reconstruction Toolbox code sets from GitHub as of 2/28/2020.
- Keyword:
- myelin, machine learning, kernel learning, magnetic resonance imaging, and scan design
- Citation to related publication:
- Whitaker, S. T., Nataraj, G., Nielsen, J.-F., & Fessler, J. A. (2020). Myelin water fraction estimation using small-tip fast recovery MRI. Magnetic Resonance in Medicine, 84(4), 1977–1990. https://doi.org/10.1002/mrm.28259
- Discipline:
- Health Sciences and Engineering
-
- Creator:
- Chen, Yang and Manchester, Ward IV
- Description:
- GOES_flare_list: contains a list of more than 10,000 flare events. The list has 6 columns, flare classification, active region number, date, start time end time, emission peak time, GOES_B_flare_list: contains time series data of SDO/HMI SHARP parameters for B class solar flares , GOES_MX_flare_list: contains time series data of SDO/HMI SHARP parameters for M and X class solar flares, SHARP_B_flare_data_300.hdf5 and SHARP_MX_flare_data_300.hdf5 files contain time series more than 20 physical variables derived from the SDO/HMI SHARP data files. These data are saved at a 12 minute cadence and are used to train the LSTM model., and B_HARPs_CNNencoded_part_xxx.hdf5 and M_X HARPs_CNNencoded_part_xxx.hdf5 include neural network encoded features derived from vector magnetogram images derived from the Solar Dynamics Observatory (SDO) Helioseismic and Magnetic Imager (HMI). These data files typically contains one or two sequences of magnetograms covering an active region for a period of 24h with a 1 hour cadence. We encode each magnetogram with frames of a fixed size of 8x16 with 512 channels.
- Keyword:
- machine learning, data science, and solar flare prediction
- Citation to related publication:
- Chen, Y., Manchester, W., Hero, A., Toth, G., DuFumier, B. Zhou, T., Wang, X., Zhu, H., Sun, Zeyu, Gombosi, T., Identifying Solar Flare Precursors Using Time Series of SDO/HMI Images and SHARP Parameters, Space Weather, 17, 1404–1426. https://doi.org/10.1029/2019SW002214 and Jiao, Z., Chen, Y., Manchester, W. (2020). Data for Solar Flare Intensity Prediction with Machine Learning Models [Data set]. University of Michigan - Deep Blue. https://doi.org/10.7302/b07j-bj08
- Discipline:
- Engineering and Science