More Like this

1. META: Metadata-Empowered Weak Supervision for Text Classification

   https://doi.org/10.18653/v1/2020.emnlp-main.670  

   Mekala, Dheeraj; Zhang, Xinyang; Shang, Jingbo (January 2020, Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP))

   null (Ed.)

   Recent advances in weakly supervised learn- ing enable training high-quality text classifiers by only providing a few user-provided seed words. Existing methods mainly use text data alone to generate pseudo-labels despite the fact that metadata information (e.g., author and timestamp) is widely available across various domains. Strong label indicators exist in the metadata and it has been long overlooked mainly due to the following challenges: (1) metadata is multi-typed, requiring systematic modeling of different types and their combinations, (2) metadata is noisy, some metadata entities (e.g., authors, venues) are more compelling label indicators than others. In this paper, we propose a novel framework, META, which goes beyond the existing paradigm and leverages metadata as an additional source of weak supervision. Specifically, we organize the text data and metadata together into a text-rich network and adopt network motifs to capture appropriate combinations of metadata. Based on seed words, we rank and filter motif instances to distill highly label-indicative ones as “seed motifs”, which provide additional weak supervision. Following a boot-strapping manner, we train the classifier and expand the seed words and seed motifs iteratively. Extensive experiments and case studies on real-world datasets demonstrate superior performance and significant advantages of leveraging metadata as weak supervision. 

   more » « less
2. Biodiversity Image Quality Metadata Augments Convolutional Neural Network Classification of Fish Species

   Leipzig, J.; Bakis, Y; Wang, X; Elhamod, M.; Diamond, K.; Dahdul, W; Karpante, A; Maga, M; Mabee, P; Bart, H; et al (March 2021, Metadata and Semantic Research. MTSR 2020. Communications in Computer and Information Science)

   null; null (Ed.)

   Biodiversity image repositories are crucial sources of training data for machine learning approaches to biological research. Metadata, specifically metadata about object quality, is putatively an important prerequisite to selecting sample subsets for these experiments. This study demonstrates the importance of image quality metadata to a species classification experiment involving a corpus of 1935 fish specimen images which were annotated with 22 metadata quality properties. A small subset of high quality images produced an F1 accuracy of 0.41 compared to 0.35 for a taxonomically matched subset of low quality images when used by a convolutional neural network approach to species identification. Using the full corpus of images revealed that image quality differed between correctly classified and misclassified images. We found the visibility of all anatomical features was the most important quality feature for classification accuracy. We suggest biodiversity image repositories consider adopting a minimal set of image quality metadata to support future machine learning projects. 

   more » « less
3. Biodiversity Image Quality Metadata Augments Convolutional Neural Network Classification of Fish Species

   https://doi.org/10.1007/978-3-030-71903-6_1  

   Leipzig, J (March 2021, MTSR 2020. Communications in Computer and Information Science)

   Garoufallou E., Ovalle-Perandones MA. (Ed.)

   Biodiversity image repositories are crucial sources for training machine learning approaches to support biological research. Metadata about object (e.g. image) quality is a putatively important prerequisite to selecting samples for these experiments. This paper reports on a study demonstrating the importance of image quality metadata for a species classification experiment involving a corpus of 1935 fish specimen images which were annotated with 22 metadata quality properties. A small subset of high quality images produced an F1 accuracy of 0.41 compared to 0.35 for a taxonomically matched subset low quality images when used by a convolutional neural network approach to species identification. Using the full corpus of images revealed that image quality differed between correctly classified and misclassified images. We found anatomical feature visibility was the most important quality feature for classification accuracy. We suggest biodiversity image repositories consider adopting a minimal set of image quality metadata to support machine learning. 

   more » « less
4. Understanding iOS Privacy Nutrition Labels: An Exploratory Large-Scale Analysis of App Store Data

   https://doi.org/10.1145/3491101.3519739  

   Li, Yucheng; Chen, Deyuan; Li, Tianshi; Agarwal, Yuvraj; Cranor, Lorrie Faith; Hong, Jason I. (April 2022, SIGCHI Conference: Human Factors in Computing Systems)

   Since December 2020, the Apple App Store has required all developers to create a privacy label when submitting new apps or app updates. However, there has not been a comprehensive study on how developers responded to this requirement. We present the frst measurement study of Apple privacy nutrition labels to understand how apps on the U.S. App Store create and update privacy labels. We collected weekly snapshots of the privacy label and other metadata for all the 1.4 million apps on the U.S. App Store from April 2 to November 5, 2021. Our analysis showed that 51.6% of apps still do not have a privacy label as of November 5, 2021. Although 35.3% of old apps have created a privacy label, only 2.7% of old apps created a privacy label without app updates (i.e., voluntary adoption). Our findings suggest that inactive apps have little incentive to create privacy labels. 

   more » « less
5. A Unified Framework for Post-disaster Hazard and Structural Assessment Data Collection Across Hazards and Infrastructure Typologies

   Alam, Mohammad; Kijewski-Correa, Tracy; Roueche, David; Mosalam, Khalid; Prevatt, David; Robertson, Ian (May 2025, Frontiers in built environment)

   Post-disaster field observations of the built environment are critical for advancing fundamental research that links hazard data to structural performance, cascading community impacts, and the development of effective mitigation strategies. Yet, data collection remain fragmented across hazard types and infrastructure systems due to varying objectives, methodologies, protocols, and standards among investigators and organizations. To address this, a Unified Assessment Framework has been developed for standardized post-disaster hazard and structural assessment data and metadata collection across multiple natural hazards (earthquake, windstorm, coastal events) and infrastructure typologies. The framework encompasses a tiered performance assessment structure with increasing rigor and fidelity levels: Basic Assessment (BA), Load Path Assessment (LPA), and Detailed Component Assessment (DCA). The framework has been implemented as an open-access mobile application, the Structural Extreme Events Reconnaissance (StEER) Network’s StEER Unified App, hosted on Fulcrum data collection platform . Along with unification of data fields, preliminary mapping rules were developed to map out existing hazard-specific damage rating scales (e.g., wind, surge/flooding, rainwater ingress) to the European Macroseismic Scale (EMS-98) compatible unified damage scale, enabling consolidation of global damage ratings into a common data field, facilitating the unification of multiple hazards within a single app. In the mapping of damage ratings, overarching level definitions were retained (e.g., slight, moderate, severe damage) while customizing the specific descriptors to reflect hazard-specific damage mechanisms. Two use cases are presented to demonstrate the application of this framework through the StEER Unified App: a supervised pilot after the 2022 Hurricane Ian, Florida and an unsupervised deployment for the 2023 Turkey earthquake sequence. These deployments illustrate the framework’s flexibility and scalability, validate the feasibility of standardized assessments, and offer insights into how data quality is influenced by assessor pre-deployment training and assessment tier—particularly for complex tasks such as load path evaluation. This work advances the field by providing a scalable, standardized, and hazard-agnostic approach to structural field reconnaissance. The open-access framework and app support real-time deployments and enable integration of legacy datasets into a unified platform—laying the foundation for longitudinal analyses, cross-hazard comparisons, and expanded data reuse in the Natural Hazards Engineering community. 

   more » « less