Few-shot adaptation improves minority-label Bangla hate speech detection
FewShot AdaptiveMLC raised macro-F1 by 0.072 on a multi-label Bangla hate speech dataset by fine-tuning task-specific heads for minority labels ...
Tag: Natural Language Processing
Hospital study maps electronic records for earlier disease reporting
A retrospective study of 284 notifiable-disease episodes found delayed and inadequate hospital notifications while mapping electronic records that could support ...
Extra Trees leads 13-model cyberbullying benchmark on English tweets
A benchmark of 13 models on more than 47,000 English tweets found that TF-IDF-based Extra Trees led overall, while class-level ...
AI Learns to Check Its Own X-Ray Reports by Searching Past Cases Before Writing
Researchers have built a modular AI framework that improves automated chest X-ray report generation by retrieving similar past cases at ...
AI Reads Literature Like a Critic: SymbolFusion Maps Symbols and Emotions in Poetry and Prose
A new neuro-symbolic framework called SymbolFusion applies explicit literary relation constraints to graph learning, achieving state-of-the-art performance in modeling symbolic ...
How Language Models Learned to Read Spreadsheets: A New Survey Maps the Rise of Tabular AI
A new survey in Artificial Intelligence Review provides the first comprehensive map of language modeling for tabular data, tracing the ...
AI Reads Crypto Whitepapers but Still Trusts the Hype Too Much
A mixed-methods study testing large language models on token-offering whitepapers finds that structured summarization with Chain-of-Thought prompting reduces rating error, ...
DistilBERT Outperforms Perceiver in the Hunt for Fake Hotel Reviews
A comparative study finds that a fine-tuned DistilBERT pipeline reliably detects deceptive hotel reviews while a Perceiver-based pipeline collapses toward ...
How Machines Spot Real Events in the Social Media Noise
The publication of a comprehensive, domain-agnostic survey on event detection in online social networks arrives at a moment when the ...
When Every Second Counts: Text Mining Reveals What Weighs Most on Nurses’ Minds
A text mining study of 1,650 South Korean nurses reveals that concurrent task demands, patient flow, and communication load drive ...
AI Model Reads Like a Human Brain, Tracking Hidden Cognitive States Word by Word
Researchers have built a brain-inspired model that uses a large language model to trace five hidden cognitive states as readers ...
AI Engine That Reads the News and Rewrites Supply Chain Recovery Plans in Seconds
A new AI decision engine combines BERT-based reading of disruption reports with a fuzzy-controlled bat algorithm to cut supply chain ...
AI Reads Brief Clinical Notes to Flag Suicide Risk, Sometimes Outperforming Specialists
Researchers trained AI models on short Spanish first-contact clinical notes and found the algorithms detected suicidal behavior more accurately than ...
Fuzzy Logic Meets Deep Learning to Crack Handwritten Tamil and Translate It Across 50 Languages
Researchers at Vellore Institute of Technology have built a fuzzy logic-augmented deep learning pipeline that recognizes handwritten Tamil with 98.23 ...
AI Framework Compresses Thousands of MOOC Reviews Into Concise, Accurate Summaries
Researchers at BRAC University developed ClusSumm, a framework combining hierarchical clustering with adaptive transformer models that compresses thousands of MOOC ...
AI Steps Into the Emergency Room and the Disaster Zone, But Nurses Are Still Waiting
A new scoping review maps how artificial intelligence is being tested across emergency care and disaster response, and finds that ...
New AI Framework Tackles the Trickiest Problem in Reading News: Spotting Events
Researchers at Liaoning Normal University have developed MCC-ED, a multi-module AI framework that combines contextual interaction, closed-loop graph fusion, and ...
Fox-Inspired AI Search and Context-Aware Transformer Push Document Translation Coherence to New Heights
A new framework combining a discourse-aware Context-Aware Transformer with a dynamically perturbed fox-inspired optimization algorithm achieves significant gains in coherent ...
Hidden Messages in AI Text: New Steganography Method Boosts Capacity Twentyfold
A new framework called TE-CSA aggregates embedding opportunities across the entire denoising trajectory of diffusion language models, boosting hidden-message capacity ...
AI Learns to Summarize Conversations Without Annotated Examples Using Bridge Domains
Researchers have developed an unsupervised domain adaptation method that lets dialogue summarization models learn from annotated news data and bridge ...
POPULAR NEWS
-
CRISPR Beta-Cell Regeneration Shows Early Promise for Long-Term Diabetes Management
-
Sudan Study Reveals Gaps in Kidney Stone Prevention Care
-
Machine Learning Model Predicts Pediatric Sepsis Risk Using Routine Clinical Indicators
-
ChatMPC Integrates Natural Language Processing with Predictive Control for Autonomous Driving