Tag: BERT

Researchers at the University of Kashan have developed GGIP2P, a modular pipeline that combines BERT-based entity recognition, pronoun resolution, spatial ...

A dual-encoder AI grammar correction model combining syntactic and semantic pathways outperformed existing systems on standard benchmarks and fixed more ...

A new systematic study shows that low-rank adaptation can match or approach full fine-tuning of transformer models across four sentiment ...

Researchers at Hanoi University of Science and Technology have developed a fuzzy-set-based hesitation score that makes multi-task AI models robust ...

A new survey in Machine Learning provides an extensive taxonomy of attention mechanisms and more than thirty Transformer variants that ...

Researchers at the Norwegian University of Science and Technology have developed two new explainability methods that integrate transformer attention weights ...

Researchers in Qinghai have developed MHCNet, a hypergraph-based multimodal AI that classifies Tibetan medicinal plants with 98.11 percent accuracy by ...

Researchers have developed EDIRL, an adaptive relation linking method that boosts multihop question answering over knowledge graphs by choosing between ...

Turkish researchers show that behavior-preserving code refactoring used to synthesize training data, combined with GraphCodeBERT embeddings, markedly improves deep learning ...

New research analyzing millions of Reddit interactions finds that negative emotions and multimedia cues are the strongest drivers of viral ...