A5102686105- Topic Modeling
- Natural Language Processing Techniques
- Speech and dialogue systems
- Language, Metaphor, and Cognition
- Kidney Stones and Urolithiasis Treatments
- Speech Recognition and Synthesis
- Pelvic floor disorders treatments
- Pediatric Urology and Nephrology Studies
- Multimodal Machine Learning Applications
- Advanced Graph Neural Networks
- Educational Technology and Assessment
- Recommender Systems and Techniques
North China University of Technology
2024
Zhejiang University
2024
Sir Run Run Shaw Hospital
2024
Abstract Chinese spelling correction has achieved significant progress, but critical challenges remain, especially in handling visually and phonetically similar errors within complex syntactic structures. This paper introduces a novel approach combining Long Short-Term Memory Network (LSTM)-enhanced Transformer for error detection Bidirectional Encoder Representations from Transformers (BERT)-based with dynamic adaptive weighting scheme. uses global attention mechanism to capture...
Metaphor recognition plays a crucial role in natural language understanding and semantic analysis. This paper introduces metaphor model called EGSNet (Enhanced Gloss Siamese Network). Previous research has demonstrated that the gloss of words contributes to their comprehension recognition. To leverage this information, incorporates annotations into model. Additionally, through data augmentation techniques enrich representations sentences target words, combined with effectively captures...
Abstract Background Obesity is an important risk factor for kidney stones(KS). Chinese Visceral Adiposity Index (CVAI), as a specific indicator visceral obesity in the population, can more accurately assess fat content individuals compared to (VAI). However, association between CVAI and KS has not been studied. Methods A total of 97,645 participants from health screening cohort underwent ultrasound examinations diagnosis stones, along with measurements their CVAI. Logistic regressions were...
With the rise of large-scale language models (LLMs), it is currently popular and effective to convert multimodal information into text descriptions for multi-hop question answering. However, we argue that current methods multi-modal answering still mainly face two challenges: 1) The retrieved evidence containing a large amount redundant information, inevitably leads significant drop in performance due irrelevant misleading prediction. 2) reasoning process without interpretable steps makes...