AI Networking Cookbook: Practical recipes for AI-assisted network automation and development
TL;DR
A zero-shot translation enhancement algorithm based on dual semantic decoupling and explicit path regularization that improves the robustness and controllability of zero-shot translation in low-resource and cross-linguistic scenarios.
A semantic distance augmentation (SDA) method that integrates syntactic information from constituency parse trees into the NMT encoder to optimize self-attention and achieves statistically significant improvements in translation quality over the strong baseline, without requiring additional training data or increasing model complexity.
An Efficient Dual-BERT Adversarial Network (DBAN) is proposed to improve the translation of noisy UGC by integrating contextual representation learning with adversarial training and significantly improves contextual understanding and cross-lingual semantic alignment while maintaining computational efficiency.
This paper proposes an alignment-aware multi-granularity tagging framework. First, this method uses a cross-lingual pre-trained model to encode source and target language contexts jointly while explicitly modeling-level bilingual correspondences via a learnable soft alignment layer. Second, a gated local enhancement module is introduced to dynamically fuse n-gram-level surface features on top of the Transformer high-level representation, preserving fine-grained error signals that are easily smoothed by attention mechanisms. Finally, a label-aware focus loss function is designed to alleviate the extreme imbalance between positive and negative samples. This function supports joint prediction of four types of fine-grained error labels. Results showed that on the large-scale benchmark of WMT QE Shared Task 2023, covering six language pairs, alignment-aware multi-granularity tagging achieved 78.6% in Micro-F1, 72.3% in Macro-F1, and improved Recall@Rare to 62.3%, especially excelling in low-frequency error types and significantly outperforming the baseline model.
Recent advances in large language models (LLMs) have led sign language translation (SLT), the task of converting sign-language videos into spoken-language text, to increasingly adopt LLMs as textual backbones. However, despite their strong language modeling capabilities, existing LLM-based SLT methods often undermine rather than exploit this language prior, producing disfluent translations, a failure we term language-prior degradation. Meanwhile, existing methods typically align videos and text at the sentence level, which does not ensure accurate lexical details and creates a lexical fidelity gap. To address both issues, we propose DualAnchor, a gloss-free LLM-based SLT training framework that couples two complementary anchors for linguistically fluent and visually faithful generation. Token-level Prior Anchoring (TPA) preserves the LLM's language prior by regularizing the multimodal decoder at each decoding step toward the next-token distribution of a frozen LLM conditioned on the same autoregressive prefix. Optimal Transport Alignment (OTA) improves lexical fidelity by formulating visual-textual matching as entropy-regularized partial optimal transport, with Sinkhorn optimization inducing a soft alignment between visual tokens and textual content tokens under a cosine cost. DualAnchor achieves strong overall performance on both PHOENIX-2014T and CSL-Daily. Targeted analyses attribute these gains to the complementary effects of the two anchors: TPA improves fluency, whereas OTA reduces fine-grained lexical errors.
The results show that multilingual transfer is the dominant factor in extremely low-resource Bantu translation while eliminating the need for heuristic proxy selection, and all systems fail to preserve tonal diacritics, highlighting an open challenge.
Translation across multiple forms and different languages shows limitations from problems with combining features and problems with moving meaning between languages, and these problems affect how the approach performs in settings that involve various conditions. This study aims to design a unified multimodal Transformer architecture that strengthens cross-linguistic semantic alignment and improves translation stability under heterogeneous input conditions. This study conducted initial processing that combined image and text data, allowing separate encoding of features at higher levels. It then used attention across different forms to combine meaning from multiple sources, producing a representation that showed unity. The approach also introduced space for meaning that multiple languages share at a level beyond surface forms, thereby improving the stability of the mapping between languages in the results. Training without modality and applying noise-enhancement techniques together improved this model’s robustness to incomplete and degraded input conditions that approximate practical deployment scenarios.The model was trained and evaluated on a structured English-German bilingual vision-language dataset composed of aligned image-text pairs collected under controlled experimental conditions.Experimental evaluation in fusion mode yielded BLEU scores of 39.2, METEOR scores of 31.7, and CIDEr scores of 113.8, indicating stronger semantic consistency and improved bilingual generation quality. The results of this research demonstrate that multimodal fusion and cross-language alignment mechanisms offer effective approaches to improve the accuracy and reliability of multimodal translation systems. The findings remain constrained by dataset scale and language coverage, and future research will extend validation across broader linguistic domains and more diverse multimodal scenarios.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.