Aug 2026· Advanced Electromagnetics· Vol 15, pp. 849-861· 0 citations
TL;DR
A GAT-BiLSTM fusion model that integrates dependency syntactic analysis with graph attention mechanisms and bidirectional long short-term memory networks improves translation fidelity and semantic consistency while exhibiting strong robustness for structurally complex inputs.
Abstract
Accurate semantic analysis and translation of complex English sentences are essential for intelligent information interaction and multilingual communication in modern digital systems, including semantic communication frameworks and electromagnetic-enabled intelligent networks. To address semantic omissions and logical inconsistencies caused by long-distance dependencies and referential ambiguity, this study proposes a GAT-BiLSTM fusion model that integrates dependency syntactic analysis with graph attention mechanisms and bidirectional long short-term memory networks. A lightweight semantic graph is first constructed to capture structural dependencies, after which graph representations are adaptively fused with contextual features through a gating mechanism to obtain unified semantic embeddings. During decoding, semantic gating and multi-head attention collaboratively enhance contextual coherence and semantic alignment. Experimental results demonstrate that the proposed model achieves a BLEU score exceeding 68.7, subject and action semantic matching scores of 0.88 and 0.84, respectively, and a syntactic structure retention rate of 72% for complex sentences. The proposed framework effectively improves translation fidelity and semantic consistency while exhibiting strong robustness for structurally complex inputs. Furthermore, the semantic modeling strategy provides methodological support for multilingual information processing, semantic communication, and intelligent human– machine interaction in electromagnetic wave propagation and wireless communication environments.
Chinese English neural machine translation remains challenging due to substantial syntactic divergence, word-order variation, lexical ambiguity, and cross-lingual semantic mismatch. These challenges often lead to semantic omissions, over-translation, and weak source–target semantic alignment in transformer-based translation systems. Although transformer architecture has achieved remarkable progress, they frequently exhibit limitations in capturing complementary lexical and sentence-level semantic information while offering limited interpretability of the translation process. To address these challenges, this paper proposes DSF–MarianMT, a semantic fusion enhanced neural machine translation framework built upon MarianMT. The proposed framework integrates word-level and sentence-level semantic representations through a dynamic semantic fusion mechanism and employs a contrastive semantic learning objective to improve source–target semantic consistency during training. Experimental evaluation on a Chinese English translation dataset demonstrates the effectiveness of the proposed approach, achieving 36.5 BLEU, 60.2 chrF, 39.8 TER, and a COMET score of 0.78, outperforming the standard MarianMT baseline across all evaluation metrics. In addition to improved translation quality, interpretability analyses reveal reduced attention entropy, lower redundancies among attention heads, and stable token-level semantic learning. Furthermore, error analysis indicates fewer semantic omissions and over-translation errors, while consistent performance is maintained across sentences of varying lengths. These findings demonstrate that the proposed framework effectively enhances semantic representation learning and translation fidelity for Chinese–English neural machine translation.
A robust WSD model that integrates a bidirectional long shortterm memory network (BiLSTM) with an attention mechanism, specifically designed for Chinese patent texts is proposed, providing reliable support for knowledge mining and intelligent text processing in technical domains.
L. X. Gao, T. Dong, M. H. Yang· Advanced Electromagnetics· 0 citations
Accurate semantic alignment and comprehensive multilingual coverage remain major challenges in constructing Chinese–Japanese–English translation databases for technical knowledge sharing and engineering information exchange. This study proposes a cross-lingual translation database construction framework based on Transformerbased neural networks and hierarchical contrastive representation learning. Using InfoXLM as the shared semantic encoder, dependency-syntax perturbation and cross-lingual lexical substitution are employed to generate codeswitching hard negative samples, while a three-way triplet contrastive optimization strategy jointly constrains Chinese, Japanese, and English semantic spaces to improve representation consistency. The fine-tuned model is further integrated with semantic similarity filtering and bidirectional consistency verification to construct a large-scale trilingual translation database from multilingual candidate corpora. Experimental results demonstrate an alignment precision of 94.7%, a recall of 92.3%, an F1-score of 93.5%, and a translation coverage of 89.8%, significantly outperforming conventional multilingual alignment approaches. The proposed framework provides an efficient solution for multilingual engineering knowledge organization, technical document retrieval, and intelligent information processing, offering valuable support for cross-language electromagnetic engineering documentation, antenna technology resources, and multilingual communication systems requiring robust semantic alignment and signal-aware information management.
X. Wang, D. L. Wu· Advanced Electromagnetics· 0 citations
A novel fusion-based Japanese-English translation system is presented to address the persistent challenges posed by heterogeneous, multi-domain text data in practical engineering environments. The study proposes an algorithmic framework designed to efficiently handle formal, informal, and technical language forms. It uses rule-based symbolic processing and advanced neural network translation models. The method aims to prepare complex Japanese input translation thru strict multistage data normalization, highly featured word segmentation, and adaptive contextual annotation. The core dual-path architecture dynamically controls the contribution of symbolic and neural components thru a context-aware weighting mechanism. This helps optimize translation fidelity based on the language characteristics and domain relevance of the input text. According to a comprehensive empirical evaluation, the system performs well in high variability domains and rare/ambiguous language patterns, and outperforms traditional neural network and symbolic baseline models in terms of efficiency and accuracy. Quantitative results show that the BLEU's appropriateness indicators, scores, and human fluency have been greatly improved. The technical design ensures industrial-grade throughput and scalability thru parallel reasoning and optimization. By combining symbolic language features and neural representation learning, the framework provides a reliable and adaptable solution to the limitations of current machine translation, and lays a technical foundation for advanced multilingual applications.
Na Qi· The 2026 International Confe...· 0 citations
Word Sense Disambiguation (WSD) is a critical task in Natural Language Processing (NLP) that aims to determine the intended meaning of ambiguous words based on their contextual usage. Although Transformer-based language models have significantly improved contextual understanding, their decision-making process often lacks semantic transparency and explainability. Conversely, knowledge-based approaches leverage lexical databases and ontological resources to provide interpretable semantic relationships but are constrained by limited adaptability to diverse linguistic contexts.
This research introduces a hybrid WSD framework that combines contextual representations generated by Transformer architectures with structured semantic knowledge extracted from ontology-driven repositories. By integrating contextual embeddings with knowledge embeddings, the proposed model enhances both contextual sensitivity and semantic consistency during the sense prediction process. The knowledge infusion mechanism enables the model to validate contextual interpretations using explicit semantic relationships, thereby improving the reliability of word sense assignments.
Experimental evaluation demonstrates that the proposed framework achieves superior performance compared with standalone contextual and knowledge-based approaches, yielding statistically significant improvements in precision, recall, and F1-score. The results indicate that combining deep contextual learning with structured semantic knowledge provides a robust and interpretable solution for accurate word sense disambiguation across diverse linguistic scenarios
Roopa H. R., P. S., Meenatchi Sundaram· THE SCIENTIFIC TEMPER· 0 citations
The proposed WILO-BiLSTM model can perform superior to the conventional approaches and its performance results in terms of METEOR, BLEU, ROUGE, and SPICE score at training data 90% is 0.28, 0.50, 0.56, and 26.93 for the SquAD dataset, respectively.
Pallavi Yevale, Nilesh Uke· Journal of Intelligent Decis...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
