Translation-Service

A novel retrieval-augmented NMT framework that uses contrastive retrieval and hierarchical encoding to leverage diverse, non-redundant translation memories for improved translation quality.

Analysis of degeneracies in inverse propensity scoring for offline machine translation learning from deterministic logs, with proposed solutions.

An analysis comparing printed/online dictionaries and term bases, focusing on their evolution, reliability, and future in translation technology.

Analysis of the first application of Neural Machine Translation to Arabic, comparing it with phrase-based systems, exploring preprocessing effects, and evaluating robustness to domain shift.

Introducing WOKIE, an open-source pipeline for automated translation of SKOS thesauri using external services and LLM refinement to enhance accessibility and cross-lingual interoperability in Digital Humanities.

Research on enhancing LLM-based machine translation using Translation Memory prompts, achieving significant BLEU score improvements across multiple languages and domains.

Research on lightweight, local deep learning models for translating low-resource, obfuscated, and niche languages using LSTM-RNN encoder-decoder architecture.

An analysis of Machine Translation systems developed for Indian languages, covering approaches like Direct, Rule-Based, and Corpus-Based methods, key systems, and future research directions.

Analysis of using LLMs for few-shot translation of English semantic parsing datasets to train multilingual parsers, outperforming translate-train baselines across 50 languages.

Analysis of DGT-TM, a freely available translation memory covering 22 EU languages and 231 language pairs, its creation, applications in language technology, and future impact.

Analysis of a research paper proposing a novel Advantage Actor-Critic (A2C) reinforcement learning model for multimodal machine translation, integrating visual and textual data.

Analysis of a hybrid NMT-SMT framework that integrates SMT recommendations into NMT decoding to address fluency-adequacy trade-offs, with experimental results on Chinese-English translation.

An in-depth analysis of Neural Machine Translation (NMT) covering its history, core neural network concepts, encoder-decoder architectures, refinements, and current challenges.

An end-to-end deep learning framework integrating quality estimation and automatic post-editing to enhance machine translation output and reduce human translator workload.

Analysis of TM-LevT, a novel variant of the Levenshtein Transformer designed to effectively edit translations from a Translation Memory, achieving performance on par with autoregressive models.

An analysis of the EuroTermBank Toolkit, an open-source solution for managing and sharing terminology across federated databases to support translators, NLP applications, and AI systems.

Analysis of submodular function-based algorithms for optimal example selection in retrieval-augmented neural machine translation, focusing on coverage optimization.

Analysis of a novel post-training domain adaptation method for NMT, exploring incremental specialization, experimental results, and future applications.

An analysis of challenges and opportunities in legal translator training, exploring updated models, technology integration, and the evolving role of translators as intercultural mediators.

Analysis of global text augmentation methods (Word2Vec, WordNet, round-trip translation) and mixup for improving short text classification performance and model robustness.

Introduces Structure-Invariant Testing (SIT), a metamorphic testing method for validating machine translation software by analyzing structural consistency in translated outputs.

An analysis of how professional translation ethics codes apply to non-professional contexts like crowdsourcing, fan translation, and FOSS localization, highlighting differences and innovations.

An analysis of fuzzy match algorithms for Translation Memory systems, evaluating their correlation with human judgments and proposing a new weighted n-gram precision method.

Analysis of TM-augmented NMT from a probabilistic and variance-bias decomposition viewpoint, explaining performance contradictions and proposing an effective ensemble method.

Analysis of modern QA tools for translation, their integration with CAT tools, industry standards, and practical evaluation of standalone QA software outputs.

Analysis of how globalization and technology redefine translation demands, positioning translators as cultural mediators and strategic business assets.

Analysis of a variational encoder-decoder model for NMT that introduces continuous latent variables to explicitly model underlying semantics, improving translation quality.

Analysis of SM2, a streaming Transformer Transducer model for multilingual ASR and speech translation, featuring truly zero-shot capability and weak supervision.