首页|Vellore Institute of Technology Researchers Provide New Data on Machine Translat ion (Efficient incremental training using a novel NMT-SMT hybrid framework for t ranslation of low-resource languages)
Vellore Institute of Technology Researchers Provide New Data on Machine Translat ion (Efficient incremental training using a novel NMT-SMT hybrid framework for t ranslation of low-resource languages)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
2024 OCT 03 (NewsRx)-By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-New study results on machine translati on have been published. According to news originating from Tamil Nadu, India, by NewsRx editors, the research stated, "The data-hungry statistical machine trans lation (SMT) and neural machine translation (NMT) models offer state-of-the-art results for languages with abundant data resources. However, extensive research is imperative to make these models perform equally well for low-resource languag es." Our news correspondents obtained a quote from the research from Vellore Institut e of Technology: "This paper proposes a novel approach to integrate the best fea tures of the NMT and SMT systems for improved translation performance of low-res ource English-Tamil language pair. The suboptimal NMT model trained with the sma ll parallel corpus translates the monolingual corpus and selects only the best t ranslations, to retrain itself in the next iteration. The proposed method employ s the SMT phrase-pair table to determine the best translations, based on the max imum match between the words of the phrasepair dictionary and each of the indiv idual translations. This repeating cycle of translation and retraining generates a large quasi-parallel corpus, thus making the NMT model more powerful. SMT-int egrated incremental training demonstrates a substantial difference in translatio n performance as compared to the existing approaches for incremental training. T he model is strengthened further by adopting a beam search decoding strategy to produce k best possible translations for each input sentence."
Vellore Institute of TechnologyTamil N aduIndiaAsiaEmerging TechnologiesMachine LearningMachine Translation
NETL
NSTL
