Improving Accuracy and Fluency: Recent Developments in Machine Translation

EasyChair Preprint 12207, version 1

Abstract

Machine Translation (MT) has witnessed remarkable progress in recent years, driven by advancements in neural network architectures, training techniques, and data augmentation strategies. This abstract provides an overview of the latest developments aimed at improving the accuracy and fluency of machine translation systems. Furthermore, data augmentation strategies, including back-translation and data synthesis, have been instrumental in addressing the issue of data scarcity for low-resource languages. Back-translation involves generating synthetic parallel data by translating monolingual corpora, while data synthesis techniques create diverse training examples through paraphrasing and textual manipulation. These approaches have significantly improved the robustness and fluency of MT systems, particularly for underrepresented languages.

Keyphrases: architectures, network, neural

@booklet{EasyChair:12207,
  author    = {Kurez Oroy and Robert Thomas},
  title     = {Improving Accuracy and Fluency: Recent Developments in Machine Translation},
  howpublished = {EasyChair Preprint 12207},
  year      = {EasyChair, 2024}}