FGraDA (Fine-Grained Domain Adaptation Dataset)

我们在机器翻译（FGraDA）中提出了一个现实世界的细粒度领域适应任务。 FGraDA 数据集包含四个信息技术子领域的中英翻译任务：自动驾驶汽车、人工智能教育、实时网络和智能手机。每个子域都配备了用于评估目的的开发集和测试集。为了更贴近实际，FGraDA没有使用任何域内双语训练数据，而是提供双语词典和wiki知识库，可以在短时间内更容易获得。We propose a real-world fine-grained domain adaptation task for machine translation, termed FGraDA. The FGraDA dataset encompasses Chinese-English translation tasks across four information technology subfields: autonomous vehicles, AI education, real-time networks, and smartphones. Each subdomain is equipped with a development set and a test set for evaluation purposes. To better align with real-world scenarios, FGraDA does not provide any in-domain bilingual training data. Instead, it offers bilingual dictionaries and Wikipedia knowledge bases that can be easily obtained within a short time frame.

搜集汇总

背景与挑战

背景概述

FGraDA是一个面向机器翻译的细粒度领域适应数据集，涵盖自动驾驶、人工智能教育、实时网络和智能手机四个信息技术子领域的中英翻译任务。该数据集不包含域内双语训练数据，但提供了双语词典和wiki知识库，以支持更贴近实际场景的评估。