German-Greek Journalistic Machine Translated Corpus with Sentence-Level Alignment and Human Translation Samples

This dataset consists of a corpus of German-Greek journalistic texts (DE–GR) and domain-specific journalistic terminology, designed to support research in journalistic translation, terminology studies, and machine translation evaluation. The data is provided in an Excel (.xlsx) file and is organized at the sentence level. Each row corresponds to a sentence extracted from German journalistic texts containing specialized terminology and references to journalistic content. Structure of the Dataset The Excel file includes the following columns: 1. Source References Contains URL links to the original journalistic sources and references of journalistic content. 2. German Journalistic Texts Containing Specialized Terminology The original German journalistic text, segmented at the sentence level, with each sentence appearing in a separate cell. 3. Neural Machine Translation Output Generated by Google Translate (2021) Machine-translated output produced using Google Translate (2021 version). 4. Translations Generated by a Deep Learning–Based Neural Machine Translation System (2025) Machine-translated output generated by a neural machine translation system (2025). 5. Human-Validated Corrections of Automatically Rendered Terminology Human-reviewed and corrected translations, focusing on improving the accuracy of specialized journalistic terminology. Purpose This corpus enables:- Comparative analysis of machine translation quality across time and MT systems (2021 Google Translate vs 2025 DeepLearning Translate)- Evaluation of terminology translation in journalistic contexts- Research on human-in-the-loop correction workflows- Development and benchmarking of NLP systems for journalistic language Format - File format: Excel (.xlsx)- Unit of analysis: Sentence-level alignment- Language pair: German → Greek- Encoding: UTF-8 Notes - The dataset focuses on domain-specific terminology found in journalistic texts and media discourse.- Source references include journalistic articles and bibliographic material relevant to media content.- Human corrections are selectively applied and aim to enhance terminological precision.