nuredinali/Tigrinya-English-MT-Evaluation-Datasets

**About**: This is the dataset for the project “Error Analysis of Tigrinya-English Machine Translation Systems.” In this project, we evaluated the status of three state-of-the-art MT systems that support the translation of Tigrinya to and from English: Google translate, Microsoft translator, and Lesan. Tigrinya is a language that is spoken by more than 10 million native speakers mainly in Tigray, Ethiopia and Eritrea. The details of the dataset are described below: **Schema** - Source - This is the source url where the original data is scraped from. Typically takes a URL of a web resource as value but in some cases it could be page number of a textbook. - Domain - refers to the area from which the text was gathered. The dataset reflects a diverse content. This feature can take one of the following values: Art and Culture, Business and Economy, Politics as well as Science and Technology. - SourceText - is the source text given as input to the system for translation. - SourceLanguage - the language of the source text. This takes values: English (en) or Tigrinya (ti). - TargetLanguage - the language of the target translation output. This takes values: English (en) or Tigrinya (ti). - OutputGoogle - text translation output from Google Translate. - EvalSentenceLevelGoogle - sentence level evaluation of translation output from Google Translate. - EvalSnippetLevelGoogle - snippet level evaluation of translation output from Google Translate. - ErrorTypeGoogle - one or more Error type(s) according to MQM on translation output from Google Translate. - OutputMicrosoft- translation output from Microsoft Translator. - EvalSentenceLevelMicrosoft - sentence level evaluation of translation output from Microsoft Translator - EvalSnippetLevelMicrosoft - snippet level evaluation of translation output from Microsoft Translator. - ErrorTypeMcirosoft: One or more Error type(s) according to MQM on translation output from Microsoft Translator. …