Igbo-English Translated Data

This dataset includes several files that provide linguistic resources for machine translation of Igbo to English. The files whose names start with scraped_ consist of monolingual data scraped and processed from BBC News (Igbo edition), Tedtalk, and a popular online newspaper in Igbo. Files whose names begin with gpt_3.5 and gpt_4.0 contain parallel data, translated from the monolingual datasets using gpt-3.5 and gpt-4.0, respectively. Lastly, the MTurk_Pidgin file features a high-quality parallel dataset translated by native Nigerian speakers with an acceptance rate of over 50% on Amazon Mechanical Turk.