IndicLink is a Multilingual Fact Linking (MFL) dataset of sentences and a set of WikiData facts (subject; relation; object) contained in each sentence. IndicLink contains sentences from English and 6 Indian languages - Hindi, Telugu, Tamil, Urdu, Gujarati and Assamese. The correct facts are chosen from an oracle of 4.7 million Wikidata facts with fact labels/descriptions available in these 7 languages. The dataset is intended only to act as a test set to evaluate models trained for the task of MFL. For more details, please see https://arxiv.org/abs/2109.14364 - View it on GitHub
Star
0
Rank
11084539