SNLI Dataset


The v1.0 corpus is a collection of 570k human-generated sentence pairs in the English language that are manually labeled for a balanced classification.The entailment, contradiction, and neutral, of the labels support tasks like NLI (natural language inference), often known as RTE (recognizing textual entailment). It is geared towards serving both as a benchmark for evaluating text representational systems, such as ones induced by representation learning methods, as well as a useful resource for developing various NLP (natural language processing) models.