RTE3-FR

RTE3-FR dataset is the French translation of the Textual Entailment English dataset used in the RTE-3 Challenge (https://nlp.stanford.edu/RTE3-pilot).

Like its English counterpart, the French RTE-3 dataset is composed of a development set and a test set, each containing 800 T/H pairs.

The dataset is annotated for a 3-way task with the following labels: entailment (0), neutral (1), contradiction (2).

RTE3-FR is available both in XML and TSV format.