Update README.md

README.md

@@ -96,9 +96,9 @@ EUROPA is a dataset designed for training and evaluating multilingual keyphrase
 - **keyphrases:** reference keyphrases drafted by the CJEU.
 
 As explained in our paper, the dataset is split chronologically for assessing temporal generalization of models:
-- training set: 1957 to 2010 (131 076 instances);
-- validation set: 2011 to 2015 (63 373 instances);
-- test set: 2016 to 2023 (90 508 instances).
+- **training set**: 1957 to 2010 (131 076 instances);
+- **validation set**: 2011 to 2015 (63 373 instances);
+- **test set**: 2016 to 2023 (90 508 instances).
 
 ## Citation