Update README.md

README.md

@@ -38,22 +38,15 @@ Abstraction Level: The model tends to be more extractive than abstractive in its
 ## Training and evaluation data
 
 
-News Articles Dataset:
+Dataset:
 
-Source: CNN/Daily Mail dataset (version 3.0.0)
-Size: Approximately 200,000 articles
-Time Range: 2007-2021
+Source: PARANMT-50M 
+Size: Approximately 50M
+Time Range: 2007-2017
 Language: English
-Content: Wide range of topics including politics, sports, entertainment, and world events
-
-
-Academic Articles Dataset:
-
-Source: arXiv and PubMed Open Access Subset
-Size: Approximately 150,000 articles
-Time Range: 2010-2022
-Language: English
-Content: Research papers from various scientific fields including physics, mathematics, computer science, and biomedical sciences
+Content: more than 50 million English-English
+sentential paraphrase pairs
+https://arxiv.org/pdf/1711.05732v2
 
 
 Pre-processing Steps: