Update README.md

README.md

@@ -5,4 +5,11 @@ pretty_name: cc100-sr derivation by JeRTeh
 ---
 
 As the title suggests, this dataset is a derivative of the cc100-sr common-crawl based dataset for Serbian. 
-It was deduplicated on the sentence level, and transliterated into latin alphabet.
+It was deduplicated on the sentence level, and transliterated into latin alphabet.
+
+Dataset is constituted of JSON files, where the textual sentences are located in the "sents" attribute of the object root and can be obtianed via:
+
+from json import load
+
+with open("cc100-sr-ded.json") as jf:
+  sentences = load(jf)["sents"]