Update files from the datasets library (from 1.5.0)

Release notes: https://github.com/huggingface/datasets/releases/tag/1.5.0

bc2gm_corpus.py

@@ -119,7 +119,7 @@ class Bc2gmCorpus(datasets.GeneratorBasedBuilder):
             tokens = []
             ner_tags = []
             for line in f:
-                if line.startswith("-DOCSTART-") or line == "" or line == "\n":
+                if line == "" or line == "\n":
                     if tokens:
                         yield guid, {
                             "id": str(guid),