Simple converter to ConLL-2003 NER format for spaCy model training

convert.py

DATA = [
  [
    [['Who', 'is', 'Shaka', 'Khan', '?'], ['O', 'O', 'I-PER', 'I-PER', 'O']]
  ],
  [
    [['I', 'like', 'London', 'and', 'Berlin', '.'], ['O', 'O', 'I-LOC', 'O', 'I-LOC', 'O']]
  ]
]

with open('output.conll', 'w') as f:
  for doc in DATA:
    f.write('-DOCSTART- -X- O O\n')
    for sentence, sent_entities in doc:
      for token, BIO_tag in zip(sentence, sent_entities):
        f.write('{} -X- _ {}\n'.format(token, BIO_tag))
      f.write('\n')

## Result
# -DOCSTART- -X- O O
# Who -X- _ O
# is -X- _ O
# Shaka -X- _ I-PER
# Khan -X- _ I-PER
# ? -X- _ O
# 
# -DOCSTART- -X- O O
# I -X- _ O
# like -X- _ O
# London -X- _ I-LOC
# and -X- _ O
# Berlin -X- _ I-LOC
# . -X- _ O

Hey is there a way to go from Conll format to spacy format?

You can check out my repo for the solution.
https://github.com/dipansh-girdhar/NLP/tree/master/NER/Spacy%20NER