annotator/utils.py

import re


def divide_prompt(prompt):
    # seqis like ". [TOP-LEFT]:"
    match_sep = re.compile(r"\.\s+[A-Z0-9-\[\]]+:")
    seps = match_sep.findall(prompt)
    # Add '.' for each sentence
    sub_prompts = [
        p + '.' if p.strip()[-1] != '.' else p
        for p in re.split('|'.join(map(re.escape, seps)), prompt)
    ]
    return sub_prompts