def clean_up_highlights()

in summarize_from_feedback/datasets/cnndm.py [0:0]


def clean_up_highlights(line):
    if line.startswith("NEW: "):
        return line.split("NEW: ")[1].strip()
    bad_prefixes = ["READ: ", "for all the latest "]
    for prefix in bad_prefixes:
        if line.startswith(prefix):
            return ""
    # these are bad even if they appear mid-highlight
    bad_after = [
        "CLICK HERE ",
        "Click HERE ",
        "Click here ",
        "For confidential support call the Samaritans in the UK",
    ]
    for bad in bad_after:
        if bad in line:
            line = line.split(bad)[0].strip()
    return line