import neologdn
import MeCab
import re
from transformers.models.whisper.english_normalizer import BasicTextNormalizer

wakati = MeCab.Tagger("-Owakati")
special_characters = '[\,\、\。\．\「\」\…\？\・\!\-\;\:\"\“\%\‘\”\�]'


def norm_everything(batch): 
    batch["sentence"] = neologdn.normalize(batch["sentence"]).strip() 
    batch["sentence"] = normalizer(batch["sentence"]).strip() 
    batch["sentence"] = wakati.parse(batch["sentence"]).strip() 
    batch["sentence"] = re.sub(special_characters,'', batch["sentence"]).strip() 
    return batch 

ds = ds.map(norm_everything)