import pandas as pd 

df = pd.read_parquet("acl.parquet")

articles = df
articles["article_id"] = df.index.astype(int)
articles["date_created"] = articles["year"]
articles["categories"] = None

articles_authors = articles[["article_id", "acl_id", "author"]]
articles_authors["bai"] = articles_authors["author"].str.split("and\n").fillna("")
articles_authors = articles_authors.explode("bai")
articles_authors["bai"] = articles_authors["bai"].str.lower().str.strip()
articles_authors["bai"] = articles_authors["bai"].str.extract(r"^((?:.*), [a-z])", expand=False)
articles_authors = articles_authors[articles_authors["bai"].fillna("").map(len)>1]
articles_authors = articles_authors[["article_id", "acl_id", "bai"]]
articles_authors.to_parquet("articles_authors.parquet")

articles.to_parquet("articles.parquet")