guess_importance.py

import os
import re

import tf_idf

INBOX = "S:/Merchant Notices/Current Action/"
G1_WORDS = [
    "sue", "suit", "seize", "seizure", "lien", "levy", "terminate",
    "termination", "final", "court date", "garnish", "litigation",
    "proceedings", "legal action", "dispute"]


def build_profile():
    target_pdfs = list(target_files(INBOX))
    target_bags = tf_idf.get_word_bags(target_pdfs)
    target_idf = tf_idf.build_idf(target_bags)
    return target_idf


def target_files(inbox=INBOX):
    return (
        INBOX + filename for filename in os.listdir(inbox)
        if is_from_july_2019_or_august_2019(filename)
        if is_from_kansas_or_wisconsin(filename)
        if is_g1(filename))


def is_from_july_2019_or_august_2019(filename):
    return any([
        re.search(r"0[7-8]\d\d19", filename),
        re.search(r"2019-?0[7-8]-?\d\d", filename)])


def is_from_kansas_or_wisconsin(filename):
    return re.search(r"^(KS|WI)", filename)


def is_g1(filename):
    return any(word in filename.casefold() for word in G1_WORDS)