nltk.word_tokenize isn't fast enough

Question

import pandas as pd
import numpy as np
import csv
import nltk
from nltk.corpus import stopwords
from nltk.stem import SnowballStemmer
from nltk.tokenize import word_toke