fleurs / text_processor /text_processor.py
cahya's picture
add fleurs dataset with text normaliser for indonesian
230a13a
import re
from num2words import num2words
import os
def get_abs_path(rel_path):
"""
Get absolute path
Args:
rel_path: relative path to this file
Returns absolute path
"""
return os.path.dirname(os.path.abspath(__file__)) + '/' + rel_path
class TextProcessor:
thousands = ["ratus", "ribu", "juta", "miliar", "milyar", "triliun"]
months = ["Januari", "Februari", "Maret", "April",
"Mei", "Juni", "Juli", "Agustus",
"September", "Oktober", "November", "Desember"]
measurements_path = get_abs_path("measurements.tsv")
currencies_path = get_abs_path("currency.tsv")
timezones_path = get_abs_path("timezones.tsv")
def __init__(self):
self.measurements = {}
with open(TextProcessor.measurements_path, "r") as file:
for line in file:
line = line.strip().split("\t")
self.measurements[line[0]] = line[1]
self.currencies = {}
with open(TextProcessor.currencies_path, "r") as file:
for line in file:
line = line.strip().split("\t")
self.currencies[line[0]] = line[1]
self.timezones = {}
with open(TextProcessor.timezones_path, "r") as file:
for line in file:
line = line.strip().split("\t")
self.timezones[line[0]] = line[1]
self.re_thousands = '|'.join([t for t in TextProcessor.thousands])
self.re_currencies = r'\b' + re.sub(r'\|([^|$£€¥₩]+)', r'|\\b\1', '|'.join([c for c in self.currencies]))
self.re_currencies = re.sub(r'([$£€¥₩])', r'\\\1', self.re_currencies)
self.re_moneys = r'(({}) ?([\d\.\,]+)( ({})?(an)?)?)'.format(self.re_currencies, self.re_thousands)
self.re_measurements = '|'.join([t for t in self.measurements])
self.re_measurements = r'(\b([\d\.\,]+) ?({})\b)'.format(self.re_measurements)
self.re_timezones = '|'.join([c for c in self.timezones])
self.re_timezones = r'((\d{1,2})[\.:](\d{1,2}) ' + r'\b({})\b)'.format(self.re_timezones)
self.re_http = r'(https?://(www\.)?[-a-zA-Z0-9@:%._\+~#=]{1,256}\.[a-zA-Z0-9()]{1,6}\b[-a-zA-Z0-9()@:%_\+.~#?&//=]*)'
@staticmethod
def is_integer(number):
try:
int(number)
return True
except ValueError:
return False
@staticmethod
def is_float(number):
try:
float(number)
return True
except ValueError:
return False
def normalize(self, text):
found_errors = False
# Remove URL
urls = re.findall(self.re_http, text)
for url in urls:
text = text.replace(url[0], "")
# Currency
moneys = re.findall(self.re_moneys, text)
for money in moneys:
number = re.sub(',', '.', re.sub(r'\.', '', money[2].strip(" ,.")))
try:
if number == "":
continue
if self.is_integer(number):
number = int(number)
elif self.is_float(number):
number = float(number)
else:
number = re.sub(r'[.,]', "", number)
number = int(number)
number = num2words(number, to='cardinal', lang='id')
text = text.replace(money[0].strip(" ,."), f'{number} {money[3]} {self.currencies[money[1]]}')
except Exception as error:
found_errors = True
print(error)
print(f'Problem with money: <{text}>: {number}')
# Measurements
units = re.findall(self.re_measurements, text)
for unit in units:
number = re.sub(',', '.', re.sub(r'\.', '', unit[1].strip(" ,.")))
try:
if number == "":
continue
if re.search(r'\.', number):
number = float(number)
else:
number = int(number)
number = num2words(number, to='cardinal', lang='id')
text = text.replace(unit[0].strip(" ,."), f'{number} {self.measurements[unit[2]]}')
except Exception as error:
found_errors = True
print(error)
print(f'Problem with measurements: <{text}>: {number}')
# Date
dates = re.findall(r'(\((\d{1,2})/(\d{1,2})(/(\d+))?\))', text)
for date in dates:
try:
day = num2words(int(date[1]), to='cardinal', lang='id')
month = int(date[2]) - 1
if month >= 12:
month = 0
month = self.months[month]
if date[4] != "":
year = num2words(int(date[4]), to='cardinal', lang='id')
date_string = f'{day} {month} {year}'
else:
date_string = f'{day} {month}'
text = text.replace(date[0], f' {date_string} ')
except Exception as error:
found_errors = True
print(error)
print(f'Problem with dates: <{text}>: {date}')
# Timezones
timezones = re.findall(self.re_timezones, text)
for timezone in timezones:
try:
hour = num2words(int(timezone[1]), to='cardinal', lang='id')
minute = num2words(int(timezone[2]), to='cardinal', lang='id')
zone = self.timezones[timezone[3]]
if minute == "nol":
time_string = f'{hour} {zone}'
else:
time_string = f'{hour} lewat {minute} menit {zone}'
text = text.replace(timezone[0], f'{time_string}')
except Exception as error:
found_errors = True
print(error)
print(f'Problem with timezones: <{text}>: {timezone}')
# Any number
re_numbers = [r'([\d.,]+)', r'\d+']
for re_number in re_numbers:
number_len = 0
for i in re.finditer(re_number, text):
start = i.start() + number_len
end = i.end() + number_len
number = text[start:end]
number = re.sub(',', '.', re.sub(r'\.', '', number.strip(" ,.")))
if number == "":
continue
if self.is_integer(number) or self.is_float(number):
try:
if self.is_integer(number):
number = int(number)
else:
number = float(number)
number = num2words(number, to='cardinal', lang="id")
text = text[:start] + number + text[end:]
number_len += len(number) - (end - start)
except Exception as error:
found_errors = True
print(error)
print(f'Problem with number: <{text}>: {number}')
text = re.sub(r"\s+", " ", text)
if found_errors:
print(f'>>> {text}')
return text