multilingual-ecommerce / normalizer.py
Abdul-Ib's picture
Create normalizer.py
28a38bd verified
raw
history blame
16.2 kB
import asyncio
import string, re
import pandas as pd
from aiogoogletrans import Translator
from spellchecker import SpellChecker
from nltk.tokenize import RegexpTokenizer
class Normalizer:
"""
A class for text normalization tasks such as converting to lowercase,
removing whitespace, punctuation, HTML tags, emojis, etc.
"""
def __init__(self):
"""
Initializes the Normalizer object.
"""
# Letter variations dictionary
self._letter_variations = {
"aàáâãäåāăą": "a",
"cçćĉċč": "c",
"eèéêëēĕėęě": "e",
"gğ": "g",
"hħĥ": "h",
"iìíîïīĭįı": "i",
"jĵ": "j",
"nñńņň": "n",
"oòóôõöøōŏő": "o",
"ś": "s",
"ß": "ss",
"uùúûüūŭůűų": "u",
"yýÿŷ": "y",
"æ": "ae",
"œ": "oe",
}
# Generate regex pattern including single characters
pattern_parts = []
for variation in self._letter_variations.keys():
pattern_parts.append(variation)
for char in variation:
if len(char) == 1:
pattern_parts.append(re.escape(char))
self._pattern = "|".join(pattern_parts)
# RegexpTokenizer
self._regexp = RegexpTokenizer("[\w']+")
# Dictionary of acronyms
acronyms_url = "https://raw.githubusercontent.com/sugatagh/E-commerce-Text-Classification/main/JSON/english_acronyms.json"
self._acronyms_dict = pd.read_json(acronyms_url, typ="series")
self._acronyms_list = list(self._acronyms_dict.keys())
# Dictionary of contractions
contractions_url = "https://raw.githubusercontent.com/sugatagh/E-commerce-Text-Classification/main/JSON/english_contractions.json"
self._contractions_dict = pd.read_json(contractions_url, typ="series")
self._contractions_list = list(self._contractions_dict.keys())
# Initialize translator for language detection
self._translator = Translator()
# Converting to lowercase
def _convert_to_lowercase(self, text):
"""
Convert the input text to lowercase.
Args:
text (str): The input text to be converted.
Returns:
str: The input text converted to lowercase.
"""
try:
return text.lower()
except Exception as e:
print(f"An error occurred during lowercase conversion: {e}")
return text
# Removing whitespaces
def _remove_whitespace(self, text):
"""
Remove leading and trailing whitespaces from the input text.
Args:
text (str): The input text to be processed.
Returns:
str: The input text with leading and trailing whitespaces removed.
"""
try:
return text.strip()
except Exception as e:
print(f"An error occurred during whitespace removal: {e}")
return text
# Removing punctuations
def _remove_punctuation(self, text):
"""
Remove punctuation marks from the input text, except for apostrophes and percent signs.
Args:
text (str): The input text to be processed.
Returns:
str: The input text with punctuation marks removed.
"""
try:
punct_str = string.punctuation
punct_str = punct_str.replace("'", "").replace(
"%", ""
) # discarding apostrophe from the string to keep the contractions intact
return text.translate(str.maketrans("", "", punct_str))
except Exception as e:
print(f"An error occurred during punctuation removal: {e}")
return text
# Removing HTML tags
def _remove_html(self, text):
"""
Remove HTML tags from the input text.
Args:
text (str): The input text containing HTML tags.
Returns:
str: The input text with HTML tags removed.
"""
try:
html = re.compile(r"<.*?>")
return html.sub(r"", text)
except Exception as e:
print(f"An error occurred during HTML tag removal: {e}")
return text
# Removing emojis
def _remove_emoji(self, text):
"""
Remove emojis from the input text.
Args:
text (str): The input text containing emojis.
Returns:
str: The input text with emojis removed.
"""
try:
emoji_pattern = re.compile(
"["
"\U0001F600-\U0001F64F" # emoticons
"\U0001F300-\U0001F5FF" # symbols & pictographs
"\U0001F680-\U0001F6FF" # transport & map symbols
"\U0001F1E0-\U0001F1FF" # flags (iOS)
"\U00002702-\U000027B0"
"\U000024C2-\U0001F251"
"]+",
flags=re.UNICODE,
)
return emoji_pattern.sub(r"", text)
except Exception as e:
print(f"An error occurred during emoji removal: {e}")
return text
# Removing other unicode characters
def _remove_http(self, text):
"""
Remove HTTP links from the input text.
Args:
text (str): The input text containing HTTP links.
Returns:
str: The input text with HTTP links removed.
"""
try:
http = "https?://\S+|www\.\S+" # matching strings beginning with http (but not just "http")
pattern = r"({})".format(http) # creating pattern
return re.sub(pattern, "", text)
except Exception as e:
print(f"An error occurred during HTTP link removal: {e}")
return text
# Function to convert contractions in a text
def _convert_acronyms(self, text):
"""
Convert acronyms in the text.
Example of acronyms dictionary:
{"LOL": "laugh out loud", "BRB": "be right back", "IDK": "I don't know"}
Args:
text (str): The input text containing acronyms.
Returns:
str: The input text with acronyms expanded.
"""
try:
words = []
for word in self._regexp.tokenize(text):
if word in self._acronyms_list:
words = words + self._acronyms_dict[word].split()
else:
words = words + word.split()
text_converted = " ".join(words)
return text_converted
except Exception as e:
print(f"An error occurred during acronym conversion: {e}")
return text
# Function to convert contractions in a text
def _convert_contractions(self, text):
"""
Convert contractions in the text.
Example of contractions dictionary:
{"I'm": "I am", "he's": "he is", "won't": "will not"}
Args:
text (str): The input text containing contractions.
Returns:
str: The input text with contractions expanded.
"""
try:
words = []
for word in self._regexp.tokenize(text):
if word in self._contractions_list:
words = words + self._contractions_dict[word].split()
else:
words = words + word.split()
text_converted = " ".join(words)
return text_converted
except Exception as e:
print(f"An error occurred during contraction conversion: {e}")
return text
def _fix_letter_variations(self, query):
"""
Replace variations of letters with their original counterparts.
Args:
query (str): The input query containing variations of letters.
Returns:
str: The normalized query with variations replaced by their original counterparts.
"""
def replace_variation(match):
"""
Helper function to replace variations with original counterparts.
Args:
match (re.Match): The match object representing the found variation.
Returns:
str: The original character if match is not found in letter_variations, otherwise its original counterpart.
"""
for key in self._letter_variations.keys():
if match.group(0) in key:
return self._letter_variations[key]
return match.group(0)
try:
# Fixing the query
normalized_query = re.sub(self._pattern, replace_variation, query)
return normalized_query
except Exception as e:
print(f"An error occurred during letter variation fixing: {e}")
return query
def _normalize_query(self, word: str):
"""
Clean the input text by performing the following steps:
1. Remove non-alphabetic characters and keep specific characters like spaces, dashes, asterisks, and Arabic characters.
2. Remove non-alphabetic characters between alphabetic characters.
3. Remove repeating characters.
4. Remove preceding numbers (e.g. 123phone -> phone).
5. Add space between numbers and letters.
6. Remove extra spaces.
Args:
word (str): The input text to be cleaned.
Returns:
str: The cleaned text.
"""
try:
# Remove non-alphabetic characters and keep specific characters like spaces, dashes, asterisks, and Arabic characters
word = re.sub(
r"[^A-Za-z\s\-%*.$\u0621-\u064A0-9\u00E4\u00F6\u00FC\u00C4\u00D6\u00DC\u00df]",
"",
word,
flags=re.UNICODE,
)
# Remove non-alphabetic characters between alphabetic characters
clean_text = re.sub(
r"(?<=[a-zA-Z])([^A-Za-z\u0621-\u064A\s]+)(?=[a-zA-Z])", "", word
)
# Remove non-alphabetic characters between alphabetic characters
clean_text = re.sub(r"(?<=[a-zA-Z])([^A-Za-z\s]+)(?=[a-zA-Z])", "", clean_text)
# Remove non-alphabetic characters between Arabic characters
clean_text = re.sub(
r"(?<=[\u0621-\u064A])([^\u0621-\u064A\s]+)(?=[\u0621-\u064A])",
"",
clean_text,
)
# Remove repeating characters
clean_text = re.sub(r"(.)(\1+)", r"\1\1", clean_text)
# Remove preceding non latin alpha (e.g. صصphone -> phone)
clean_text = re.sub(r"([\u0621-\u064A]+)([a-zA-Z]+)", r"\2", clean_text)
# Add space between numbers and letters
clean_text = re.sub(r"([a-zA-Z]+)([\u0621-\u064A]+)", r"\1", clean_text)
# Remove preceding latin alpha (from arabic words) (e.g. phoneصص -> phone)
clean_text = re.sub(r"([a-zA-Z]+)([\u0621-\u064A]+)", r"\2", clean_text)
# Add space between numbers and letters
clean_text = re.sub(r"([\u0621-\u064A]+)([a-zA-Z]+)", r"\1", clean_text)
# Remove preceding numbers (e.g. 123phone -> phone)
clean_text = re.sub(r"(\d+)([a-zA-Z\u0621-\u064A]+)", r"\1 \2", clean_text)
# Add space between numbers and letters
clean_text = re.sub(r"([a-zA-Z\u0621-\u064A]+)(\d+)", r"\1 \2", clean_text)
# Remove extra spaces
clean_text = re.sub(r"\s+", " ", clean_text)
return clean_text.strip()
except Exception as e:
print(f"An error occurred during query normalization: {e}")
return word
def keep_one_char(self, word: str) -> str:
"""
Keep only one occurrence of consecutive repeated characters in the input word.
Args:
- word (str): The input word to modify.
Returns:
- str: The modified word with only one occurrence of consecutive repeated characters.
"""
try:
return re.sub(r"(.)(\1+)", r"\1", word)
except Exception as e:
print(f"An error occurred during character repetition removal: {e}")
return word
def translate_text(self, text: str) -> str:
"""
Translate the given text to English and return the translated text.
Args:
- text (str): The text to translate.
Returns:
- str: The translated text.
"""
try:
loop = asyncio.get_event_loop()
translated_text = (
loop.run_until_complete(self._translator.translate(text))
.text.lower()
.strip()
)
except Exception as e:
print(f"Text Translation failed: {e}")
translated_text = (
text.lower().strip()
) # Use original text if translation fails
return translated_text
def check_spelling(self, query: str) -> str:
"""
Check the spelling of the input query and return the corrected version.
Args:
- query (str): The input query to check its spelling.
Returns:
- str: The corrected query.
"""
try:
# Detect the language of the input query using Google Translate API
# input_language = self._translator.detect(query)
input_language = "en" if query.encode().isalpha() else "ar"
# Initialize SpellChecker with detected language, fallback to English if language detection fails
try:
spell_checker = SpellChecker(language=input_language)
except:
spell_checker = SpellChecker(language="en")
# Initialize an empty string to store the corrected query
result_query = ""
# Iterate through each word in the query
for word in query.split(" "):
# Get the corrected version of the word
corrected_word = spell_checker.correction(word)
# If the corrected word is not found, try correcting with keeping one character
if corrected_word is None:
corrected_word = spell_checker.correction(self.keep_one_char(word))
# If still not found, keep the original word
if corrected_word is None:
result_query += word + " "
else:
result_query += corrected_word + " "
else:
result_query += corrected_word + " "
# Remove trailing whitespace and return the corrected query
return result_query.strip()
except Exception as e:
print(f"An error occurred during spelling check: {e}")
return query
def clean_text(self, text):
"""
Normalize the input text.
Args:
text (str): The input text to be normalized.
Returns:
str: The normalized text.
"""
try:
# Convert text to lowercase
text = self._convert_to_lowercase(text)
# Remove whitespace
text = self._remove_whitespace(text)
# Convert text to one line
text = re.sub("\n", " ", text)
# Remove square brackets
text = re.sub("\[.*?\]", "", text)
# Remove HTTP links
text = self._remove_http(text)
# Remove HTML tags
text = self._remove_html(text)
# Remove emojis
text = self._remove_emoji(text)
# Fix letter variations
text = self._fix_letter_variations(text)
# Normalize queries
text = self._normalize_query(text)
return text
except Exception as e:
print(f"An error occurred during text cleaning: {e}")
return text