functions_preprocess.py

import string
import pandas as pd
from bs4 import BeautifulSoup
import re
from torchtext.vocab import build_vocab_from_iterator, GloVe
import numpy as np
from sklearn.base import TransformerMixin
from sklearn.metrics import ConfusionMatrixDisplay
from keras.preprocessing.text import Tokenizer
import nltk
nltk.download('stopwords')
nltk.download('punkt')
from nltk.corpus import stopwords
from nltk.stem import WordNetLemmatizer
from nltk.tokenize import word_tokenize
from nltk.corpus import wordnet
import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
from torchtext.data.utils import get_tokenizer

def download_if_non_existent(res_path, res_name):
  try:
    nltk.data.find(res_path)
  except LookupError:
    print(f'resource {res_path} not found. Downloading now...')
    nltk.download(res_name)

download_if_non_existent('corpora/stopwords', 'stopwords')
download_if_non_existent('taggers/averaged_perceptron_tagger', 'averaged_perceptron_tagger')
download_if_non_existent('corpora/wordnet', 'wordnet')

def fit_model(pipeline, x_train, y_train, x_test, y_test):
  pipeline.fit(x_train, y_train)
  return ConfusionMatrixDisplay.from_estimator(pipeline, x_test, y_test, normalize="true")

class LinguisticPreprocessor(TransformerMixin):
  def __init__(self, ):
    super().__init__()
    self.lemmatizer = WordNetLemmatizer()
    self.tokenizer = Tokenizer()
    self.stop_words = set(stopwords.words('english'))
    self.stop = stopwords.words('english')

  def fit(self, X, y=None):
    return self

  def transform(self, X, y=None):
    X = self._remove_html_tags(X)
    X = self._remove_all_punctuations(X)
    X = self._remove_double_spaces(X)
    X = self._lemmatize(X)
    X = self._remove_stopwords(X)
    return X

  def _remove_html_tags(self, X):
    X = list(map( lambda x: BeautifulSoup(x, 'html.parser').get_text(), X))
    return X

  def _remove_all_punctuations(self, X):
    X = list(
        map(
            lambda text: re.sub('[%s]' % re.escape(string.punctuation), '', text),
            X
        )
    )
    return X

  def _remove_double_spaces(self, X):
    X = list(map(lambda text: re.sub(" +", " ", text), X))
    return X

  def _remove_stopwords(self, X):
    X = list(map(
            lambda text:  " ".join(
                [
                    word for word in text.split() if word not in (self.stop_words)
                ]
            ),
            X
        )
    )
    return X

  def _lemmatize(self, X):
    X = list(map(lambda text: self._lemmatize_one_sentence(text), X))
    return X

  def _lemmatize_one_sentence(self, sentence):
    sentence = nltk.word_tokenize(sentence)
    sentence = list(map(lambda word: self.lemmatizer.lemmatize(word), sentence))
    return " ".join(sentence)

def training_data(dataset_1, dataset_2, dataset_3):
  X_test = dataset_1['test']['text']
  y_test = dataset_1['test']['label']

  test_df = pd.DataFrame({
      'text':X_test,
      'label': y_test
  })

  combined_train_df = pd.DataFrame({
      'text': dataset_1['train']['text'] + dataset_2['train']['text'] + dataset_3['train']['text'],
      'label': dataset_1['train']['label'] + dataset_2['train']['label'] + dataset_3['train']['label']
  })

  combined_train_df.drop_duplicates(subset=['text'], inplace=True)

  merged_df = pd.merge(combined_train_df, test_df, on="text", how='left', indicator=True)
  result_df = merged_df[merged_df['_merge'] == 'left_only'].drop(columns=['_merge'])


  X_train = result_df['text'].tolist()
  y_train = result_df['label_x'].tolist()
  X_test = np.array(X_test)
  X_train = np.array(X_train)

  return X_train, y_train, X_test, y_test

class CNN(nn.Module):
    def __init__(self, vocab_size, embed_size, n_filters, filter_sizes, dropout, num_classes):
        super(CNN, self).__init__()
        self.embedding = nn.Embedding(vocab_size, embed_size)
        self.convs = nn.ModuleList([nn.Conv2d(in_channels=1, out_channels=n_filters, kernel_size=(fs, embed_size)) for fs in filter_sizes])
        self.dropout = nn.Dropout(dropout)
        self.fc1 = nn.Linear(len(filter_sizes) * n_filters, num_classes)

    def forward(self, text):
        embedded = self.embedding(text)
        embedded = embedded.unsqueeze(1)
        conved = [F.leaky_relu(conv(embedded)).squeeze(3) for conv in self.convs]
        pooled = [F.max_pool1d(conv, conv.shape[2]).squeeze(2) for conv in conved]
        cat = self.dropout(torch.cat(pooled, dim=1))
        return self.fc1(cat)

def build_vocab(data_iter):
    tokenizer = get_tokenizer("basic_english")

    def yield_tokens():
        for example in data_iter:
            cleaned_text = clean_text(example['text'])
            yield tokenizer(cleaned_text)

    vocab = build_vocab_from_iterator(yield_tokens(), specials=["<unk>", "<pad>"])
    vocab.set_default_index(vocab["<unk>"])
    return vocab, tokenizer

def clean_text(text):
    text = text.lower()
    text = re.sub(r'\d+', '', text)
    text = re.sub(r'[^\w\s]', '', text)
    text = ' '.join([lemmatizer.lemmatize(word) for word in text.split() if word not in stop_words])
    return text