import gradio as gr # Library for building User Interfaces
import nltk
from sklearn.feature_extraction.text import CountVectorizer 
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.preprocessing import StandardScaler
import joblib
from nltk.stem.porter import PorterStemmer
from nltk.corpus import stopwords
import re
import numpy as np
from gensim.models import Word2Vec

nltk.download('stopwords')

# import warnings
# from sklearn.exceptions import InconsistentVersionWarning
# warnings.filterwarnings("ignore", category=InconsistentVersionWarning)

ps = PorterStemmer()
scaler = StandardScaler()

vectorizer = joblib.load('./vectorizer.pkl')
nb_classifier = joblib.load('./nb_classifier.pkl')
tfidf_vectorizer = joblib.load('./tfidf_vectorizer.pkl')
random_forest = joblib.load('./random_forest.pkl')
word2vec_model = Word2Vec.load("./word2vec_model.bin")
svm_classifier = joblib.load('./svm_classifier.pkl')

def preprocess_for_bow_and_tfidf(text):
    corpus = []
    text = re.sub('[^a-zA-Z0-9$£€¥%]',' ',text)
    text = text.lower()
    text = text.split()
    text = [ps.stem(t) for t in text if t not in stopwords.words('english')]
    corpus.append(' '.join(text))

    return corpus

def preprocess_for_word2vec(text):
    corpus = []
    text = re.sub('[^a-zA-Z0-9$£€¥%]',' ',text)
    text = text.lower()
    text = text.split()
    text = [ps.stem(t) for t in text if t not in stopwords.words('english')]
    corpus.append(text)

    return corpus

def average_word_vectors(words, model, vocabulary, num_features):
    feature_vector = np.zeros((num_features,), dtype="float64")
    nwords = 0.
    
    for word in words:
        if word in vocabulary:
            nwords = nwords + 1.
            feature_vector = np.add(feature_vector, model.wv[word])
    
    if nwords:
        feature_vector = np.divide(feature_vector, nwords)
        
    return feature_vector
    


def classify(text,choice):
  corpus=[text]
  if(choice == 1):
      corpus = preprocess_for_bow_and_tfidf(text)
      features = vectorizer.transform(corpus).toarray()
      prediction = nb_classifier.predict(features)
      
  elif(choice == 2):
      corpus = preprocess_for_bow_and_tfidf(text)
      features = tfidf_vectorizer.transform(corpus).toarray()
      prediction = random_forest.predict(features)
      
  elif(choice == 3):
      corpus = preprocess_for_word2vec(text)
      X = [average_word_vectors(title, word2vec_model, word2vec_model.wv.key_to_index, 100) for title in corpus]
      X = np.array(X)
      X = scaler.transform(X)
      prediction = svm_classifier.predict(X)
      
      
  if(prediction == 1):
     return "Fake News"
  else:
     return "Not Fake News"
      
GUI = gr.Interface(
    inputs = [
    gr.Textbox(label = "Input" , placeholder = "Enter News Title here"),
    gr.Radio( choices = [("Naive Bayes",1) , ("Random Forest",2) , ("Support Vector Machine",3) ] , value = 1 , label = "Model") ],
    outputs = ['text'],
    fn = classify,
    title = "Fake News Detection System"
)
GUI.launch(debug = True)