AmaarAli/NewsTopicModeling

1from flask import Flask,request,render_template,jsonify2import pandas as pd3import numpy as np4import matplotlib.pyplot as plt5import seaborn as sns6import nltk7nltk.data.path.append('./nltk_data')8from nltk.corpus import stopwords9from nltk.tokenize import word_tokenize10import joblib11import string12from flask_cors import CORS13 14 15app = Flask(__name__)16 17CORS(app)18 19try:20    model = joblib.load('NMF_Model.joblib')21    vectorizer = joblib.load('Vectorizer.joblib')22    print('Model and vectorizer were loaded')23    predicted_ready = True24    feature_names = vectorizer.get_feature_names_out()25except Exception as e:26    print(f'Error occured while loadin the model :- {e}')27    vectorizer = None28    model = None29    feature_names = None30    prediction_ready = False31 32 33def lowercasing(txt):34    return txt.lower()35 36def remove_stopwords(txt):37    stop_words = set(stopwords.words('english'))38    try:39        words = word_tokenize(txt)40    except Exception as e:41        print(f'Error occured while removing stopwords :- {e}')42        return ""43    44    cleaned_words = [i.lower() for i in words if i.isalpha() and i.lower() not in stop_words]45    return " ".join(cleaned_words)46 47def remove_numbers(txt):48    return "".join([i for i in txt if not i.isdigit()])49 50def remove_punctuation(txt):51    return txt.translate(str.maketrans('','',string.punctuation))52 53 54def preprocess_txt(text):55    text = lowercasing(text)56    text = remove_numbers(text)57    text = remove_punctuation(text)58    text = text.replace('\n','').replace('\r','')59    text = remove_stopwords(text)60    return text61 62 63topic_names = {64    0: "Business & Finance",65    1: "Politics",66    2: "Sports (Rugby & General)",67    3: "Film & Awards",68    4: "Economics & Growth",69    5: "Government & Law",70    6: "Olympics & Athletics",71    7: "Energy & Oil",72    8: "Technology & Software",73    9: "Mobile & Digital Media"74}75 76def get_top_words(topic_id,num_words=300):77    if not predicted_ready:78        return []79    80    topic_weights = model.components_[topic_id]81    top_word_indices = topic_weights.argsort()[:-num_words -1:-1]82 83    top_words = []84    for i in top_word_indices:85        word = feature_names[i]86        weight = float(topic_weights[i])87        top_words.append([word,weight])88    return top_words89    90    91 92 93@app.route('/')94def index():95    return render_template('index.html')96 97@app.route('/predict',methods=['POST'])98def predict():99    if model is None or vectorizer is None:100        return jsonify({'error' : 'Model not loaded. Please check server logs'})101    102    data = request.get_json()103    if not data or 'text' not in data:104        return jsonify({'error' : 'No Text field provided, please provide it'})105    106    input_text = data['text']107    print('Preprocessed before :-', input_text)108    preprocessed_text = preprocess_txt(input_text)109    print('Preprocessed After :-',preprocessed_text)110    111 112    if not preprocessed_text.strip():113        return jsonify({114            'predicted_topic' : 'N/A',115            'topic_id' : -1,116            'confidence' : 0.00,117            'message' : 'Input text was filtered out during preprocessing. Try a longer, more descriptive text.'118        })119    120 121    vectorized_text = vectorizer.transform([preprocessed_text])122    topic_distribution = model.transform(vectorized_text)123 124    dominant_topic_id = topic_distribution.argmax()125    topic_distribution_probability = topic_distribution.max()126 127    predicted_topic_name = topic_names.get(dominant_topic_id,'Unknown Id')128 129    top_words = get_top_words(dominant_topic_id)130 131    response = {132        'predicted_topic' : predicted_topic_name,133        'topic_id' : int(dominant_topic_id),134        'confidence' : float(topic_distribution_probability),135        'top_words' : top_words136    }137 138    return jsonify(response)139 140 141if __name__ == '__main__':142    app.run(debug=True,host='0.0.0.0')143 144 145 146 147 148 149 150