analyzer.py

from flask import Flask, request, jsonify
import os
import re
import json
import nltk
from nltk.stem import WordNetLemmatizer
from nltk.corpus import stopwords
import joblib
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.naive_bayes import MultinomialNB
from sklearn.pipeline import Pipeline

nltk.download('punkt', download_dir='/tmp/nltk_data')
nltk.download('stopwords', download_dir='/tmp/nltk_data')
nltk.download('wordnet', download_dir='/tmp/nltk_data')

nltk.data.path.append('/tmp/nltk_data')

app = Flask(__name__)

# Function to preprocess text data
def clean_text_with_lemmatization(texts):
    stopwords_set = set(stopwords.words('english'))
    lemmatizer = WordNetLemmatizer()
    cleaned_texts = []
    for text in texts:
        text = text.lower()
        text = re.sub(r'[^\w\s]', ' ', text)
        words = [lemmatizer.lemmatize(word) for word in text.split() if word not in stopwords_set]
        cleaned_texts.append(' '.join(words))
    return cleaned_texts

# Function to determine sentiment label based on probability
def get_sentiment_label(prob):
    if prob < 0.30:
        return "Negative"
    elif 0.35 <= prob < 0.5:
        return "Slightly Negative"
    elif prob == 0.5:
        return "Neutral"
    elif 0.5 < prob < 0.70:
        return "Slightly Positive"
    else:
        return "Positive"

# Function to load or train the model
def load_model():
    model_path = 'sentiment_pipeline.pkl'
    if os.path.exists(model_path):
        return joblib.load(model_path)
    else:
        return train_model('trainData.json')

# Function to train the model
def train_model(json_file_path):
    with open(json_file_path, 'r') as file:
        data = json.load(file)
    X = [entry['text'] for entry in data]
    y = [entry['label'] for entry in data]
    pipeline = Pipeline([
        ('vectorizer', CountVectorizer()),
        ('classifier', MultinomialNB())
    ])
    pipeline.fit(X, y)
    joblib.dump(pipeline, 'sentiment_pipeline.pkl')
    return pipeline

# Endpoint to process new reviews
@app.route('/', methods=['POST'])
def predict_sentiment():
    pipeline = load_model()
    new_reviews_json = request.json
    new_reviews = [review['CUSTOMERREVIEWS'] for review in new_reviews_json['reviewsModel']]
    cleaned_new_reviews = clean_text_with_lemmatization(new_reviews)
    predicted_probabilities = pipeline.predict_proba(cleaned_new_reviews)
    results = []
    for i, review_info in enumerate(new_reviews_json['reviewsModel']):
        original_review = review_info['CUSTOMERREVIEWS']
        negative_probability_new = predicted_probabilities[i, 0]
        positive_probability_new = predicted_probabilities[i, 1]
        sentiment = get_sentiment_label(positive_probability_new)
        formatted_negative_probability = f"{negative_probability_new:.4f}"
        formatted_positive_probability = f"{positive_probability_new:.4f}"
        results.append({
            "review_text": original_review,
            "negative_probability": formatted_negative_probability,
            "positive_probability": formatted_positive_probability,
            "sentiment": sentiment
        })
    overall_negative_probability = predicted_probabilities.mean(axis=0)[0]
    overall_positive_probability = predicted_probabilities.mean(axis=0)[1]
    overall_sentiment = get_sentiment_label(overall_positive_probability)
    results.append({
        "overall_negative_probability": f"{overall_negative_probability:.4f}",
        "overall_positive_probability": f"{overall_positive_probability:.4f}",
        "overall_sentiment": overall_sentiment
    })
    return jsonify(results)

#if __name__ == '__main__':
#    app.run(debug=True)