app.py

import gradio as gr
import numpy as np
from keras.models import load_model
import re
from keras.preprocessing.sequence import pad_sequences
from keras.preprocessing.text import tokenizer_from_json
import re
import joblib
import json

# Load the JSON data from the file
with open('tok.json', 'r') as json_file:
    tokenizer_json = json.load(json_file)

# Assuming tokenizer_class is the class of your tokenizer
tokenizer = tokenizer_from_json(tokenizer_json)
rnn_model = load_model("rnn_model.h5")

# setting the joblib
lr_model = joblib.load("logistic_model.pkl")
svm_model = joblib.load("svm_model.pkl")
nn_model = load_model("dl_model.h5")
mnb_model = joblib.load("mnb_model.pkl")

# Load other necessary files like vectorizers or scalers
tfidf_vectorizer = joblib.load("tfidf_vectorizer.pkl")

TAG_RE = re.compile(r'<[^>]+>')

def remove_tags(text):
    return TAG_RE.sub('', text)

def preprocess_text(sen):
    # Removing html tags
    sentence = remove_tags(sen)

    # Remove punctuations and numbers
    sentence = re.sub('[^a-zA-Z]', ' ', sentence)

    # Single character removal
    sentence = re.sub(r"\s+[a-zA-Z]\s+", ' ', sentence)

    # Removing multiple spaces
    sentence = re.sub(r'\s+', ' ', sentence)

    return sentence

def preprocess_text_for_rnn(text, tokenizer, maxlen):
    text = preprocess_text(text)
    sequence = tokenizer.texts_to_sequences([text])
    padded_sequence = pad_sequences(sequence, padding='post', maxlen=maxlen)
    return padded_sequence

def predict_lr(text):
    preprocessed_text = preprocess_text(text)
    vectorized_text = tfidf_vectorizer.transform([preprocessed_text])
    dense_vectorized_text = vectorized_text.toarray()
    prediction_probs = lr_model.predict_proba(dense_vectorized_text)[0]
    prediction = int(np.argmax(prediction_probs))
    return prediction, prediction_probs

def predict_svm(text):
    preprocessed_text = preprocess_text(text)
    vectorized_text = tfidf_vectorizer.transform([preprocessed_text])
    dense_vectorized_text = vectorized_text.toarray()
    prediction_probs = svm_model.predict_proba(dense_vectorized_text)[0]
    prediction = int(np.argmax(prediction_probs))
    return prediction, prediction_probs

def predict_nn(text):
    preprocessed_text = preprocess_text(text)
    vectorized_text = tfidf_vectorizer.transform([preprocessed_text])
    dense_vectorized_text = vectorized_text.toarray()
    prediction_probs = nn_model.predict(dense_vectorized_text)[0]
    prediction = int(np.argmax(prediction_probs))
    return prediction, prediction_probs

def predict_mnb(text):
    preprocessed_text = preprocess_text(text)
    vectorized_text = tfidf_vectorizer.transform([preprocessed_text])
    dense_vectorized_text = vectorized_text.toarray()
    prediction_probs = mnb_model.predict_proba(dense_vectorized_text)[0]
    prediction = int(np.argmax(prediction_probs))
    return prediction, prediction_probs

def predict_rnn(text):
    processed_text = preprocess_text_for_rnn(text, tokenizer, maxlen=170)
    prediction_probs = rnn_model.predict(processed_text)[0]
    prediction = int(np.argmax(prediction_probs))
    return prediction, prediction_probs

def sentiment_prediction(text, model):
    prediction, percentages = 0, []
    if model == "Logistic Regression":
        prediction, percentages = predict_lr(text)
    elif model == "SVM":
        prediction, percentages = predict_svm(text)
    elif model == "Neural Network":
        prediction, percentages = predict_nn(text)
    elif model == "Multinomial Naive Bayes":
        prediction, percentages = predict_mnb(text)
    elif model == "Recurrent Neural Network":
        prediction, percentages = predict_rnn(text)

    # Displaying emojis based on sentiment
    emoji_positive = "😃"
    emoji_negative = "😢"
    emoji = emoji_positive if prediction == 1 else emoji_negative

    # Create labels for predictions
    labels = ["Negative", "Positive"]

    # Create label for the prediction
    prediction_label = labels[prediction]

    # Calculate percentages for both labels
    percentage_negative = percentages[0]
    percentage_positive = percentages[1]

    return prediction_label, f"{labels[0]}: {percentage_negative:.2%}, Percentage {labels[1]}: {percentage_positive:.2%}", emoji

# Create the Gradio interface
iface = gr.Interface(
    fn=sentiment_prediction,
    inputs=[gr.Textbox(type="text", label="Enter Text"), gr.Dropdown(["Logistic Regression", "SVM", "Neural Network", "Multinomial Naive Bayes", "Recurrent Neural Network"], label="Select Model")],
    outputs=[gr.Label(), gr.Label(), gr.Label()],
    live=True,
    title="Sentiment Analysis with Model Selection",
    description="Enter a text and choose a model for sentiment prediction.",
)

# Launch the Gradio interface
iface.launch()