Add Gradio app

app.py

@@ -0,0 +1,190 @@
+# app.py
+import gradio as gr
+import tensorflow as tf
+from tensorflow.keras.models import load_model
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+import numpy as np
+import json
+import pickle
+import nltk
+from nltk.tokenize import word_tokenize
+from nltk.stem import WordNetLemmatizer
+import re
+import string
+from huggingface_hub import hf_hub_download
+
+# Download NLTK resources
+nltk.download('punkt', quiet=True)
+nltk.download('wordnet', quiet=True)
+nltk.download('omw-1.4', quiet=True)
+
+# Initialize lemmatizer
+lemmatizer = WordNetLemmatizer()
+
+# Define LuongAttention (placeholder for loading)
+class LuongAttention(tf.keras.layers.Layer):
+    def __init__(self, **kwargs):
+        super(LuongAttention, self).__init__(**kwargs)
+
+    def build(self, input_shape):
+        self.W = self.add_weight(
+            name='attention_weight',
+            shape=(input_shape[-1], input_shape[-1]),
+            initializer='glorot_uniform',
+            trainable=True
+        )
+        self.b = self.add_weight(
+            name='attention_bias',
+            shape=(input_shape[-1],),
+            initializer='zeros',
+            trainable=True
+        )
+        super(LuongAttention, self).build(input_shape)
+
+    def call(self, inputs):
+        lstm_output = inputs
+        score = tf.matmul(lstm_output, self.W) + self.b
+        score = tf.tanh(score)
+        attention_weights = tf.nn.softmax(score, axis=1)
+        context = lstm_output * attention_weights
+        context = tf.reduce_sum(context, axis=1)
+        return context, attention_weights
+
+    def get_config(self):
+        config = super(LuongAttention, self).get_config()
+        return config
+
+# Load model, tokenizer, label encoder from Hugging Face Hub
+model_path = hf_hub_download(repo_id="logasanjeev/sentiment-analysis-bilstm-luong", filename="sentiment_model.h5")
+tokenizer_path = hf_hub_download(repo_id="logasanjeev/sentiment-analysis-bilstm-luong", filename="tokenizer.json")
+encoder_path = hf_hub_download(repo_id="logasanjeev/sentiment-analysis-bilstm-luong", filename="label_encoder.pkl")
+model = load_model(model_path, custom_objects={"LuongAttention": LuongAttention})
+with open(tokenizer_path, "r") as f:
+    tokenizer = tf.keras.preprocessing.text.tokenizer_from_json(json.load(f))
+with open(encoder_path, "rb") as f:
+    label_encoder = pickle.load(f)
+
+# Text cleaning function
+def clean_text(text):
+    if not isinstance(text, str):
+        text = str(text)
+    text = text.lower()
+    text = re.sub(r'http\S+|www\S+|https\S+', '', text, flags=re.MULTILINE)
+    text = re.sub(r'@\w+|\#\w+', '', text)
+    text = text.translate(str.maketrans('', '', string.punctuation))
+    text = re.sub(r'\d+', '', text)
+    tokens = word_tokenize(text)
+    tokens = [lemmatizer.lemmatize(token) for token in tokens]
+    return ' '.join(tokens).strip()
+
+# Prediction function
+def predict_sentiment(text):
+    if not text or not isinstance(text, str) or len(text.strip()) < 3:
+        return "Please enter a valid sentence.", None, None
+    
+    # Clean and preprocess
+    cleaned = clean_text(text)
+    seq = tokenizer.texts_to_sequences([cleaned])
+    if not seq or not any(x > 1 for x in seq[0]):
+        return "Text too short or invalid.", None, None
+    
+    # Pad sequence
+    max_len = 35
+    pad = pad_sequences(seq, maxlen=max_len, padding='post', truncating='post')
+    
+    # Predict
+    with tf.device('/CPU:0'):
+        pred = model.predict(pad, verbose=0)[0]
+    sentiment = label_encoder.inverse_transform([np.argmax(pred)])[0]
+    probs = pred.tolist()
+    
+    # Format output
+    emoji = {"negative": "😣", "neutral": "😐", "positive": "😊"}
+    probs_dict = {
+        "Negative": probs[0],
+        "Neutral": probs[1],
+        "Positive": probs[2]
+    }
+    
+    return (
+        f"**Sentiment**: {sentiment.capitalize()} {emoji[sentiment]}",
+        probs_dict,
+        cleaned
+    )
+
+# Custom CSS for slick UI
+css = """
+body { font-family: 'Arial', sans-serif; }
+.gradio-container { max-width: 800px; margin: auto; }
+h1 { color: #1a73e8; text-align: center; }
+.textbox { border-radius: 8px; }
+.output-text { font-size: 1.2em; font-weight: bold; }
+.footer { text-align: center; color: #666; }
+.prob-bar { margin-top: 10px; }
+button { border-radius: 6px; }
+.dark { background: #1e1e1e; color: #ffffff; }
+.dark h1 { color: #4a90e2; }
+.dark .footer { color: #aaa; }
+"""
+
+# Gradio interface
+with gr.Blocks(theme=gr.themes.Soft(), css=css) as demo:
+    gr.Markdown(
+        """
+        # Sentiment Analysis App
+        Predict the sentiment of your text (negative, neutral, positive) using a Bi-LSTM model with Luong attention. Try it out!
+        """
+    )
+    
+    with gr.Row():
+        with gr.Column(scale=3):
+            text_input = gr.Textbox(
+                label="Your Text",
+                placeholder="e.g., The food service is not good at all",
+                lines=2
+            )
+            predict_btn = gr.Button("Analyze Sentiment", variant="primary")
+        with gr.Column(scale=1):
+            theme_toggle = gr.Button("Toggle Theme")
+    
+    output_text = gr.Markdown()
+    prob_plot = gr.Label(label="Probability Distribution")
+    cleaned_text = gr.Textbox(label="Cleaned Text", interactive=False)
+    
+    examples = gr.Examples(
+        examples=[
+            "the food service is not good at all",
+            "this is not recommended at all",
+            "This place sucks!",
+            "I’m so happy with this!",
+            "It’s alright, I guess."
+        ],
+        inputs=text_input
+    )
+    
+    # Theme toggle logic
+    def toggle_theme():
+        return gr.themes.Dark() if demo.theme.name == "soft" else gr.themes.Soft()
+    
+    # Bind functions
+    predict_btn.click(
+        fn=predict_sentiment,
+        inputs=text_input,
+        outputs=[output_text, prob_plot, cleaned_text]
+    )
+    theme_toggle.click(
+        fn=toggle_theme,
+        outputs=None,
+        _js="() => { document.body.classList.toggle('dark'); }"
+    )
+    
+    gr.Markdown(
+        """
+        <div class='footer'>
+            Created by logasanjeev | Powered by Hugging Face & Gradio
+        </div>
+        """
+    )
+
+# Launch app
+demo.launch()