app.py

# import streamlit as st
# from transformers import pipeline
# import re

# # Load the model
# classifier = pipeline("text-classification", model="Mpavan45/Telugu_Sentimental_Analysis")

# # Label mapping and emojis
# labels = ["neutral", "positive", "negative"]
# emojis = {"positive": "🤗", "negative": "😔", "neutral": "😐"}

# # Background and Title Styling
# st.markdown("""
#     <style>
#     .stApp {
#         background-image: url('https://cdn-uploads.huggingface.co/production/uploads/675fab3a2d0851e23d23cad3/_YKXYHCbjM44ubGwnAKeQ.jpeg');
#         background-size: cover;
#         background-position: center;
#         background-repeat: no-repeat;
#         background-attachment: fixed;
#     }
#     .radium-title {
#         font-size: 40px;
#         text-align: center;
#         color: #fff;
#         padding: 10px;
#         border-radius: 10px;
#         background: linear-gradient(90deg, #8E2DE2, #4A00E0);
#         box-shadow: 0 0 20px #8E2DE2, 0 0 30px #4A00E0;
#     }
#     .radium-label {
#         font-size: 28px;
#         font-weight: bold;
#         color: white;
#         padding: 10px 20px;
#         border-radius: 12px;
#         background: linear-gradient(90deg, #7F00FF, #E100FF);
#         display: inline-block;
#         margin-top: 20px;
#     }
#     </style>
# """, unsafe_allow_html=True)

# # Check Telugu text validity
# def is_mostly_telugu(text):
#     if not text.strip():
#         return False
#     telugu_pattern = r'[\u0C00-\u0C7F]'
#     allowed_pattern = r'[a-zA-Z0-9\s.,!?]'
#     telugu_chars = len(re.findall(telugu_pattern, text))
#     allowed_chars = len(re.findall(allowed_pattern, text))
#     total_chars = len(text)
#     telugu_ratio = telugu_chars / total_chars if total_chars > 0 else 0
#     valid_chars = telugu_chars + allowed_chars == total_chars
#     return telugu_ratio >= 0.7 and valid_chars

# # Clean text
# def clean_input(text):
#     cleaned_text = re.sub(r'[^a-zA-Z0-9\u0C00-\u0C7F\s?.!]', ' ', text)
#     cleaned_text = re.sub(r'([?.!])(?![?.!]\s|$)', '', cleaned_text)
#     return ' '.join(cleaned_text.split())

# # Title
# st.markdown('<div class="radium-title">Telugu Sentiment Analysis</div>', unsafe_allow_html=True)

# # Input
# user_input = st.text_area("Enter your Telugu text:")

# if st.button("Predict"):
#     if not user_input.strip():
#         st.warning("Please enter some Telugu text.")
#     else:
#         cleaned = clean_input(user_input)
#         if not is_mostly_telugu(cleaned):
#             st.error("Please enter text primarily in Telugu script.")
#         else:
#             result = classifier(cleaned)[0]
#             label = result['label']
#             try:
#                 index = int(label.split('_')[-1])
#                 sentiment = labels[index]
#             except (ValueError, IndexError):
#                 sentiment = label.lower() if label.lower() in labels else "neutral"
            
#             # Display sentiment with emoji using radium label style
#             sentiment_display = f'{sentiment.capitalize()} {emojis.get(sentiment, "")}'
#             st.markdown(f'<div class="radium-label">{sentiment_display}</div>', unsafe_allow_html=True)
import streamlit as st
from transformers import pipeline
import re

# Load the model
classifier = pipeline("text-classification", model="Mpavan45/Telugu_Sentimental_Analysis")

# Label mapping and emojis
labels = ["neutral", "positive", "negative"]
emojis = {"positive": "🤗", "negative": "😔", "neutral": "😐"}

# UI Styling
st.markdown("""
    <style>
    .stApp {
        background-image: url('https://cdn-uploads.huggingface.co/production/uploads/675fab3a2d0851e23d23cad3/_YKXYHCbjM44ubGwnAKeQ.jpeg');
        background-size: cover;
        background-position: center;
        background-repeat: no-repeat;
        background-attachment: fixed;
        font-family: 'Segoe UI', sans-serif;
    }

    .radium-title {
        font-size: 30px;
        text-align: center;
        color: #fff;
        padding: 10px;
        border-radius: 10px;
        background: linear-gradient(90deg, #8E2DE2, #4A00E0);
        box-shadow: 0 0 20px #8E2DE2, 0 0 30px #4A00E0;
        margin-bottom: 20px;
    }

    .radium-label {
        font-size: 28px;
        font-weight: bold;
        color: white;
        padding: 10px 20px;
        border-radius: 12px;
        background: linear-gradient(90deg, #7F00FF, #E100FF);
        display: inline-block;
        margin-top: 20px;
    }

    .radium-button > button {
        font-size: 20px !important;
        font-weight: bold !important;
        color: white !important;
        border: none !important;
        padding: 12px 28px !important;
        border-radius: 12px !important;
        background: linear-gradient(90deg, #8E2DE2, #4A00E0) !important;
        box-shadow: 0 0 10px #8E2DE2, 0 0 20px #4A00E0;
        transition: all 0.3s ease-in-out;
    }

    .radium-button > button:hover {
        box-shadow: 0 0 20px #fff, 0 0 30px #8E2DE2;
        transform: scale(1.05);
    }

    textarea {
        font-size: 20px !important;
        line-height: 1.5 !important;
        padding: 10px !important;
    }

    .example-box {
        background-color: rgba(255, 255, 255, 0.1);
        color: white;
        padding: 10px 15px;
        border-radius: 10px;
        font-size: 15px;
        margin-bottom: 10px;
    }
    </style>
""", unsafe_allow_html=True)

# Title
st.markdown('<div class="radium-title">Telugu Sentiment Analysis</div>', unsafe_allow_html=True)

# Functions
def is_mostly_telugu(text):
    if not text.strip():
        return False
    telugu_pattern = r'[\u0C00-\u0C7F]'
    allowed_pattern = r'[a-zA-Z0-9\s.,!?]'
    telugu_chars = len(re.findall(telugu_pattern, text))
    allowed_chars = len(re.findall(allowed_pattern, text))
    total_chars = len(text)
    telugu_ratio = telugu_chars / total_chars if total_chars > 0 else 0
    valid_chars = telugu_chars + allowed_chars == total_chars
    return telugu_ratio >= 0.7 and valid_chars

def clean_input(text):
    cleaned_text = re.sub(r'[^a-zA-Z0-9\u0C00-\u0C7F\s?.!]', ' ', text)
    cleaned_text = re.sub(r'([?.!])(?![?.!]\s|$)', '', cleaned_text)
    return ' '.join(cleaned_text.split())

# Show examples
st.markdown("### 📝 You can only enter pure Telugu text. Try one of the examples below if you'd like:")
st.markdown('<div class="example-box">ఆమెతో మాట్లాడిన తర్వాత నా మనసు తేలికపడింది.</div>', unsafe_allow_html=True)
st.markdown('<div class="example-box">ఈ రోజు నేను చాలా నిరాశతో ఉన్నాను. ఏది కూడా సరిగ్గా జరగడం లేదు.</div>', unsafe_allow_html=True)

# Input
user_input = st.text_area(" ", height=180, key="input_box")

# Buttons
col1, col2 = st.columns(2)

with col1:
    if st.markdown('<div class="radium-button">', unsafe_allow_html=True):
        if st.button("🔮 Predict"):
            if not user_input.strip():
                st.warning("Please enter some Telugu text.")
            else:
                cleaned = clean_input(user_input)
                if not is_mostly_telugu(cleaned):
                    st.error("Please enter text primarily in Telugu script.")
                else:
                    result = classifier(cleaned)[0]
                    label = result['label']
                    try:
                        index = int(label.split('_')[-1])
                        sentiment = labels[index]
                    except (ValueError, IndexError):
                        sentiment = label.lower() if label.lower() in labels else "neutral"

                    sentiment_display = f'{sentiment.capitalize()} {emojis.get(sentiment, "")}'
                    st.markdown(f'<div class="radium-label">{sentiment_display}</div>', unsafe_allow_html=True)


# Clear button with session state reset
with col2:
    if st.markdown('<div class="radium-button">', unsafe_allow_html=True):
        if st.button("🧹 Clear"):
            st.session_state.input_box = ""