Update app.py
Browse files
app.py
CHANGED
|
@@ -0,0 +1,127 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import streamlit as st
|
| 2 |
+
from transformers import pipeline
|
| 3 |
+
import re
|
| 4 |
+
|
| 5 |
+
st.set_page_config(page_title="Hindi Sentiment Analysis", layout="centered")
|
| 6 |
+
|
| 7 |
+
# Custom CSS for styling
|
| 8 |
+
st.markdown("""
|
| 9 |
+
<style>
|
| 10 |
+
@import url('https://fonts.googleapis.com/css2?family=Poppins&display=swap');
|
| 11 |
+
|
| 12 |
+
html, body, [class*="css"] {
|
| 13 |
+
font-family: 'Poppins', sans-serif;
|
| 14 |
+
text-align: center;
|
| 15 |
+
}
|
| 16 |
+
|
| 17 |
+
.centered-title {
|
| 18 |
+
font-size: 50px;
|
| 19 |
+
font-weight: bold;
|
| 20 |
+
margin-bottom: 10px;
|
| 21 |
+
}
|
| 22 |
+
|
| 23 |
+
.sentiment-result {
|
| 24 |
+
font-size: 26px;
|
| 25 |
+
font-weight: 600;
|
| 26 |
+
margin-top: 20px;
|
| 27 |
+
}
|
| 28 |
+
|
| 29 |
+
.confidence {
|
| 30 |
+
font-size: 20px;
|
| 31 |
+
margin-top: 10px;
|
| 32 |
+
}
|
| 33 |
+
|
| 34 |
+
.original-text {
|
| 35 |
+
font-style: italic;
|
| 36 |
+
margin-top: 10px;
|
| 37 |
+
}
|
| 38 |
+
|
| 39 |
+
textarea {
|
| 40 |
+
text-align: left !important;
|
| 41 |
+
}
|
| 42 |
+
</style>
|
| 43 |
+
""", unsafe_allow_html=True)
|
| 44 |
+
|
| 45 |
+
# Load model
|
| 46 |
+
pipe = pipeline("text-classification", model="NeonSamurai/hindi_bert_model")
|
| 47 |
+
|
| 48 |
+
names = ["neutral", "positive", "negative"]
|
| 49 |
+
emojis = {"positive": "🤗", "negative": "😔", "neutral": "😐"}
|
| 50 |
+
|
| 51 |
+
# Utility functions
|
| 52 |
+
def is_mostly_hindi(text):
|
| 53 |
+
if not text.strip():
|
| 54 |
+
return False
|
| 55 |
+
devanagari_pattern = r'[\u0900-\u097F]'
|
| 56 |
+
allowed_pattern = r'[a-zA-Z0-9\s.,!?]'
|
| 57 |
+
devanagari_chars = len(re.findall(devanagari_pattern, text))
|
| 58 |
+
allowed_chars = len(re.findall(allowed_pattern, text))
|
| 59 |
+
total_chars = len(text)
|
| 60 |
+
hindi_proportion = devanagari_chars / total_chars if total_chars > 0 else 0
|
| 61 |
+
valid_chars = devanagari_chars + allowed_chars == total_chars
|
| 62 |
+
return hindi_proportion >= 0.7 and valid_chars
|
| 63 |
+
|
| 64 |
+
def clean_input(text):
|
| 65 |
+
cleaned_text = re.sub(r'[^a-zA-Z0-9\u0900-\u097F\s?.!]', ' ', text)
|
| 66 |
+
cleaned_text = re.sub(r'([?.!])(?![?.!]\s|$)', '', cleaned_text)
|
| 67 |
+
cleaned_text = ' '.join(cleaned_text.split())
|
| 68 |
+
return cleaned_text
|
| 69 |
+
|
| 70 |
+
# Title
|
| 71 |
+
st.markdown("<div class='centered-title'>📊 Hindi Sentiment Analysis</div>", unsafe_allow_html=True)
|
| 72 |
+
|
| 73 |
+
# Instructions
|
| 74 |
+
with st.expander("ℹ️ Instructions"):
|
| 75 |
+
st.markdown("""
|
| 76 |
+
- Please enter a sentence or paragraph in **Hindi (Devanagari script)**.
|
| 77 |
+
- Example: *यह फिल्म बहुत अच्छी थी और अभिनय शानदार था।*
|
| 78 |
+
- The app will classify the text as **positive**, **negative**, or **neutral**.
|
| 79 |
+
""")
|
| 80 |
+
|
| 81 |
+
# Text input
|
| 82 |
+
user_input = st.text_area("✍️ Enter Hindi text below:", height=150, placeholder="Type your Hindi text here...")
|
| 83 |
+
|
| 84 |
+
# Predict button
|
| 85 |
+
if st.button("🔍 Predict Sentiment"):
|
| 86 |
+
if not user_input.strip():
|
| 87 |
+
st.warning("⚠️ Please enter some text.")
|
| 88 |
+
else:
|
| 89 |
+
cleaned_input = clean_input(user_input)
|
| 90 |
+
if not is_mostly_hindi(cleaned_input):
|
| 91 |
+
st.error("❌ Input should be primarily in Hindi (Devanagari script).")
|
| 92 |
+
else:
|
| 93 |
+
result = pipe(cleaned_input)[0]
|
| 94 |
+
sentiment_index = int(result['label'].split("_")[1])
|
| 95 |
+
sentiment = names[sentiment_index]
|
| 96 |
+
emoji = emojis[sentiment]
|
| 97 |
+
|
| 98 |
+
st.markdown(f"<div class='sentiment-result'>✅ Sentiment: <b>{sentiment.capitalize()} {emoji}</b></div>", unsafe_allow_html=True)
|
| 99 |
+
st.markdown(f"<div class='confidence'>Confidence Score: <code>{result['score']:.2f}</code></div>", unsafe_allow_html=True)
|
| 100 |
+
st.markdown(f"<div class='original-text'>“{user_input}”</div>", unsafe_allow_html=True)
|
| 101 |
+
|
| 102 |
+
# Background-Image
|
| 103 |
+
st.markdown(
|
| 104 |
+
"""
|
| 105 |
+
<style>
|
| 106 |
+
.stApp {
|
| 107 |
+
background-image: url("https://cdn-uploads.huggingface.co/production/uploads/673f5e166c2774fcc8a82f0b/OIaLFHTZGXeTvsizCCNsK.png");
|
| 108 |
+
background-size: cover;
|
| 109 |
+
background-position: center;
|
| 110 |
+
height: 100vh;
|
| 111 |
+
}
|
| 112 |
+
|
| 113 |
+
/* Semi-transparent overlay */
|
| 114 |
+
.stApp::before {
|
| 115 |
+
content: "";
|
| 116 |
+
position: absolute;
|
| 117 |
+
top: 0;
|
| 118 |
+
left: 0;
|
| 119 |
+
width: 100%;
|
| 120 |
+
height: 100%;
|
| 121 |
+
background: rgba(0, 0, 0, 0.4); /* 40% transparency */
|
| 122 |
+
z-index: -1;
|
| 123 |
+
}
|
| 124 |
+
</style>
|
| 125 |
+
""",
|
| 126 |
+
unsafe_allow_html=True
|
| 127 |
+
)
|