Update app.py

app.py

@@ -65,12 +65,16 @@
 import streamlit as st
 from transformers import pipeline
 import re
+import nltk
 from collections import Counter
 
 # Load the sentiment analysis model
 classifier = pipeline("text-classification", model="Mpavan45/Telugu_Sentimental_Analysis")
 
-# Styling
+# Download NLTK tokenizer
+nltk.download('punkt_tab')
+
+# CSS styling
 st.markdown("""
     <style>
     .radium-title {
@@ -82,76 +86,51 @@ st.markdown("""
         background: linear-gradient(90deg, #ff416c, #ff4b2b);
         box-shadow: 0 0 20px #ff416c, 0 0 30px #ff4b2b;
     }
-    .summary-label {
-        font-size: 26px;
+    .radium-label {
+        font-size: 24px;
         font-weight: bold;
         color: white;
-        padding: 12px;
-        border-radius: 10px;
-        background: linear-gradient(90deg, #00b09b, #96c93d);
-        text-align: center;
-        margin-top: 20px;
+        padding: 10px;
+        border-radius: 8px;
+        background: linear-gradient(90deg, #36d1dc, #5b86e5);
+        display: inline-block;
+        margin-top: 10px;
     }
     </style>
 """, unsafe_allow_html=True)
 
-st.markdown('<div class="radium-title">Telugu Paragraph Sentiment Analyzer</div>', unsafe_allow_html=True)
-st.write("This app analyzes a **Telugu paragraph**, splits it into sentences, and predicts the **overall sentiment** based on those sentences.")
+st.markdown('<div class="radium-title">Sentiment Analysis with BERT</div>', unsafe_allow_html=True)
+st.write("This app uses a fine-tuned BERT model to classify **Telugu text** as Positive, Negative, or Neutral.")
 
-# Label map
+# Emoji map
 label_map = {
     "LABEL_0": ("Negative", "😞"),
     "LABEL_1": ("Neutral", "😐"),
     "LABEL_2": ("Positive", "😊")
 }
 
-# Sentence splitter
-def split_telugu_text(text):
-    sentences = re.split(r'[.!?।\n]', text)
-    return [s.strip() for s in sentences if s.strip()]
-
-# Telugu validation
+# Telugu validation function
 def is_telugu_text(text):
     cleaned = re.sub(r'[\u0C00-\u0C7F\s\d\-–—\.,!@#\$%\^&\*\(\)\[\]\{\}\'\"“”…‘’<>?/|]+', '', text)
     return len(cleaned.strip()) == 0
 
-# Streamlit session
-if "text_input" not in st.session_state:
-    st.session_state.text_input = ""
-if "result_shown" not in st.session_state:
-    st.session_state.result_shown = False
+# Overall sentiment from paragraph
+def get_overall_sentiment(paragraph):
+    sentences = nltk.sent_tokenize(paragraph)
+    sentiments = [classifier(sent)[0]['label'] for sent in sentences]
+    most_common = Counter(sentiments).most_common(1)[0][0]
+    sentiment, emoji = label_map.get(most_common, (most_common, ""))
+    return sentiment, emoji
 
 # Text input
-text_input = st.text_area("Enter a Telugu paragraph:", value=st.session_state.text_input, height=200)
-
-if st.button("Analyze Paragraph"):
-    if is_telugu_text(text_input):
-        st.session_state.text_input = text_input
-        st.session_state.result_shown = True
-
-        telugu_sentences = split_telugu_text(text_input)
-        sentence_sentiments = []
-
-        for sent in telugu_sentences:
-            result = classifier(sent)
-            label = result[0]['label']
-            sentiment, _ = label_map.get(label, (label, ""))
-            sentence_sentiments.append(sentiment)
-
-        sentiment_counts = Counter(sentence_sentiments)
-        overall = sentiment_counts.most_common(1)[0][0]
-        emoji = [v[1] for k, v in label_map.items() if v[0] == overall][0]
-
-        st.session_state.overall_sentiment = overall
-        st.session_state.overall_emoji = emoji
+text_input = st.text_area("Enter Telugu text to analyze sentiment:", height=150)
 
+# Analyze sentiment
+if st.button("Analyze Sentiment"):
+    if is_telugu_text(text_input) and text_input.strip():
+        sentiment, emoji = get_overall_sentiment(text_input)
+        st.markdown(f'<div class="radium-label">{sentiment} {emoji}</div>', unsafe_allow_html=True)
     else:
-        st.session_state.result_shown = False
         st.warning("Please enter valid Telugu text only (no English or symbols)!")
 
-# Display overall paragraph sentiment only
-if st.session_state.result_shown:
-    overall = st.session_state.overall_sentiment
-    emoji = st.session_state.overall_emoji
-    st.markdown(f'<div class="summary-label" Sentiment: {overall} {emoji}</div>', unsafe_allow_html=True)