Update src/streamlit_app.py

src/streamlit_app.py

@@ -1,42 +1,63 @@
 import streamlit as st
 from transformers import pipeline
+from pypdf import PdfReader
 
 # --- PAGE SETUP ---
-st.set_page_config(page_title="Hugging Face Sentiment AI", page_icon="🤗")
+st.set_page_config(page_title="Twitter Sentiment AI", page_icon="🐦")
 
 # --- MODEL LOADING (Cached) ---
 @st.cache_resource
-def load_sentiment_model():
-    # This uses the default 'distilbert-base-uncased-finetuned-sst-2-english'
-    return pipeline("sentiment-analysis")
+def load_twitter_model():
+    # Using the specialized Twitter RoBERTa model
+    model_path = "cardiffnlp/twitter-roberta-base-sentiment-latest"
+    return pipeline("sentiment-analysis", model=model_path, tokenizer=model_path)
 
-classifier = load_sentiment_model()
+classifier = load_twitter_model()
+
+def get_pdf_text(pdf_file):
+    reader = PdfReader(pdf_file)
+    text = ""
+    for page in reader.pages:
+        text += page.extract_text()
+    return text
 
 # --- UI ELEMENTS ---
-st.title("🤗 AI Sentiment Analyzer")
-st.write("This app uses a Hugging Face Transformer model to detect sentiment.")
+st.title("🐦 Twitter-RoBERTa Sentiment AI")
+st.markdown("Analyze sentiment using a model trained on **124M+ tweets**.")
+
+# Tabbed interface for Input
+tab1, tab2 = st.tabs(["💬 Text Input", "📄 PDF Upload"])
+
+with tab1:
+    user_input = st.text_area("Paste your text or tweet:", height=150)
 
-user_input = st.text_area("Enter text to analyze:", placeholder="I am so excited to build this app!")
+with tab2:
+    uploaded_file = st.file_uploader("Upload a PDF file", type="pdf")
+    if uploaded_file:
+        with st.spinner("Extracting text from PDF..."):
+            user_input = get_pdf_text(uploaded_file)
+            st.info(f"Extracted {len(user_input)} characters from PDF.")
 
 if st.button("Analyze Sentiment"):
     if user_input.strip():
-        with st.spinner("Analyzing..."):
-            # Run prediction
-            results = classifier(user_input)
-            
-            # Extract data
+        with st.spinner("Analyzing with RoBERTa..."):
+            # The Twitter model outputs labels like 'positive', 'neutral', 'negative'
+            results = classifier(user_input[:2000]) # Truncate to avoid errors on huge PDFs
             label = results[0]['label']
             score = results[0]['score']
             
-            # --- DISPLAY RESULTS ---
             st.divider()
-            if label == "POSITIVE":
-                st.success(f"### {label} 😊")
+            
+            # Custom formatting based on labels
+            if label.lower() == "positive":
+                st.success(f"### Result: {label.upper()} 😊")
+            elif label.lower() == "negative":
+                st.error(f"### Result: {label.upper()} 😡")
             else:
-                st.error(f"### {label} 😡")
+                st.warning(f"### Result: {label.upper()} 😐")
                 
-            st.metric(label="Confidence Score", value=f"{score:.2%}")
+            st.metric(label="Confidence Level", value=f"{score:.2%}")
     else:
-        st.warning("Please enter some text first!")
+        st.warning("Please provide some text or a PDF file.")
 
-st.caption("Running on Hugging Face Spaces with Streamlit")
+st.caption("Note: Large PDFs are truncated to the first 2000 characters for processing speed.")