Update app.py

app.py

@@ -2,7 +2,7 @@ import streamlit as st
 import torch
 from transformers import AutoTokenizer, AutoModelForSequenceClassification
 import matplotlib.pyplot as plt
-import numpy  # Importing numpy
+import numpy as np
 
 @st.cache_resource
 def load_model():
@@ -12,9 +12,11 @@ def load_model():
 
 tokenizer, model = load_model()
 
-st.title("Sentiment Analysis App")
+st.title("Advanced Sentiment Analysis App")
 
 text = st.text_input("Enter text to analyze:")
+threshold = st.slider("Set sentiment strength threshold:", 0.0, 1.0, 0.5, 0.01)
+
 if st.button("Analyze") and text:
     encoding = tokenizer.encode_plus(text, return_tensors="pt", padding=True, truncation=True)
     input_ids = encoding["input_ids"]
@@ -23,19 +25,31 @@ if st.button("Analyze") and text:
     with torch.no_grad():
         output = model(input_ids, attention_mask)
     logits = output.logits.squeeze()
-    
-    # Determine the number of sentiment classes from the model output
+
     num_classes = logits.shape[0]
     sentiments = ["Very Negative", "Negative", "Neutral", "Positive", "Very Positive"][:num_classes]
 
+    softmax = torch.nn.Softmax(dim=0)
+    probabilities = softmax(logits).numpy()
+
     prediction = int(torch.argmax(logits))
     sentiment = sentiments[prediction]
-    st.write(f"Sentiment: {sentiment}")
+    st.write(f"Detected Sentiment: {sentiment}")
 
-    values = logits.tolist()
+    # Normalize scores for display
+    values = probabilities.tolist()
 
     fig, ax = plt.subplots()
-    ax.bar(sentiments, values, color=plt.cm.viridis_r(numpy.linspace(0.3, 0.7, num_classes)))
-    ax.set_title("Sentiment Analysis Scores")
-    ax.set_ylabel("Score")
-    st.pyplot(fig)
+    colors = plt.cm.coolwarm(np.linspace(0, 1, num_classes))
+    bars = ax.bar(sentiments, values, color=colors)
+
+    # Highlight bars that pass the threshold
+    for bar, value in zip(bars, values):
+        if value > threshold:
+            bar.set_alpha(1.0)  # Solid color for high confidence
+        else:
+            bar.set_alpha(0.5)  # Faded color for low confidence
+
+    ax.set_title("Sentiment Analysis Scores with Confidence Threshold")
+    ax.set_ylabel("Confidence")
+    st.pyplot(fig)