Update app.py

app.py

@@ -52,7 +52,6 @@ def clean_text(text: str) -> str:
     text = re.sub(r'\s+([,.;:?!])', r'\1', text)
     return text
 
-# --- Tokenizer Normalizer Configuration ---
 newline_to_space = Replace(Regex(r'\s*\n\s*'), " ")
 join_hyphen_break = Replace(Regex(r'(\w+)[--]\s*\n\s*(\w+)'), r"\1\2")
 tokenizer.backend_tokenizer.normalizer = Sequence([
@@ -82,27 +81,22 @@ def classify_text(text):
         logits_2 = model_2(**inputs).logits
         logits_3 = model_3(**inputs).logits
 
-        # Apply softmax to get probabilities
         softmax_1 = torch.softmax(logits_1, dim=1)
         softmax_2 = torch.softmax(logits_2, dim=1)
         softmax_3 = torch.softmax(logits_3, dim=1)
 
-        # Average the probabilities from the three models
         averaged_probabilities = (softmax_1 + softmax_2 + softmax_3) / 3
         probabilities = averaged_probabilities[0]
 
-    # --- Generate Text Result ---
     human_prob = probabilities[24].item()
     ai_probs_clone = probabilities.clone()
-    ai_probs_clone[24] = 0  # Exclude human probability for AI total
+    ai_probs_clone[24] = 0
     ai_total_prob = ai_probs_clone.sum().item()
 
-    # Normalize probabilities to get a percentage-based decision
     total_decision_prob = human_prob + ai_total_prob
     human_percentage = (human_prob / total_decision_prob) * 100
     ai_percentage = (ai_total_prob / total_decision_prob) * 100
 
-    # Determine the most likely AI model
     ai_argmax_index = torch.argmax(ai_probs_clone).item()
     ai_argmax_model = label_mapping[ai_argmax_index]
 
@@ -116,38 +110,33 @@ def classify_text(text):
             f"**Identified LLM: {ai_argmax_model}**"
         )
 
-    # --- Generate Plot ---
-    # Find the top 5 AI models by probability
+
     ai_probs_for_plot = probabilities.clone()
-    ai_probs_for_plot[24] = -1 # Ensure 'human' isn't in the top 5 AI list
     top_5_probs, top_5_indices = torch.topk(ai_probs_for_plot, 5)
 
-    # Prepare data for plotting
     top_5_probs = top_5_probs.cpu().numpy()
     top_5_labels = [label_mapping[i.item()] for i in top_5_indices]
 
-    # Create a horizontal bar plot
     fig, ax = plt.subplots(figsize=(10, 5))
     bars = ax.barh(top_5_labels, top_5_probs, color='#4CAF50', alpha=0.8)
     ax.set_xlabel('Probability', fontsize=12)
     ax.set_title('Top 5 Predicted AI Models', fontsize=14, fontweight='bold')
-    ax.invert_yaxis()  # Highest probability on top
+    ax.invert_yaxis()
     ax.grid(axis='x', linestyle='--', alpha=0.6)
 
-    # Add percentage labels on the bars
+    
     for bar in bars:
         width = bar.get_width()
         label_x_pos = width + 0.01
         ax.text(label_x_pos, bar.get_y() + bar.get_height() / 2, f'{width:.2%}', va='center')
 
-    ax.set_xlim(0, max(top_5_probs) * 1.18)  # Adjust x-axis limit for labels
+    ax.set_xlim(0, max(top_5_probs) * 1.18) 
     plt.tight_layout()
 
-    # Return both the text message and the plot figure
+    
     return result_message, fig
 
 
-# --- Gradio Interface Definition ---
 
 title = "AI Text Detector"
 
@@ -165,7 +154,6 @@ Paste your text below to analyze its origin.
 """
 bottom_text = "**Developed by SzegedAI**"
 
-# Example texts
 AI_texts = [
 "Camels are remarkable desert animals known for their unique adaptations to harsh, arid environments. Native to the Middle East, North Africa, and parts of Asia, camels have been essential to human life for centuries, serving as a mode of transportation, a source of food, and even a symbol of endurance and survival. There are two primary species of camels: the dromedary camel, which has a single hump and is commonly found in the Middle East and North Africa, and the Bactrian camel, which has two humps and is native to Central Asia. Their humps store fat, not water, as commonly believed, allowing them to survive long periods without food by metabolizing the stored fat for energy. Camels are highly adapted to desert life. They can go for weeks without water, and when they do drink, they can consume up to 40 gallons in one sitting. Their thick eyelashes, sealable nostrils, and wide, padded feet protect them from sand and help them walk easily on loose desert terrain.",
 "Wines are a fascinating reflection of culture, history, and craftsmanship. They embody a rich diversity shaped by the land, climate, and traditions where they are produced. From the bold reds of Bordeaux to the crisp whites of New Zealand, each bottle tells a unique story. What makes wine so special is its ability to connect people. Whether shared at a family dinner, a celebratory event, or a quiet evening with friends, wine enhances experiences and brings people together. The variety of flavors and aromas, influenced by grape type, fermentation techniques, and aging processes, make wine tasting a complex yet rewarding journey for the senses.",
@@ -177,7 +165,6 @@ Human_texts = [
 "Fats are rich in energy, build body cells, support brain development of infants, help body processes, and facilitate the absorption and use of fat-soluble vitamins A, D, E, and K. The major component of lipids is glycerol and fatty acids. According to chemical properties, fatty acids can be divided into saturated and unsaturated fatty acids. Generally lipids containing saturated fatty acids are solid at room temperature and include animal fats (butter, lard, tallow, ghee) and tropical oils (palm,coconut, palm kernel). Saturated fats increase the risk of heart disease.",
 "To make BERT handle a variety of down-stream tasks, our input representation is able to unambiguously represent both a single sentence and a pair of sentences (e.g., h Question, Answeri) in one token sequence. Throughout this work, a “sentence” can be an arbitrary span of contiguous text, rather than an actual linguistic sentence. A “sequence” refers to the input token sequence to BERT, which may be a single sentence or two sentences packed together. We use WordPiece embeddings (Wu et al., 2016) with a 30,000 token vocabulary. The first token of every sequence is always a special classification token ([CLS]). The final hidden state corresponding to this token is used as the aggregate sequence representation for classification tasks. Sentence pairs are packed together into a single sequence."]
 
-# Define the Gradio interface with CSS styling
 iface = gr.Blocks(css="""
     @import url('https://fonts.googleapis.com/css2?family=Roboto+Mono:wght@400;700&display=swap');
     #text_input_box { border-radius: 10px; border: 2px solid #4CAF50; font-size: 18px; padding: 15px; margin-bottom: 20px; width: 60%; box-sizing: border-box; margin: auto; }