updated

app.py

@@ -116,8 +116,8 @@ def highlight_differences(original: str, generated: str) -> Tuple[str, str, Dict
             
         elif tag == 'replace':
             # Changed text
-            highlighted_original.append(f'<span style="background-color: #ffcccc; padding: 2px 4px; border-radius: 3px; text-decoration: line-through;">{original_chunk}</span>')
-            highlighted_generated.append(f'<span style="background-color: #ccffcc; padding: 2px 4px; border-radius: 3px; font-weight: 500;">{generated_chunk}</span>')
+            highlighted_original.append(f'<span style="color: #ffcccc; padding: 2px 4px; border-radius: 3px; text-decoration: line-through;">{original_chunk}</span>')
+            highlighted_generated.append(f'<span style="color: #ccffcc; padding: 2px 4px; border-radius: 3px; font-weight: 500;">{generated_chunk}</span>')
             changes_count += max(i2 - i1, j2 - j1)
             
             # Track word substitutions (limit to single word changes for clarity)
@@ -126,12 +126,12 @@ def highlight_differences(original: str, generated: str) -> Tuple[str, str, Dict
             
         elif tag == 'delete':
             # Text removed in generated
-            highlighted_original.append(f'<span style="background-color: #ffcccc; padding: 2px 4px; border-radius: 3px; text-decoration: line-through;">{original_chunk}</span>')
+            highlighted_original.append(f'<span style="color: #ffcccc; padding: 2px 4px; border-radius: 3px; text-decoration: line-through;">{original_chunk}</span>')
             deletions_count += (i2 - i1)
             
         elif tag == 'insert':
             # Text added in generated
-            highlighted_generated.append(f'<span style="background-color: #ccffcc; padding: 2px 4px; border-radius: 3px; font-weight: 500;">{generated_chunk}</span>')
+            highlighted_generated.append(f'<span style="color: #ccffcc; padding: 2px 4px; border-radius: 3px; font-weight: 500;">{generated_chunk}</span>')
             additions_count += (j2 - j1)
     
     # Join with spaces
@@ -160,49 +160,49 @@ def highlight_differences(original: str, generated: str) -> Tuple[str, str, Dict
     return final_original, final_generated, statistics
 
 def format_statistics(stats: Dict) -> str:
-    """Format statistics into a readable HTML string"""
+    """Format statistics into a readable HTML string with dark theme"""
     html = f"""
-    <div style="padding: 15px; background: linear-gradient(135deg, #667eea 0%, #764ba2 100%); border-radius: 10px; color: white; margin: 10px 0;">
+    <div style="padding: 15px; background: #000000; border-radius: 10px; color: white; margin: 10px 0;">
         <h3 style="margin-top: 0; color: white;">📊 Change Analysis</h3>
         
         <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(150px, 1fr)); gap: 10px; margin: 15px 0;">
-            <div style="background: rgba(255,255,255,0.1); padding: 10px; border-radius: 8px; text-align: center;">
+            <div style="background: rgba(255,255,255,0.05); padding: 10px; border-radius: 8px; text-align: center;">
                 <div style="font-size: 24px; font-weight: bold;">{stats['total_original']}</div>
-                <div style="font-size: 12px; opacity: 0.9;">Original Words</div>
+                <div style="font-size: 12px; opacity: 0.8;">Original Words</div>
             </div>
-            <div style="background: rgba(255,255,255,0.1); padding: 10px; border-radius: 8px; text-align: center;">
+            <div style="background: rgba(255,255,255,0.05); padding: 10px; border-radius: 8px; text-align: center;">
                 <div style="font-size: 24px; font-weight: bold;">{stats['total_generated']}</div>
-                <div style="font-size: 12px; opacity: 0.9;">Generated Words</div>
+                <div style="font-size: 12px; opacity: 0.8;">Generated Words</div>
             </div>
-            <div style="background: rgba(255,255,255,0.1); padding: 10px; border-radius: 8px; text-align: center;">
+            <div style="background: rgba(255,255,255,0.05); padding: 10px; border-radius: 8px; text-align: center;">
                 <div style="font-size: 24px; font-weight: bold; color: #90EE90;">{stats['unchanged']}</div>
-                <div style="font-size: 12px; opacity: 0.9;">Unchanged</div>
+                <div style="font-size: 12px; opacity: 0.8;">Unchanged</div>
             </div>
-            <div style="background: rgba(255,255,255,0.1); padding: 10px; border-radius: 8px; text-align: center;">
-                <div style="font-size: 24px; font-weight: bold; color: #FFB6C1;">{stats['changed']}</div>
-                <div style="font-size: 12px; opacity: 0.9;">Changed</div>
+            <div style="background: rgba(255,255,255,0.05); padding: 10px; border-radius: 8px; text-align: center;">
+                <div style="font-size: 24px; font-weight: bold; color: #FF4C4C;">{stats['changed']}</div>
+                <div style="font-size: 12px; opacity: 0.8;">Changed</div>
             </div>
         </div>
         
-        <div style="margin: 15px 0; padding: 10px; background: rgba(255,255,255,0.1); border-radius: 8px;">
+        <div style="margin: 15px 0; padding: 10px; background: rgba(255,255,255,0.05); border-radius: 8px;">
             <div style="margin-bottom: 8px;">
                 <strong>Modification Rate:</strong> {stats['percentage_changed']:.1f}% modified, {stats['percentage_unchanged']:.1f}% preserved
             </div>
             <div style="margin-bottom: 8px;">
                 <span style="color: #90EE90;">✚ Added: {stats['added']} words</span> | 
-                <span style="color: #FFB6C1;">✖ Removed: {stats['deleted']} words</span>
+                <span style="color: #FF4C4C;">✖ Removed: {stats['deleted']} words</span>
             </div>
         </div>
     """
     
     if stats['substitutions']:
         html += """
-        <div style="margin-top: 15px; padding: 10px; background: rgba(255,255,255,0.1); border-radius: 8px;">
+        <div style="margin-top: 15px; padding: 10px; background: rgba(255,255,255,0.05); border-radius: 8px;">
             <strong>🔄 Sample Word Substitutions:</strong>
             <div style="margin-top: 8px; font-size: 13px; line-height: 1.6;">
         """
         for orig, new in stats['substitutions']:
-            html += f'<div style="margin: 4px 0;"><span style="color: #FFB6C1;">{orig}</span> → <span style="color: #90EE90;">{new}</span></div>'
+            html += f'<div style="margin: 4px 0;"><span style="color: #FF4C4C;">{orig}</span> → <span style="color: #90EE90;">{new}</span></div>'
         html += """
             </div>
         </div>
@@ -211,8 +211,8 @@ def format_statistics(stats: Dict) -> str:
     html += """
         <div style="margin-top: 15px; padding: 8px; background: rgba(255,255,255,0.05); border-radius: 6px; font-size: 12px;">
             <strong>Legend:</strong> 
-            <span style="background-color: #ffcccc; padding: 2px 6px; border-radius: 3px; margin: 0 4px;">Removed/Changed</span>
-            <span style="background-color: #ccffcc; padding: 2px 6px; border-radius: 3px; margin: 0 4px;">Added/New</span>
+            <span style="background-color: #FF4C4C; padding: 2px 6px; border-radius: 3px; margin: 0 4px;">Removed/Changed</span>
+            <span style="background-color: #90EE90; padding: 2px 6px; border-radius: 3px; margin: 0 4px;">Added/New</span>
         </div>
     </div>
     """
@@ -386,6 +386,7 @@ def process_text(
         print(error_msg)
         return error_msg, "Error occurred", 0.0, "", "", ""
 
+# Create Gradio interface
 # Create Gradio interface
 with gr.Blocks(title="Text Paraphraser & Expander", theme=gr.themes.Soft()) as demo:
     gr.Markdown(
@@ -499,8 +500,6 @@ with gr.Blocks(title="Text Paraphraser & Expander", theme=gr.themes.Soft()) as d
     gr.Markdown("---")
     gr.Markdown("## 🔍 Visual Comparison - See What Changed")
     
-    change_stats = gr.HTML(label="Change Statistics")
-    
     with gr.Row():
         with gr.Column(scale=1):
             gr.Markdown("### 📄 Original Text (with changes highlighted)")
@@ -515,6 +514,8 @@ with gr.Blocks(title="Text Paraphraser & Expander", theme=gr.themes.Soft()) as d
                 label="Generated with Changes",
                 show_label=False
             )
+
+    change_stats = gr.HTML(label="Change Statistics")
     
     # Event handlers
     mode.change(

text-paraphraser-expander.ipynb

@@ -0,0 +1 @@
+{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.11.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"none","dataSources":[],"dockerImageVersionId":31153,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"pip install gradio transformers torch sentence-transformers","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"import gradio as gr\nimport torch\nfrom transformers import (\n    AutoTokenizer, \n    AutoModelForSeq2SeqLM,\n    T5ForConditionalGeneration,\n    T5Tokenizer\n)\nfrom sentence_transformers import SentenceTransformer, util\nimport numpy as np\nfrom typing import List, Tuple, Dict\nimport re\nimport numpy as np\nimport difflib","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Initialize similarity model\nsimilarity_model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"PARAPHRASE_MODELS = {\n    \"T5-Base\": \"Vamsi/T5_Paraphrase_Paws\",\n    \"PEGASUS-Paraphrase\": \"tuner007/pegasus_paraphrase\",\n    \"Parrot-Paraphraser\": \"prithivida/parrot_paraphraser_on_T5\",\n    \"BART-Paraphrase\": \"eugenesiow/bart-paraphrase\",\n    \"ChatGPT-Style-T5\": \"humarin/chatgpt_paraphraser_on_T5_base\",\n}","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"EXPANSION_MODELS = {\n    \"Flan-T5-Base\": \"google/flan-t5-base\",\n    \"Flan-T5-Large\": \"google/flan-t5-large\",\n}","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Cache for loaded models\nmodel_cache = {}","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def load_model(model_name: str, model_path: str):\n    \"\"\"Load model and tokenizer with caching\"\"\"\n    if model_name in model_cache:\n        return model_cache[model_name]\n    \n    print(f\"Loading {model_name}...\")\n    tokenizer = AutoTokenizer.from_pretrained(model_path)\n    model = AutoModelForSeq2SeqLM.from_pretrained(model_path)\n    \n    # Move to GPU if available\n    device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n    model = model.to(device)\n    \n    model_cache[model_name] = (model, tokenizer, device)\n    return model, tokenizer, device","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def chunk_text(text: str, max_sentences: int = 4) -> List[str]:\n    \"\"\"Split text into chunks based on number of sentences\"\"\"\n    import re\n    sentences = re.split(r'(?<=[.!?]) +', text.strip())\n    chunks = [' '.join(sentences[i:i+max_sentences]) for i in range(0, len(sentences), max_sentences)]\n    return [chunk for chunk in chunks if chunk.strip()]\n\ndef estimate_tokens(text: str) -> int:\n    \"\"\"Estimate number of tokens in text (approximate: 1 token ≈ 0.75 words)\"\"\"\n    word_count = len(text.split())\n    return int(word_count / 0.75)\n\ndef calculate_max_length(input_text: str, mode: str, base_max_length: int) -> int:\n    \"\"\"Calculate appropriate max_length based on input tokens\"\"\"\n    input_tokens = estimate_tokens(input_text)\n    \n    if mode == \"Paraphrase\":\n        # For paraphrasing: output should be 1.2-1.5x input tokens\n        calculated_max = int(input_tokens * 1.5) + 50\n    else:\n        # For expansion: output should be 2-3x input tokens\n        calculated_max = int(input_tokens * 3) + 100\n    \n    # Use the larger of calculated or user-specified max_length\n    final_max_length = max(calculated_max, base_max_length)\n    \n    # Cap at reasonable maximum to avoid memory issues\n    return min(final_max_length, 1024)\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# def chunk_text(text: str, max_tokens: int = 300) -> List[str]:\n#     \"\"\"Split text into chunks based on sentences to avoid exceeding token limits\"\"\"\n#     sentences = re.split(r'(?<=[.!?])\\s+', text)\n#     chunks = []\n#     current_chunk = []\n#     current_length = 0\n    \n#     for sentence in sentences:\n#         sentence_length = len(sentence.split())\n#         if current_length + sentence_length > max_tokens and current_chunk:\n#             chunks.append(' '.join(current_chunk))\n#             current_chunk = [sentence]\n#             current_length = sentence_length\n#         else:\n#             current_chunk.append(sentence)\n#             current_length += sentence_length\n    \n#     if current_chunk:\n#         chunks.append(' '.join(current_chunk))\n    \n#     return chunks if chunks else [text]","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def calculate_similarity(text1: str, text2: str) -> float:\n    \"\"\"Calculate cosine similarity between two texts\"\"\"\n    embeddings = similarity_model.encode([text1, text2], convert_to_tensor=True)\n    similarity = util.cos_sim(embeddings[0], embeddings[1])\n    return similarity.item()","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def highlight_differences(original: str, generated: str) -> Tuple[str, str, Dict]:\n    \"\"\"\n    Create highlighted HTML versions of both texts showing differences\n    Returns: (highlighted_original, highlighted_generated, statistics)\n    \"\"\"\n    # Split into words for comparison\n    original_words = original.split()\n    generated_words = generated.split()\n    \n    # Use difflib to find differences\n    diff = difflib.SequenceMatcher(None, original_words, generated_words)\n    \n    highlighted_original = []\n    highlighted_generated = []\n    \n    changes_count = 0\n    additions_count = 0\n    deletions_count = 0\n    unchanged_count = 0\n    word_substitutions = []\n    \n    for tag, i1, i2, j1, j2 in diff.get_opcodes():\n        original_chunk = ' '.join(original_words[i1:i2])\n        generated_chunk = ' '.join(generated_words[j1:j2])\n        \n        if tag == 'equal':\n            # Unchanged text\n            highlighted_original.append(original_chunk)\n            highlighted_generated.append(generated_chunk)\n            unchanged_count += (i2 - i1)\n            \n        elif tag == 'replace':\n            # Changed text\n            highlighted_original.append(f'<span style=\"color: #ffcccc; padding: 2px 4px; border-radius: 3px; text-decoration: line-through;\">{original_chunk}</span>')\n            highlighted_generated.append(f'<span style=\"color: #ccffcc; padding: 2px 4px; border-radius: 3px; font-weight: 500;\">{generated_chunk}</span>')\n            changes_count += max(i2 - i1, j2 - j1)\n            \n            # Track word substitutions (limit to single word changes for clarity)\n            if i2 - i1 == 1 and j2 - j1 == 1:\n                word_substitutions.append((original_chunk, generated_chunk))\n            \n        elif tag == 'delete':\n            # Text removed in generated\n            highlighted_original.append(f'<span style=\"color: #ffcccc; padding: 2px 4px; border-radius: 3px; text-decoration: line-through;\">{original_chunk}</span>')\n            deletions_count += (i2 - i1)\n            \n        elif tag == 'insert':\n            # Text added in generated\n            highlighted_generated.append(f'<span style=\"color: #ccffcc; padding: 2px 4px; border-radius: 3px; font-weight: 500;\">{generated_chunk}</span>')\n            additions_count += (j2 - j1)\n    \n    # Join with spaces\n    final_original = ' '.join(highlighted_original)\n    final_generated = ' '.join(highlighted_generated)\n    \n    # Calculate statistics\n    total_original_words = len(original_words)\n    total_generated_words = len(generated_words)\n    \n    percentage_changed = (changes_count + deletions_count + additions_count) / max(total_original_words, 1) * 100\n    percentage_unchanged = (unchanged_count / max(total_original_words, 1)) * 100\n    \n    statistics = {\n        'total_original': total_original_words,\n        'total_generated': total_generated_words,\n        'unchanged': unchanged_count,\n        'changed': changes_count,\n        'added': additions_count,\n        'deleted': deletions_count,\n        'percentage_changed': percentage_changed,\n        'percentage_unchanged': percentage_unchanged,\n        'substitutions': word_substitutions[:10]  # Limit to first 10\n    }\n    \n    return final_original, final_generated, statistics\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def format_statistics(stats: Dict) -> str:\n    \"\"\"Format statistics into a readable HTML string with dark theme\"\"\"\n    html = f\"\"\"\n    <div style=\"padding: 15px; background: #000000; border-radius: 10px; color: white; margin: 10px 0;\">\n        <h3 style=\"margin-top: 0; color: white;\">📊 Change Analysis</h3>\n        \n        <div style=\"display: grid; grid-template-columns: repeat(auto-fit, minmax(150px, 1fr)); gap: 10px; margin: 15px 0;\">\n            <div style=\"background: rgba(255,255,255,0.05); padding: 10px; border-radius: 8px; text-align: center;\">\n                <div style=\"font-size: 24px; font-weight: bold;\">{stats['total_original']}</div>\n                <div style=\"font-size: 12px; opacity: 0.8;\">Original Words</div>\n            </div>\n            <div style=\"background: rgba(255,255,255,0.05); padding: 10px; border-radius: 8px; text-align: center;\">\n                <div style=\"font-size: 24px; font-weight: bold;\">{stats['total_generated']}</div>\n                <div style=\"font-size: 12px; opacity: 0.8;\">Generated Words</div>\n            </div>\n            <div style=\"background: rgba(255,255,255,0.05); padding: 10px; border-radius: 8px; text-align: center;\">\n                <div style=\"font-size: 24px; font-weight: bold; color: #90EE90;\">{stats['unchanged']}</div>\n                <div style=\"font-size: 12px; opacity: 0.8;\">Unchanged</div>\n            </div>\n            <div style=\"background: rgba(255,255,255,0.05); padding: 10px; border-radius: 8px; text-align: center;\">\n                <div style=\"font-size: 24px; font-weight: bold; color: #FF4C4C;\">{stats['changed']}</div>\n                <div style=\"font-size: 12px; opacity: 0.8;\">Changed</div>\n            </div>\n        </div>\n        \n        <div style=\"margin: 15px 0; padding: 10px; background: rgba(255,255,255,0.05); border-radius: 8px;\">\n            <div style=\"margin-bottom: 8px;\">\n                <strong>Modification Rate:</strong> {stats['percentage_changed']:.1f}% modified, {stats['percentage_unchanged']:.1f}% preserved\n            </div>\n            <div style=\"margin-bottom: 8px;\">\n                <span style=\"color: #90EE90;\">✚ Added: {stats['added']} words</span> | \n                <span style=\"color: #FF4C4C;\">✖ Removed: {stats['deleted']} words</span>\n            </div>\n        </div>\n    \"\"\"\n    \n    if stats['substitutions']:\n        html += \"\"\"\n        <div style=\"margin-top: 15px; padding: 10px; background: rgba(255,255,255,0.05); border-radius: 8px;\">\n            <strong>🔄 Sample Word Substitutions:</strong>\n            <div style=\"margin-top: 8px; font-size: 13px; line-height: 1.6;\">\n        \"\"\"\n        for orig, new in stats['substitutions']:\n            html += f'<div style=\"margin: 4px 0;\"><span style=\"color: #FF4C4C;\">{orig}</span> → <span style=\"color: #90EE90;\">{new}</span></div>'\n        html += \"\"\"\n            </div>\n        </div>\n        \"\"\"\n    \n    html += \"\"\"\n        <div style=\"margin-top: 15px; padding: 8px; background: rgba(255,255,255,0.05); border-radius: 6px; font-size: 12px;\">\n            <strong>Legend:</strong> \n            <span style=\"background-color: #FF4C4C; padding: 2px 6px; border-radius: 3px; margin: 0 4px;\">Removed/Changed</span>\n            <span style=\"background-color: #90EE90; padding: 2px 6px; border-radius: 3px; margin: 0 4px;\">Added/New</span>\n        </div>\n    </div>\n    \"\"\"\n    \n    return html\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def paraphrase_text(\n    text: str,\n    model_name: str,\n    temperature: float,\n    top_p: float,\n    max_length: int,\n    num_beams: int,\n    max_sentences: int,\n    target_words: int = None,\n    mode: str = \"Paraphrase\"\n) -> Tuple[str, float]:\n    \"\"\"Paraphrase or expand text based on mode\"\"\"\n    \n    if not text.strip():\n        return \"Please enter some text to process.\", 0.0\n    \n    # Select appropriate model based on mode\n    if mode == \"Paraphrase\":\n        models_dict = PARAPHRASE_MODELS\n        if model_name not in models_dict:\n            model_name = list(models_dict.keys())[0]\n        model_path = models_dict[model_name]\n        prefix = \"paraphrase: \" if \"T5\" in model_name else \"\"\n    else:  # Expand mode\n        models_dict = EXPANSION_MODELS\n        if model_name not in models_dict:\n            model_name = list(models_dict.keys())[0]\n        model_path = models_dict[model_name]\n        target_words = target_words or 300\n        prefix = f\"Expand the following text to approximately {target_words} words, adding more details and context: \"\n    \n    # Load model\n    model, tokenizer, device = load_model(model_name, model_path)\n    \n    # Chunk text based on sentences\n    chunks = chunk_text(text, max_sentences=max_sentences)\n    \n    processed_chunks = []\n    \n    print(f\"\\n{'='*60}\")\n    print(f\"Processing {len(chunks)} chunk(s) with {max_sentences} sentences per chunk\")\n    print(f\"{'='*60}\")\n    \n    for i, chunk in enumerate(chunks):\n        # Calculate dynamic max_length for this chunk\n        chunk_max_length = calculate_max_length(chunk, mode, max_length)\n        input_tokens = estimate_tokens(chunk)\n        \n        # Prepare input\n        input_text = prefix + chunk + \" </s>\" if mode == \"Paraphrase\" else prefix + chunk\n        inputs = tokenizer.encode(\n            input_text, \n            return_tensors=\"pt\", \n            max_length=512, \n            truncation=True\n        )\n        inputs = inputs.to(device)\n        \n        # Calculate min_length to ensure output isn't too short\n        if mode == \"Paraphrase\":\n            min_length_calc = int(input_tokens * 0.8)\n        else:\n            min_length_calc = int(input_tokens * 1.5)\n        \n        # Generate\n        with torch.no_grad():\n            outputs = model.generate(\n                inputs,\n                max_length=chunk_max_length,\n                min_length=min(min_length_calc, chunk_max_length - 10),\n                num_beams=num_beams,\n                temperature=temperature if temperature > 0 else 1.0,\n                top_p=top_p,\n                top_k=120 if mode == \"Paraphrase\" else 50,\n                do_sample=temperature > 0,\n                early_stopping=True,\n                no_repeat_ngram_size=3 if mode == \"Expand\" else 2,\n                length_penalty=1.0 if mode == \"Paraphrase\" else 1.5,\n                repetition_penalty=1.2,\n            )\n        \n        # Decode output\n        processed_text = tokenizer.decode(outputs[0], skip_special_tokens=True)\n        processed_chunks.append(processed_text.strip())\n        \n        output_tokens = estimate_tokens(processed_text)\n        print(f\"Chunk {i+1}/{len(chunks)}:\")\n        print(f\"  Input: {len(chunk.split())} words (~{input_tokens} tokens)\")\n        print(f\"  Output: {len(processed_text.split())} words (~{output_tokens} tokens)\")\n        print(f\"  Max length used: {chunk_max_length}\")\n        print(f\"-\" * 60)\n    \n    # Combine chunks with double newline\n    final_text = \"\\n\\n\".join(processed_chunks)\n    \n    # Calculate similarity\n    similarity_score = calculate_similarity(text, final_text)\n    \n    print(f\"{'='*60}\")\n    print(f\"Total: {len(text.split())} → {len(final_text.split())} words\")\n    print(f\"Similarity: {similarity_score:.4f}\")\n    print(f\"{'='*60}\\n\")\n    \n    return final_text, similarity_score","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def update_model_choices(mode: str):\n    \"\"\"Update model dropdown based on selected mode\"\"\"\n    if mode == \"Paraphrase\":\n        choices = list(PARAPHRASE_MODELS.keys())\n    else:\n        choices = list(EXPANSION_MODELS.keys())\n    return gr.Dropdown(choices=choices, value=choices[0])\n\ndef update_parameters_visibility(mode: str):\n    \"\"\"Show/hide target words parameter based on mode\"\"\"\n    if mode == \"Expand\":\n        return gr.Number(visible=True)\n    else:\n        return gr.Number(visible=False)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"def process_text(\n    input_text: str,\n    mode: str,\n    model_name: str,\n    temperature: float,\n    top_p: float,\n    max_length: int,\n    num_beams: int,\n    max_sentences: int,\n    target_words: int\n):\n    \"\"\"Main processing function\"\"\"\n    try:\n        output_text, similarity = paraphrase_text(\n            input_text,\n            model_name,\n            temperature,\n            top_p,\n            max_length,\n            num_beams,\n            max_sentences,\n            target_words,\n            mode\n        )\n        \n        word_count_original = len(input_text.split())\n        word_count_output = len(output_text.split())\n        \n        # Generate highlighted comparison\n        highlighted_original, highlighted_generated, statistics = highlight_differences(\n            input_text, \n            output_text\n        )\n        \n        # Format statistics\n        stats_html = format_statistics(statistics)\n        \n        # Basic stats line\n        basic_stats = f\"**Original:** {word_count_original} words | **Generated:** {word_count_output} words | **Similarity:** {similarity:.4f}\"\n        \n        return output_text, basic_stats, similarity, highlighted_original, highlighted_generated, stats_html\n    except Exception as e:\n        import traceback\n        error_msg = f\"Error: {str(e)}\\n\\n{traceback.format_exc()}\"\n        print(error_msg)\n        return error_msg, \"Error occurred\", 0.0, \"\", \"\", \"\"\n","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"# Create Gradio interface\nwith gr.Blocks(title=\"Text Paraphraser & Expander\", theme=gr.themes.Soft()) as demo:\n    gr.Markdown(\n        \"\"\"\n        # 📝 Text Paraphraser & Expander\n        Transform your text with AI-powered paraphrasing and expansion capabilities.\n        \"\"\"\n    )\n    \n    with gr.Row():\n        with gr.Column(scale=1):\n            mode = gr.Radio(\n                choices=[\"Paraphrase\", \"Expand\"],\n                value=\"Paraphrase\",\n                label=\"Mode\",\n                info=\"Choose to paraphrase or expand your text\"\n            )\n            \n            model_dropdown = gr.Dropdown(\n                choices=list(PARAPHRASE_MODELS.keys()),\n                value=list(PARAPHRASE_MODELS.keys())[0],\n                label=\"Model Selection\",\n                info=\"Choose the model for processing\"\n            )\n            \n            gr.Markdown(\"### ⚙️ Parameters\")\n            \n            temperature = gr.Slider(\n                minimum=0.0,\n                maximum=2.0,\n                value=0.7,\n                step=0.1,\n                label=\"Temperature\",\n                info=\"Higher = more creative, Lower = more focused\"\n            )\n            \n            top_p = gr.Slider(\n                minimum=0.1,\n                maximum=1.0,\n                value=0.9,\n                step=0.05,\n                label=\"Top-p (Nucleus Sampling)\",\n                info=\"Probability threshold for token selection\"\n            )\n            \n            max_length = gr.Slider(\n                minimum=128,\n                maximum=1024,\n                value=512,\n                step=32,\n                label=\"Max Length (tokens)\",\n                info=\"Maximum length of generated text per chunk\"\n            )\n            \n            num_beams = gr.Slider(\n                minimum=1,\n                maximum=10,\n                value=4,\n                step=1,\n                label=\"Number of Beams\",\n                info=\"Higher = better quality but slower\"\n            )\n            \n            max_sentences = gr.Slider(\n                minimum=1,\n                maximum=10,\n                value=4,\n                step=1,\n                label=\"Sentences per Chunk\",\n                info=\"Number of sentences to process together\"\n            )\n            \n            target_words = gr.Number(\n                value=300,\n                label=\"Target Word Count (Expand mode)\",\n                info=\"Approximate number of words for expansion\",\n                visible=False\n            )\n    \n    with gr.Row():\n        with gr.Column(scale=1):\n            gr.Markdown(\"### 📥 Input Text\")\n            input_text = gr.Textbox(\n                lines=10,\n                placeholder=\"Enter your text here...\",\n                label=\"Original Text\",\n                show_copy_button=True\n            )\n        \n        with gr.Column(scale=1):\n            gr.Markdown(\"### 📤 Generated Text\")\n            output_text = gr.Textbox(\n                lines=10,\n                label=\"Processed Text\",\n                show_copy_button=True\n            )\n    \n    with gr.Row():\n        process_btn = gr.Button(\"🚀 Generate\", variant=\"primary\", size=\"lg\")\n        clear_btn = gr.ClearButton([input_text, output_text], value=\"🗑️ Clear\")\n    \n    stats_display = gr.Markdown()\n    \n    similarity_display = gr.Number(\n        label=\"Cosine Similarity Score\",\n        precision=4,\n        interactive=False\n    )\n    \n    # Highlighted comparison section\n    gr.Markdown(\"---\")\n    gr.Markdown(\"## 🔍 Visual Comparison - See What Changed\")\n    \n    with gr.Row():\n        with gr.Column(scale=1):\n            gr.Markdown(\"### 📄 Original Text (with changes highlighted)\")\n            highlighted_original = gr.HTML(\n                label=\"Original with Changes\",\n                show_label=False\n            )\n        \n        with gr.Column(scale=1):\n            gr.Markdown(\"### ✨ Generated Text (with changes highlighted)\")\n            highlighted_generated = gr.HTML(\n                label=\"Generated with Changes\",\n                show_label=False\n            )\n\n    change_stats = gr.HTML(label=\"Change Statistics\")\n    \n    # Event handlers\n    mode.change(\n        fn=update_model_choices,\n        inputs=[mode],\n        outputs=[model_dropdown]\n    )\n    \n    mode.change(\n        fn=update_parameters_visibility,\n        inputs=[mode],\n        outputs=[target_words]\n    )\n    \n    process_btn.click(\n        fn=process_text,\n        inputs=[\n            input_text,\n            mode,\n            model_dropdown,\n            temperature,\n            top_p,\n            max_length,\n            num_beams,\n            max_sentences,\n            target_words\n        ],\n        outputs=[\n            output_text, \n            stats_display, \n            similarity_display,\n            highlighted_original,\n            highlighted_generated,\n            change_stats\n        ]\n    )\n    \n    gr.Markdown(\n        \"\"\"\n        ---\n        ### 💡 Tips:\n        - **Paraphrase Mode**: Rewrites text while preserving meaning\n        - **Expand Mode**: Adds details and elaboration to make text longer\n        - **Sentences per Chunk**: Controls how many sentences are processed together (4 recommended)\n        - Adjust temperature for creativity (0.7-1.0 for paraphrase, 1.0-1.5 for expansion)\n        - Higher beam count = better quality but slower processing\n        - Max length is automatically calculated based on input, but can be overridden\n        - Output chunks are separated by double newlines for readability\n        \"\"\"\n    )","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"if __name__ == \"__main__\":\n    demo.launch(share=True)","metadata":{"trusted":true},"outputs":[],"execution_count":null},{"cell_type":"code","source":"","metadata":{"trusted":true},"outputs":[],"execution_count":null}]}