refactor: using falcon

app.py

@@ -17,7 +17,7 @@ DATA_DIR = "/data" if os.path.exists("/data") else "."
 DATASET_DIR = os.path.join(DATA_DIR, "rag_dataset")
 DATASET_PATH = os.path.join(DATASET_DIR, "dataset")
 TOKENIZER_MODEL = "google/flan-t5-small"
-SUMMARIZATION_MODEL= "HuggingFaceTB/SmolVLM-256M-Instruct"
+SUMMARIZATION_MODEL= "Falconsai/text_summarization"
 # SUMMARIZATION_MODEL="rhaymison/t5-portuguese-small-summarization"
 
 @st.cache_resource
@@ -210,26 +210,119 @@ def generate_answer(question, context, max_length=512):
     clean_question = clean_text(question)
     
     # Format the input for T5 (it expects a specific format)
-    input_text = f"""Objective:
-Provide a clear, simple, and well-structured answer about autism that is easy to understand for a general audience. Use the provided research papers as references.
-
-Question: {clean_question}
-Research Papers:
-{clean_context}
-
-Instructions:
-Start with a simple definition 
-- Explain what autism is in a short and clear way, avoiding technical terms.
-- Use real-life examples 
-- Give practical and relatable examples to help illustrate key points.
-- Explain research in simple words 
-- Instead of just citing studies, summarize their key findings in a way that anyone can understand. Example: "A study from X University found that..."
-- Avoid complex words 
-- If a scientific term is needed, provide a short and simple explanation.
-- Use clear formatting 
-- Write in short paragraphs, bullet points, or numbered lists to improve readability.
-- Keep a friendly tone 
-- Make the response engaging and easy to follow, so people without prior knowledge can understand."""
+    input_text = f"""Context
+Input Question: {clean_question}
+Source Materials: {clean_context}
+Primary Objective
+Generate a comprehensive yet accessible summary of autism research that bridges the gap between academic knowledge and public understanding. The response should be evidence-based while remaining engaging and practical for general readers.
+Content Structure
+1. Opening Overview
+
+Begin with a concise, jargon-free definition of autism
+Frame the topic within everyday experiences
+Establish relevance to the reader's understanding
+
+2. Key Concepts Breakdown
+
+Transform complex research findings into digestible information
+Structure information in a logical progression
+Connect each point to real-world scenarios
+
+3. Research Integration
+Present research findings using this framework:
+
+Main finding: [Clear statement of what was discovered]
+Real-world meaning: [Practical implications]
+Context: [How this fits into broader understanding]
+
+4. Examples and Applications
+Include:
+
+Concrete, relatable scenarios
+Day-to-day situations
+Practical implications for families and individuals
+
+Writing Guidelines
+Language Requirements
+
+Target reading level: 8th grade
+Sentence length: Maximum 20 words
+Paragraph length: 2-4 sentences
+Technical terms: Must include plain language explanation in parentheses
+
+Tone and Style
+
+Empathetic and respectful
+Solution-focused approach
+Balanced perspective
+Inclusive language
+
+Formatting Specifications
+
+Use headers for major sections
+Include white space between concepts
+Implement bullet points for lists
+Bold key terms with immediate explanations
+
+Research Citation Format
+When referencing studies, follow this pattern:
+"Research from [Institution] shows [finding in simple terms]. This means [practical interpretation]."
+Quality Checks
+Before finalizing, ensure the summary:
+
+Answers the original question directly
+Maintains scientific accuracy while being accessible
+Provides actionable insights
+Respects neurodiversity perspectives
+Balances depth with clarity
+
+Response Framework
+
+Introduction (2-3 sentences)
+
+Core definition
+Relevance statement
+
+
+Main Body (3-4 key points)
+
+Evidence-based insights
+Practical examples
+Real-world applications
+
+
+Conclusion (2-3 sentences)
+
+Summary of key takeaways
+Actionable next steps or implications
+
+
+
+Engagement Elements
+
+Include thought-provoking questions
+Provide relatable scenarios
+Connect to common experiences
+Offer practical applications
+
+Modified Output Analysis
+The response should be evaluated against these criteria:
+
+Clarity: Is the information immediately understandable?
+Accuracy: Does it reflect the research correctly?
+Relevance: Does it address the specific question?
+Practicality: Are the insights actionable?
+Engagement: Does it maintain reader interest?
+
+Special Considerations
+
+Acknowledge spectrum nature of autism
+Respect diverse perspectives
+Focus on strengths and challenges
+Avoid deficit-based language
+Include support-oriented information
+
+Remember to adapt the depth and complexity based on the specific question while maintaining accessibility and scientific accuracy."""
     
     try:
         # T5 expects a specific format for the input