Update app.py

app.py

@@ -436,7 +436,7 @@ class ContentAnalyzer:
                 time.sleep(self.retry_delay * (2 ** attempt))
 
     def _create_analysis_prompt(self, transcript: str) -> str:
-        """Create the analysis prompt with balanced evaluation criteria"""
+        """Create the analysis prompt with stricter evaluation criteria"""
         # First try to extract existing timestamps
         timestamps = re.findall(r'\[(\d{2}:\d{2})\]', transcript)
         
@@ -462,9 +462,9 @@ Example: If a quote starts at word 300, timestamp would be [02:00] (300 words /
                 marked_transcript += word + " "
             transcript = marked_transcript
 
-        prompt_template = """Analyze this teaching content objectively. Each criterion should be evaluated independently with reasonable standards - neither too strict nor too lenient.
+        prompt_template = """Analyze this teaching content with strict standards. Each criterion must meet specific requirements for a passing score.
 
-Score 1 if the criterion meets reasonable standards, 0 if it does not.
+Score 1 ONLY if ALL requirements are met with clear evidence. Score 0 if ANY requirement is not fully met.
 
 Transcript:
 {transcript}
@@ -472,96 +472,108 @@ Transcript:
 Timestamp Instructions:
 {timestamp_instruction}
 
-Required JSON structure:
-{{
-    "Concept Assessment": {{
-        "Subject Matter Accuracy": {{
-            "Score": 0 or 1,  # Score 1 if content is generally accurate with only minor errors
-            "Citations": ["[MM:SS] Quote demonstrating accuracy or error"]
-        }},
-        "First Principles Approach": {{
-            "Score": 0 or 1,  # Score 1 if key concepts are built from fundamentals
-            "Citations": ["[MM:SS] Quote showing concept explanation"]
-        }},
-        "Examples and Business Context": {{
-            "Score": 0 or 1,  # Score 1 if at least 2 relevant examples are provided
-            "Citations": ["[MM:SS] Quote containing practical example"]
-        }},
-        "Cohesive Storytelling": {{
-            "Score": 0 or 1,  # Score 1 if content flows logically most of the time
-            "Citations": ["[MM:SS] Quote showing topic transition"]
-        }},
-        "Engagement and Interaction": {{
-            "Score": 0 or 1,  # Score 1 if there's meaningful audience engagement
-            "Citations": ["[MM:SS] Quote showing audience engagement"]
-        }},
-        "Professional Tone": {{
-            "Score": 0 or 1,  # Score 1 if tone is generally professional
-            "Citations": ["[MM:SS] Quote demonstrating tone"]
-        }}
-    }},
-    "Code Assessment": {{
-        "Depth of Explanation": {{
-            "Score": 0 or 1,  # Score 1 if code concepts are explained clearly
-            "Citations": ["[MM:SS] Quote showing code explanation"]
-        }},
-        "Output Interpretation": {{
-            "Score": 0 or 1,  # Score 1 if important outputs are explained
-            "Citations": ["[MM:SS] Quote demonstrating output explanation"]
-        }},
-        "Breaking down Complexity": {{
-            "Score": 0 or 1,  # Score 1 if complex concepts are made understandable
-            "Citations": ["[MM:SS] Quote showing concept breakdown"]
-        }}
-    }}
-}}
+Required JSON structure remains the same, but with stricter scoring criteria:
 
-Balanced Scoring Guidelines:
+Concept Assessment Criteria:
 
 Subject Matter Accuracy:
-✓ Pass: Content is technically sound with occasional minor errors
-✗ Fail: Multiple significant errors or fundamental misunderstandings
+✓ Score 1 if ALL:
+- No significant technical errors
+- Concepts explained with precise terminology
+- Clear distinction between facts and opinions
+✗ Score 0 if ANY:
+- Contains technical inaccuracies
+- Uses imprecise or incorrect terminology
+- Mixes facts with unsupported claims
 
 First Principles Approach:
-✓ Pass: Core concepts are explained from basic principles
-✗ Fail: Advanced concepts introduced without basic foundation
+✓ Score 1 if ALL:
+- Starts with fundamental concepts
+- Builds complexity systematically
+- Clear connections between basic and advanced concepts
+✗ Score 0 if ANY:
+- Jumps to advanced concepts without foundation
+- Missing logical progression
+- Unclear connections between concepts
 
 Examples and Business Context:
-✓ Pass: At least 2 relevant examples that illustrate concepts
-✗ Fail: Examples missing or irrelevant to the topic
+✓ Score 1 if ALL:
+- At least 2 relevant, detailed examples
+- Clear business context for each example
+- Examples directly support learning objectives
+✗ Score 0 if ANY:
+- Fewer than 2 examples
+- Examples lack business context
+- Examples don't clearly support learning
 
 Cohesive Storytelling:
-✓ Pass: Topics generally flow well with clear connections
-✗ Fail: Frequent jumps between unrelated topics
+✓ Score 1 if ALL:
+- Clear narrative structure
+- Logical topic transitions
+- Consistent theme throughout
+✗ Score 0 if ANY:
+- Disjointed narrative
+- Abrupt topic changes
+- Inconsistent theme
 
 Engagement and Interaction:
-✓ Pass: Some effective engagement with audience
-✗ Fail: One-way lecture with no audience consideration
+✓ Score 1 if ALL:
+- Regular audience engagement
+- Effective use of questions
+- Clear response to audience cues
+✗ Score 0 if ANY:
+- Minimal audience interaction
+- One-way lecture style
+- Misses engagement opportunities
 
 Professional Tone:
-✓ Pass: Generally professional with occasional casual moments
-✗ Fail: Consistently unprofessional or inappropriate
+✓ Score 1 if ALL:
+- Consistently professional language
+- Appropriate level of formality
+- Clear and confident delivery
+✗ Score 0 if ANY:
+- Casual or inappropriate language
+- Inconsistent formality
+- Uncertain or unclear delivery
+
+Code Assessment Criteria:
 
 Depth of Explanation:
-✓ Pass: Key code concepts explained with reasonable detail
-✗ Fail: Code presented without meaningful explanation
+✓ Score 1 if ALL:
+- Explains code purpose and structure
+- Covers implementation details
+- Addresses potential issues/alternatives
+✗ Score 0 if ANY:
+- Surface-level explanation
+- Missing implementation details
+- No discussion of alternatives
 
 Output Interpretation:
-✓ Pass: Important outputs and their significance explained
-✗ Fail: Outputs shown without context or explanation
+✓ Score 1 if ALL:
+- Clear explanation of expected outputs
+- Error handling discussion
+- Performance implications covered
+✗ Score 0 if ANY:
+- Unclear output expectations
+- No error handling discussion
+- Missing performance context
 
 Breaking down Complexity:
-✓ Pass: Complex topics broken into understandable parts
-✗ Fail: Complex concepts left unexplained
+✓ Score 1 if ALL:
+- Complex concepts broken into digestible parts
+- Clear step-by-step explanation
+- Logical progression of difficulty
+✗ Score 0 if ANY:
+- Overwhelming complexity
+- Missing steps in explanation
+- Illogical difficulty progression
 
 Important Notes:
-- Evaluate each criterion independently
-- Perfect delivery is not required for a passing score
-- Look for evidence of competent teaching rather than flawless execution
-- Consider the overall effectiveness for the target audience
-- One or two minor issues should not result in failure
-- Citations must support the scoring decision
-- Different criteria can and should receive different scores based on their individual merits"""
+- Each criterion is evaluated independently
+- Citations must directly support scoring decision
+- No partial credit - must meet ALL requirements for a score of 1
+- Look for explicit evidence in transcript
+- Different criteria can receive different scores based on evidence"""
 
         return prompt_template.format(
             transcript=transcript,