Update app.py

app.py

@@ -17,21 +17,25 @@ nlp = spacy.load("en_core_web_sm")
 model = SentenceTransformer('all-MiniLM-L6-v2')
 
 weights = {
-    "information_density": 0.2,  
-    "unique_key_points": 0.8,   
-    "strength_word_count": 0.002,  
-    "weakness_word_count": 0.004, 
-    "discussion_word_count": 0.01  
+    "information_density": 0.1468,
+    "unique_key_points": 0.8323,
+    "strength_word_count": 0.0034,
+    "weakness_word_count": 0.0047,
+    "discussion_word_count": 0.0127
 }
 
+# Dynamic thresholds derived via GMM
+THRESH_POOR = 4.5941101516587315
+THRESH_EXCELLENT = 10.581525294084457
+
 THRESHOLDS = {
-    "normalized_length": (0.15, 0.25),
-    "unique_key_points": (3, 10),
-    "information_density": (0.01, 0.02),
-    "unique_insights_per_word": 0.002,
-    "optimization_score": 0.7,
-    "composite_score": 5,
-    "adjusted_argument_strength": 0.75
+  "normalized_length": (0.1510989010989011, 0.3722527472527472),
+  "unique_key_points": (2.0, 5.0),
+  "information_density": (0.00727669710202455, 0.011291644452510826),
+  "unique_insights_per_word": 13.607476635514018,
+  "optimization_score": 0.032395506665225886,
+  "composite_score": (4.5941101516587315, 10.581525294084457),
+  "adjusted_argument_strength": 0.0616150390018116,
 }
 
 def chunk_text(text, max_length):
@@ -164,12 +168,13 @@ def calculate_composite_score(df):
     return df
 
 def classify_review_quality(row):
-    if row['composite_score'] > 12: 
-        return 'Excellent Review Quality'
-    elif row['composite_score'] < 3:  
-        return 'Poor Review Quality'
+    cs = row['composite_score']
+    if cs < THRESH_POOR:
+        return 'Poor'
+    elif cs > THRESH_EXCELLENT:
+        return 'Excellent'
     else:
-        return 'Moderate Review Quality'
+        return 'Moderate'
 
 def determine_review_quality(df):
 
@@ -209,8 +214,8 @@ def heuristic_optimization(row):
     elif row["unique_key_points"] > THRESHOLDS["unique_key_points"][1]:
         suggestions.append("Streamline ideas for clarity.")
 
-    if row["composite_score"] < THRESHOLDS["composite_score"]:
-        suggestions.append("Enhance clarity, evidence, and argumentation.")
+    if row["optimization_score"] < THRESHOLDS["optimization_score"]:
+        suggestions.append("Restructure content and improve phrasing.")
 
     if row["review_quality"] == "Low":
         suggestions.append("Significant revisions required.")
@@ -241,7 +246,7 @@ iface = gr.Interface(
     inputs=[gr.Textbox(label="Peer Review Comments"), gr.Textbox(label="Paper Abstract")],
     outputs=[gr.Textbox(label="Predicted Review Quality"), gr.Textbox(label="Suggestions")],
     title="# Dynamic Length Optimization of Peer Review",
-    description="A framework which dynamically provides suggestion to improve a peer review.",
+    description="Predicts review quality and suggests improvements.",
 )
 
 if __name__ == "__main__":