Update gradio_app.py

gradio_app.py

@@ -60,7 +60,8 @@ def toggle_feedback(request_data, response_data, like_clicked, dislike_clicked):
     else:
         return "Feedback sent successfully!"
 
-def preprocess_and_flatten(json_results, nested_fields, meta_fields=None):
+
+def preprocess_and_flatten(json_results, mode, meta_fields=None):
     # Ensure 'meta_fields' is a list or set default fields
     if meta_fields is None:
         meta_fields = ['doc_id', 'details', 'domain']
@@ -70,45 +71,44 @@ def preprocess_and_flatten(json_results, nested_fields, meta_fields=None):
         print(f"Invalid JSON results: Expected a dictionary but got {type(json_results)}")
         return pd.DataFrame()  # Return an empty DataFrame if json_results is not a dictionary
 
-    # Preprocess each field in the nested_fields list
-    for field in nested_fields:
-        if field in json_results:
-            nested_field = json_results.get(field)
-            if isinstance(nested_field, dict):
-                json_results[field] = [nested_field]  # Wrap dict in a list if it's a dict
-            elif not isinstance(nested_field, list):
-                json_results[field] = []  # Default to empty list if not a list or dict
-
-    # Flatten the JSON for each field in nested_fields
-    dataframes = []
-    for field in nested_fields:
-        # Check if the field exists in json_results
-        if field not in json_results or not isinstance(json_results[field], list):
-            print(f"Field '{field}' not found or is not a list in the JSON results.")
-            continue  # Skip this field if it's missing or not a list
-
-        try:
-            # Flatten the JSON for the specified nested field
-            dataframe_results = pd.json_normalize(
-                json_results,
-                record_path=[field],                     # Nested array to flatten
-                meta=meta_fields,                        # Top-level fields to retain
-                sep='_',                                 # Separator for nested fields
-                errors='ignore'                          # To handle cases with missing keys
-            )
-            if not dataframe_results.empty:
-                dataframes.append(dataframe_results)     # Append the flattened DataFrame
-        except Exception as e:
-            print(f"Error while flattening field '{field}': {e}")
-            continue  # Skip this field if an error occurs
-
-    # Concatenate all the DataFrames if there are any
-    if dataframes:
-        final_dataframe = pd.concat(dataframes, ignore_index=True)
-    else:
-        final_dataframe = pd.DataFrame()  # Return an empty DataFrame if nothing was flattened
-
-    return final_dataframe
+    # Collect flattened data
+    flattened_data = []
+
+    # Mode-based logic
+    if mode == 'news_analysis':
+        # Handle 'claim_objects' for news analysis mode
+        claim_objects = json_results.get('claim_objects', [])
+        if isinstance(claim_objects, list):
+            for item in claim_objects:
+                flattened_data.append({
+                    'doc_id': json_results.get('doc_id'),
+                    'details': json_results.get('details'),
+                    'domain': json_results.get('domain'),
+                    'topic': item.get('topic', ''),
+                    'claim': item.get('claim', ''),
+                    'claimer': item.get('claimer', '')
+                })
+
+    elif mode == 'claim_verification':
+        # Handle 'support', 'refute', 'no_info' for claim verification mode
+        nested_fields = ['support', 'refute', 'no_info']
+        for field in nested_fields:
+            nested_items = json_results.get(field, [])
+            if not isinstance(nested_items, list):
+                continue
+
+            # Loop over each item in the nested field and flatten
+            for item in nested_items:
+                flattened_data.append({
+                    'doc_id': json_results.get('doc_id'),
+                    'details': json_results.get('details'),
+                    'category': field,  # Mark which category the item belongs to (support/refute/no_info)
+                    'sentence': item.get('sentence', ''),
+                    'doi': item.get('doi', '')
+                })
+    
+    # Convert to DataFrame
+    return pd.DataFrame(flattened_data)
     
 # Define the functions to handle the inputs and outputs
 def news_analysis(text):
@@ -126,7 +126,7 @@ def news_analysis(text):
         # Prepare results for JSON output
         json_results = response.json()
         # Flatten 'claim_objects' field
-        dataframe_results = preprocess_and_flatten(json_results, ['claim_objects'], ['doc_id', 'details', 'domain'])
+        dataframe_results = preprocess_and_flatten(json_results, mode='news_analysis')
         return json_results, dataframe_results
     except Exception as e:
         results = {'error': str(e)}
@@ -146,8 +146,8 @@ def claim_verification(text):
         response.raise_for_status()
         # Prepare results for JSON output
         json_results = response.json()
-        # Call preprocess_and_flatten to flatten 'support', 'refute', and 'no_info' fields
-        dataframe_results = preprocess_and_flatten(json_results, ['support', 'refute', 'no_info'], ['doc_id', 'details'])
+        # Flatten 'support', 'refute', and 'no_info' fields
+        dataframe_results = preprocess_and_flatten(json_results, mode='claim_verification')
         return json_results, dataframe_results
     except Exception as e:
         results = {'error': str(e)}