Update app.py

app.py

@@ -15,16 +15,9 @@ def load_and_analyze_csv(file, text_field, event_model):
     if text_field not in df.columns:
         raise gr.Error(f"Error: Enter text column'{text_field}' not in CSV file.")
 
-    # floods, fires, nones, scores = [], [], [], []
     labels, scores = [], []
     for post in df[text_field].to_list():
         res = classify(post, event_model, HFTOKEN)
-        # if res["event"] == 'flood':
-        #     floods.append(post)
-        # elif res["event"] == 'fire':
-        #     fires.append(post)
-        # else:
-        #     nones.append(post)
         labels.append(res["event"])
         scores.append(res["score"])
 
@@ -46,11 +39,21 @@ def analyze_selected_texts(selections):
     result_df = pd.DataFrame({"Selected Text": selected_texts, "Analysis": analysis_results})
     return result_df
 
-def calculate_accuracy(flood_selections, fire_selections, none_selections, num_posts):
-    incorrect = len(flood_selections) + len(fire_selections) + len(none_selections)
+def calculate_accuracy(flood_selections, fire_selections, none_selections, num_posts, text_field, data_df):
+    posts = data_df[text_field].to_list()
+    selections = flood_selections + fire_selections + none_selections
+    eval = []
+    for post in posts:
+        if post in selections:
+            eval.append("incorrect")
+        else:
+            eval.append("correct")
+
+    data_df["model_eval"] = eval
+    incorrect = len(selections)
     correct = num_posts - incorrect
     accuracy = (correct/num_posts)*100
-    return incorrect, correct, accuracy
+    return incorrect, correct, accuracy, data_df
     
 
             
@@ -117,12 +120,17 @@ with gr.Blocks() as demo:
 
         accuracy_button = gr.Button("Calculate Accuracy")
         num_posts = gr.Number(visible=False)
-        data = gr.DataFrame() #visible=False
+        data = gr.DataFrame(visible=False) 
+        data_eval = gr.DataFrame()
         
-        predict_button.click(load_and_analyze_csv, inputs=[file_input, text_field, event_model], 
-                         outputs=[flood_checkbox_output, fire_checkbox_output, none_checkbox_output, model_confidence, num_posts, data])
-        accuracy_button.click(calculate_accuracy, inputs=[flood_checkbox_output, fire_checkbox_output, none_checkbox_output, num_posts], 
-                             outputs=[incorrect, correct, accuracy])
+        predict_button.click(
+            load_and_analyze_csv, 
+            inputs=[file_input, text_field, event_model], 
+            outputs=[flood_checkbox_output, fire_checkbox_output, none_checkbox_output, model_confidence, num_posts, data])
+        accuracy_button.click(
+            calculate_accuracy, 
+            inputs=[flood_checkbox_output, fire_checkbox_output, none_checkbox_output, num_posts, text_field, data], 
+            outputs=[incorrect, correct, accuracy, data_eval])
                 
     with gr.Tab("Question Answering"):
         # XXX Add some button disabling here, if the classification process is not completed first XXX