more LM baseline

app.py

@@ -404,24 +404,31 @@ def get_wer_metrics(dataset):
                 "N-best Correction": np.nan
             }
         
-        # Create a transposed DataFrame with metrics as rows and sources as columns
-        metrics = ["Count", "No LM Baseline", "N-best LM Ranking", "N-best Correction"]
-        result_df = pd.DataFrame(index=metrics, columns=["Metric"] + all_sources + ["OVERALL"])
+        # Create flat DataFrame with labels in the first column
+        rows = []
         
-        # Add descriptive column
-        result_df["Metric"] = [
-            "Number of Examples", 
-            "Word Error Rate (No LM)",
-            "Word Error Rate (N-best LM Ranking)",
-            "Word Error Rate (N-best Correction)"
-        ]
+        # First add row for number of examples
+        example_row = {"Metric": "Number of Examples"}
+        for source in all_sources + ["OVERALL"]:
+            example_row[source] = source_results[source]["Count"]
+        rows.append(example_row)
+        
+        # Then add rows for each WER method
+        no_lm_row = {"Metric": "Word Error Rate (No LM)"}
+        lm_ranking_row = {"Metric": "Word Error Rate (N-best LM Ranking)"}
+        n_best_row = {"Metric": "Word Error Rate (N-best Correction)"}
         
         for source in all_sources + ["OVERALL"]:
-            for metric in metrics:
-                result_df.loc[metric, source] = source_results[source][metric]
+            no_lm_row[source] = source_results[source]["No LM Baseline"]
+            lm_ranking_row[source] = source_results[source]["N-best LM Ranking"]
+            n_best_row[source] = source_results[source]["N-best Correction"]
+        
+        rows.append(no_lm_row)
+        rows.append(lm_ranking_row)
+        rows.append(n_best_row)
         
-        # Set Metric as index for better display
-        result_df = result_df.set_index("Metric")
+        # Create DataFrame from rows
+        result_df = pd.DataFrame(rows)
         
         return result_df
     
@@ -438,20 +445,19 @@ def format_dataframe(df):
         
         # Find the rows containing WER values
         wer_row_indices = []
-        for idx in df.index:
-            if "WER" in idx or "Error Rate" in idx:
-                wer_row_indices.append(idx)
+        for i, metric in enumerate(df["Metric"]):
+            if "WER" in metric or "Error Rate" in metric:
+                wer_row_indices.append(i)
         
-        for wer_row_index in wer_row_indices:
-            # Convert to object type first to avoid warnings
-            df.loc[wer_row_index] = df.loc[wer_row_index].astype(object)
-            
+        # Format WER values
+        for idx in wer_row_indices:
             for col in df.columns:
-                value = df.loc[wer_row_index, col]
-                if pd.notna(value):
-                    df.loc[wer_row_index, col] = f"{value:.4f}"
-                else:
-                    df.loc[wer_row_index, col] = "N/A"
+                if col != "Metric":  # Skip the metric column
+                    value = df.loc[idx, col]
+                    if pd.notna(value):
+                        df.loc[idx, col] = f"{value:.4f}"
+                    else:
+                        df.loc[idx, col] = "N/A"
         
         return df