Add the aggregated correlations table

change_visualizer.py

@@ -110,8 +110,8 @@ if __name__ == '__main__':
             gr.Markdown(f"### Reference-only correlations")
             gr.Markdown(value=analysis_util.get_correlations_for_groups(df_synthetic, right_side="ind").to_markdown())
 
-            # gr.Markdown(f"### Aggregated correlations")
-            # gr.Markdown(value=analysis_util.get_correlations_for_groups(df_synthetic, right_side="aggr").to_markdown())
+            gr.Markdown(f"### Aggregated correlations")
+            gr.Markdown(value=analysis_util.get_correlations_for_groups(df_synthetic, right_side="aggr").to_markdown())
 
         application.load(update_dataset_view_manual, inputs=slider_manual,
                          outputs=view_manual)

generation_steps/metrics_analysis.py

@@ -1,6 +1,3 @@
-import functools
-import operator
-
 import Levenshtein
 import evaluate
 import pandas as pd
@@ -85,6 +82,7 @@ def edit_distance_fn(pred, ref, **kwargs):
         return sum(scores) / len(scores)
     return Levenshtein.distance(pred, ref)
 
+
 def edit_distance_norm_fn(pred, ref, **kwargs):
     if "refs" in kwargs:
         scores = [Levenshtein.distance(pred, ref) / len(pred) for ref in kwargs["refs"]]
@@ -227,7 +225,6 @@ def compute_metrics(df):
     return df
 
 
-
 def compute_correlations(df: pd.DataFrame):
     grouped_df = df.groupby(by=["end_to_start", "start_to_end"])
     correlations = grouped_df.apply(correlations_for_group, include_groups=False)