Update app.py

app.py

@@ -1,18 +1,25 @@
 import pandas as pd
 import gradio as gr
-import os
 
-def compare_csv_files():
+def compare_csv_files(selected_languages):
     max_num = 10
     
+    # Load data
     df1 = pd.read_csv("result_1.5.csv")
     df2 = pd.read_csv("result_1.4.csv")
     
-    merged_df = pd.merge(df1, df2, on="SourceText", suffixes=("_1.5", "_1.4"))
+    # Merge with Language column
+    merged_df = pd.merge(df1, df2, on=["SourceText", "Language"], suffixes=("_1.5", "_1.4"))
     
+    # Filter by selected languages
+    if selected_languages:
+        merged_df = merged_df[merged_df["Language"].isin(selected_languages)]
+    
+    # Calculate differences
     merged_df["WordErrorRate_Diff"] = merged_df["WordErrorRate_1.5"] - merged_df["WordErrorRate_1.4"]
     merged_df["CharacterErrorRate_Diff"] = merged_df["CharacterErrorRate_1.5"] - merged_df["CharacterErrorRate_1.4"]
     
+    # Add comparison columns
     merged_df["WordErrorRate_Comparison"] = merged_df["WordErrorRate_Diff"].apply(
         lambda x: "1.4 is the same as 1.5 (Ignored due to large diff)" if abs(x) > max_num else (
             f"1.5 is stronger than 1.4 ({x:.8f})" if x < 0 else (
@@ -28,6 +35,7 @@ def compare_csv_files():
         )
     )
     
+    # Overall averages
     avg_word_diff = merged_df["WordErrorRate_Diff"].loc[merged_df["WordErrorRate_Diff"].abs() <= max_num].mean()
     avg_char_diff = merged_df["CharacterErrorRate_Diff"].loc[merged_df["CharacterErrorRate_Diff"].abs() <= 1].mean()
     overall_summary = f"""
@@ -36,19 +44,25 @@ def compare_csv_files():
     <p>Average CharacterErrorRate Difference (excluding large diffs): {f'1.5 is stronger ({avg_char_diff:.8f})' if avg_char_diff < 0 else f'1.4 is stronger ({0 - avg_char_diff:.8f})'}</p>
     """
     
+    # Generate result HTML
     result_html = overall_summary + merged_df[[
+        "Language",
         "SourceText",
         "WordErrorRate_1.5", "WordErrorRate_1.4", "WordErrorRate_Comparison",
         "CharacterErrorRate_1.5", "CharacterErrorRate_1.4", "CharacterErrorRate_Comparison",
-        # "WhisperText_1.5", "WhisperText_1.4"
     ]].to_html(escape=False, index=False)
     
     return result_html
 
+# Load unique languages from the data
+df1 = pd.read_csv("result_1.5.csv")
+df2 = pd.read_csv("result_1.4.csv")
+languages = sorted(set(df1["Language"]).union(set(df2["Language"])))
+
 gr.Interface(
     fn=compare_csv_files,
-    inputs=None,
+    inputs=gr.CheckboxGroup(choices=languages, label="Select Languages to Compare"),
     outputs="html",
     title="Fish Speech Benchmark",
-    description="This is a non-official model performance test from Fish Speech / Whisper Base / More data will be added later (not too much)"
+    description="Select specific languages to compare the results of WordErrorRate and CharacterErrorRate."
 ).launch()