app.py

import pandas as pd
import gradio as gr

def compare_csv_files():
    df1 = pd.read_csv("fish-speech-1.5.csv")
    df2 = pd.read_csv("fish-speech-1.4.csv")
    
    merged_df = pd.merge(df1, df2, on="SourceText", suffixes=("_1.5", "_1.4"))
    
    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"]
    
    merged_df["WordErrorRate_Comparison"] = merged_df["WordErrorRate_Diff"].apply(
        lambda x: f"1.5 is stronger than 1.4 ({x:.8f})" if x > 0 else (
            f"1.4 is stronger than 1.5 ({-x:.8f})" if x < 0 else "1.4 is the same as 1.5 (0)"
        )
    )
    
    merged_df["CharacterErrorRate_Comparison"] = merged_df["CharacterErrorRate_Diff"].apply(
        lambda x: f"1.5 is stronger than 1.4 ({x:.8f})" if x > 0 else (
            f"1.4 is stronger than 1.5 ({-x:.8f})" if x < 0 else "1.4 is the same as 1.5 (0)"
        )
    )
    
    def overall_comparison(row):
        word_diff = row["WordErrorRate_Diff"]
        char_diff = row["CharacterErrorRate_Diff"]
        if word_diff > 0 and char_diff > 0:
            return "1.5 is overall stronger than 1.4"
        elif word_diff < 0 and char_diff < 0:
            return "1.4 is overall stronger than 1.5"
        elif word_diff == 0 and char_diff == 0:
            return "1.4 is the same as 1.5 overall"
        else:
            return "Mixed results: check individual metrics"
    
    merged_df["Overall_Comparison"] = merged_df.apply(overall_comparison, axis=1)
    
    result = merged_df[[
        "SourceText",
        "WordErrorRate_1.5", "WordErrorRate_1.4", "WordErrorRate_Comparison",
        "CharacterErrorRate_1.5", "CharacterErrorRate_1.4", "CharacterErrorRate_Comparison",
        "Overall_Comparison"
    ]]
    
    return result.to_html(index=False)


gr.Interface(
    fn=compare_csv_files,
    inputs=None,
    outputs="html"
).launch()