Use abs delta

app.py

@@ -1,7 +1,7 @@
 import gradio as gr
 import pandas as pd
 import json
-from constants import BANNER, INTRODUCTION_TEXT, CITATION_TEXT, METRICS_TAB_TEXT, DIR_OUTPUT_REQUESTS
+from constants import BANNER, INTRODUCTION_TEXT, CITATION_TEXT, METRICS_TAB_TEXT, DIR_OUTPUT_REQUESTS, CUSTOM_MESSAGE
 from init import is_model_on_hub, upload_file, load_all_info_from_dataset_hub
 from utils_display import AutoEvalColumn, fields, make_clickable_model, styled_error, styled_message
 from datetime import datetime, timezone
@@ -131,11 +131,11 @@ for col in original_df.columns:
 original_df.rename(columns=column_names, inplace=True)
 
 # Compute delta between average WER and CV WER
-original_df['Detla Avg. C.V. WER'] = original_df['Average WER ⬇️'] - original_df['Common Voice WER ⬇️']
-original_df['Detla Avg. C.V. WER'] = pd.to_numeric(original_df['Detla Avg. C.V. WER'], errors='coerce') # Convert to numerical data type
-original_df['Detla Avg. C.V. WER'] = original_df['Detla Avg. C.V. WER'].apply(lambda x: round(x, 2) if not pd.isna(x) else x) # Round and handle NaN values
+original_df['Abs. Detla WER'] = abs(original_df['Average WER ⬇️'] - original_df['Common Voice WER ⬇️'])
+original_df['Abs. Detla WER'] = pd.to_numeric(original_df['Abs. Detla WER'], errors='coerce') # Convert to numerical data type
+original_df['Abs. Detla WER'] = original_df['Abs. Detla WER'].apply(lambda x: round(x, 2) if not pd.isna(x) else x) # Round and handle NaN values
 
-original_df.sort_values(by='Detla Avg. C.V. WER', inplace=True)
+original_df.sort_values(by='Abs. Detla WER', inplace=True)
 
 
 COLS = [c.name for c in fields(AutoEvalColumn)]
@@ -195,21 +195,7 @@ def request_model(model_text, chbcoco2017):
 
 with gr.Blocks() as demo:
     gr.HTML(BANNER, elem_id="banner")
-    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
-    
-    CUSTOM_MESSAGE = """## Legend:
-This space is a fork of the original [hf-audio/open_asr_leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard). It aims to demonstrate how the CommonVoice Test Set provides a relatively accurate approximation of the average WER/CER (Word Error Rate/Character Error Rate) at a significantly lower computational cost.
-
-#### Why is this useful?
-This space offers a way to achieve standardized test set for most languages, enabling us to programmatically select a reasonably effective model for any language supported by CommonVoice.
-
-Model, RTF (1e-3) ⬇️, and Average WER ⬇️ were sourced from [hf-audio/open_asr_leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard) using the version from September 7, 2023.
-
-### Results
-The CommonVoice Test provides a Word Error Rate (WER) within a 20-point margin of the average WER.
-
-While not perfect, this indicates that CommonVoice can be a useful tool for quickly identifying a suitable ASR model for a wide range of languages in a programmatic manner. However, it's important to note that it is not sufficient as the sole criterion for choosing the most appropriate architecture. Further considerations may be needed depending on the specific requirements of your ASR application.
-"""
+    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text") 
     gr.Markdown(CUSTOM_MESSAGE, elem_classes="markdown-text")
     
 

constants.py

@@ -96,3 +96,17 @@ are ranked based on their average WER scores, from lowest to highest.
 
 For more details on the individual datasets and how models are evaluated to give the ESB score, refer to the [ESB paper](https://arxiv.org/abs/2210.13352).
 """
+
+CUSTOM_MESSAGE = """## Legend:
+This space is a fork of the original [hf-audio/open_asr_leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard). It aims to demonstrate how the CommonVoice Test Set provides a relatively accurate approximation of the average WER/CER (Word Error Rate/Character Error Rate) at a significantly lower computational cost.
+
+#### Why is this useful?
+This space offers a way to achieve standardized test set for most languages, enabling us to programmatically select a reasonably effective model for any language supported by CommonVoice.
+
+Model, RTF (1e-3) ⬇️, and Average WER ⬇️ were sourced from [hf-audio/open_asr_leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard) using the version from September 7, 2023.
+
+### Results
+The CommonVoice Test provides a Word Error Rate (WER) within a 20-point margin of the average WER.
+
+While not perfect, this indicates that CommonVoice can be a useful tool for quickly identifying a suitable ASR model for a wide range of languages in a programmatic manner. However, it's important to note that it is not sufficient as the sole criterion for choosing the most appropriate architecture. Further considerations may be needed depending on the specific requirements of your ASR application.
+"""