Spaces:
Sleeping
Sleeping
File size: 4,956 Bytes
af4ff35 1550706 aff6746 af4ff35 aff6746 820ab2f af4ff35 820ab2f af4ff35 820ab2f af4ff35 aff6746 1550706 aff6746 af4ff35 aff6746 af4ff35 aff6746 1550706 aff6746 af4ff35 aff6746 820ab2f af4ff35 aff6746 af4ff35 aff6746 af4ff35 aff6746 c155fa9 aff6746 c155fa9 aff6746 c155fa9 aff6746 af4ff35 aff6746 af4ff35 aff6746 af4ff35 aff6746 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 |
import jiwer
import spaces
import numpy as np
import gradio as gr
@spaces.GPU()
def calculate_wer(reference, hypothesis):
"""
Calculate the Word Error Rate (WER) using jiwer.
"""
wer = jiwer.wer(reference, hypothesis)
return wer
@spaces.GPU()
def calculate_cer(reference, hypothesis):
"""
Calculate the Character Error Rate (CER) using jiwer.
"""
cer = jiwer.cer(reference, hypothesis)
return cer
@spaces.GPU()
def calculate_sentence_wer(reference, hypothesis):
"""
Calculate WER for each sentence and overall statistics.
"""
reference_sentences = jiwer.split_into_sentences(reference)
hypothesis_sentences = jiwer.split_into_sentences(hypothesis)
if len(reference_sentences) != len(hypothesis_sentences):
raise ValueError("Reference and hypothesis must contain the same number of sentences")
sentence_wers = []
for ref, hyp in zip(reference_sentences, hypothesis_sentences):
sentence_wer = jiwer.wer(ref, hyp)
sentence_wers.append(sentence_wer)
if not sentence_wers:
return {
"sentence_wers": [],
"average_wer": 0.0,
"std_dev": 0.0
}
average_wer = np.mean(sentence_wers)
std_dev = np.std(sentence_wers)
return {
"sentence_wers": sentence_wers,
"average_wer": average_wer,
"std_dev": std_dev
}
@spaces.GPU()
def process_files(reference_file, hypothesis_file):
try:
with open(reference_file.name, 'r') as f:
reference_text = f.read()
with open(hypothesis_file.name, 'r') as f:
hypothesis_text = f.read()
wer_value = calculate_wer(reference_text, hypothesis_text)
cer_value = calculate_cer(reference_text, hypothesis_text)
sentence_wer_stats = calculate_sentence_wer(reference_text, hypothesis_text)
return {
"WER": wer_value,
"CER": cer_value,
"Sentence WERs": sentence_wer_stats["sentence_wers"],
"Average WER": sentence_wer_stats["average_wer"],
"Standard Deviation": sentence_wer_stats["std_dev"]
}
except Exception as e:
return {"error": str(e)}
def format_sentence_wer_stats(sentence_wers, average_wer, std_dev):
if not sentence_wers:
return "All sentences match perfectly!"
md = "### Sentence-level WER Analysis\n\n"
md += f"* Average WER: {average_wer:.2f}\n"
md += f"* Standard Deviation: {std_dev:.2f}\n\n"
md += "### WER for Each Sentence\n\n"
for i, wer in enumerate(sentence_wers):
md += f"* Sentence {i+1}: {wer:.2f}\n"
return md
def main():
with gr.Blocks() as demo:
gr.Markdown("# ASR Metrics Calculator")
with gr.Row():
reference_file = gr.File(label="Upload Reference File")
hypothesis_file = gr.File(label="Upload Hypothesis File")
with gr.Row():
reference_preview = gr.Textbox(label="Reference Preview", lines=3)
hypothesis_preview = gr.Textbox(label="Hypothesis Preview", lines=3)
with gr.Row():
compute_button = gr.Button("Compute Metrics")
results_output = gr.JSON(label="Results")
wer_stats_output = gr.Markdown(label="WER Statistics")
# Update previews when files are uploaded
def update_previews(ref_file, hyp_file):
ref_text = ""
hyp_text = ""
if ref_file:
with open(ref_file.name, 'r') as f:
ref_text = f.read()[:200] # Show first 200 characters
if hyp_file:
with open(hyp_file.name, 'r') as f:
hyp_text = f.read()[:200] # Show first 200 characters
return ref_text, hyp_text
reference_file.change(
fn=update_previews,
inputs=[reference_file, hypothesis_file],
outputs=[reference_preview, hypothesis_preview]
)
hypothesis_file.change(
fn=update_previews,
inputs=[reference_file, hypothesis_file],
outputs=[reference_preview, hypothesis_preview]
)
def process_and_display(ref_file, hyp_file):
result = process_files(ref_file, hyp_file)
if "error" in result:
return {}, {}, "Error: " + result["error"]
metrics = {
"WER": result["WER"],
"CER": result["CER"]
}
wer_stats_md = format_sentence_wer_stats(
result["Sentence WERs"],
result["Average WER"],
result["Standard Deviation"]
)
return metrics, wer_stats_md
compute_button.click(
fn=process_and_display,
inputs=[reference_file, hypothesis_file],
outputs=[results_output, wer_stats_output]
)
demo.launch()
if __name__ == "__main__":
main()
|