Create app.py

app.py

@@ -0,0 +1,50 @@
+import gradio as gr
+from jiwer import wer, compute_measures
+
+# Function to highlight errors
+def highlight_errors(ground_truth, hypothesis):
+    measures = compute_measures(ground_truth, hypothesis)
+    gt_words = ground_truth.split()
+    hyp_words = hypothesis.split()
+
+    highlighted_hyp = []
+    for op in measures['ops']:
+        if op[0] == 'hit':
+            highlighted_hyp.append(f'<span>{op[1]}</span>')
+        elif op[0] == 'sub':
+            highlighted_hyp.append(f'<span style="color:orange;">{op[2]}</span>')  # Substitution
+        elif op[0] == 'ins':
+            highlighted_hyp.append(f'<span style="color:red;">{op[2]}</span>')  # Insertion
+        elif op[0] == 'del':
+            highlighted_hyp.append(f'<span style="color:blue; text-decoration:line-through;">{op[1]}</span>')  # Deletion
+    
+    highlighted_hyp = ' '.join(highlighted_hyp)
+    error_rate = wer(ground_truth, hypothesis)
+    
+    # Color Legend HTML
+    legend_html = """
+    <div style="margin-top: 10px;">
+        <strong>Legend:</strong><br>
+        <span style="color:orange;">Substitution</span>: Words in orange<br>
+        <span style="color:red;">Insertion</span>: Words in red<br>
+        <span style="color:blue; text-decoration:line-through;">Deletion</span>: Words in blue with strikethrough<br>
+    </div>
+    """
+
+    return highlighted_hyp + legend_html, error_rate, measures['substitutions'], measures['insertions'], measures['deletions']
+
+# Gradio Interface
+interface = gr.Interface(
+    fn=highlight_errors,
+    inputs=["text", "text"],
+    outputs=[
+        gr.HTML(label="Highlighted Transcript with Legend"),
+        gr.Number(label="Word Error Rate"),
+        gr.Number(label="Substitutions"),
+        gr.Number(label="Insertions"),
+        gr.Number(label="Deletions")
+    ],
+    title="WER Calculator with Error Highlighting and Legend"
+)
+
+interface.launch()