feat: Enhance curated examples display with individual audio players and add a script for parsing results.

app.py

@@ -277,8 +277,8 @@ def predict_language(audio_path, fam_k=1, fam_thresh=0.0, super_k=1, super_thres
                 table_data.append([name, f"{score:.2%}"])
                 
             if not table_data:
-                return pd.DataFrame(columns=[i18n("prediction_header"), i18n("confidence_header")])
-            return pd.DataFrame(table_data, columns=[i18n("prediction_header"), i18n("confidence_header")])
+                return pd.DataFrame(columns=["Prediction / Predicción", "Confidence / Confianza"])
+            return pd.DataFrame(table_data, columns=["Prediction / Predicción", "Confidence / Confianza"])
 
         df_fam = build_df(fam_probs, fam_k, fam_thresh, label_extractor.family_labels)
         df_super = build_df(super_probs, super_k, super_thresh, label_extractor.super_labels)
@@ -304,18 +304,8 @@ except Exception as e:
     icon_html = ""
 
 # --- Load curated examples ---
-curated_examples = []
-audio_paths = []
-table_headers = [
-    i18n("col_sample"),
-    i18n("col_pred_fam"),
-    i18n("col_pred_super"),
-    i18n("col_pred_code"),
-    i18n("col_act_fam"),
-    i18n("col_act_super"),
-    i18n("col_act_code"),
-    i18n("col_match")
-]
+curated_labels = []
+curated_audio = []
 try:
     with open("results.txt", "r", encoding="utf-8") as f:
         results_content = f.read()
@@ -327,22 +317,8 @@ try:
         block = sample_blocks[i+1].strip()
         audio_path = f"samples/sample{sample_num}.wav"
         if os.path.exists(audio_path):
-            truth_match = re.search(r'Truth:\s*(.*?)\s*→\s*(.*?)\s*→\s*(\w+)', block)
-            pred_match = re.search(r'Pred:\s*(.*?)\s*→\s*(.*?)\s*→\s*(\w+)', block)
-            match_val = re.search(r'Match:\s*([0-9.\s✅]+)', block)
-            
-            t_fam, t_sup, t_code = truth_match.groups() if truth_match else ("-", "-", "-")
-            p_fam, p_sup, p_code = pred_match.groups() if pred_match else ("-", "-", "-")
-            m_val = match_val.group(1).strip() if match_val else "-"
-            
-            row_data = [
-                f"🔊 {sample_num}",
-                p_fam, p_sup, p_code,
-                t_fam, t_sup, t_code,
-                m_val
-            ]
-            curated_examples.append(row_data)
-            audio_paths.append(audio_path)
+            curated_audio.append(audio_path)
+            curated_labels.append(f"Sample {sample_num}:\n{block}")
 except Exception as e:
     print(f"Warning: Could not parse results.txt: {e}")
 
@@ -353,7 +329,8 @@ with gr.Blocks(theme=gr.themes.Ocean()) as demo:
         <div style="text-align: center; padding: 30px; background: linear-gradient(120deg, rgb(2, 132, 199) 0%, rgb(16, 185, 129) 60%, rgb(5, 150, 105) 100%); color: white; border-radius: 15px; margin-bottom: 25px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);">
             {icon_html}
             <h1 style="color: white; margin: 0; font-size: 2.5em;">🦡 Tachiwin AudioId 🦡</h1>
-            <p style="font-size: 1.2em; opacity: 0.9; margin-top: 10px;">{{i18n('subtitle')}}</p>
+            <p style="font-size: 1.2em; opacity: 0.9; margin-top: 10px;">Identify any of the 68 languages of Mexico and their 360 variants<br/>
+            <i>Identifica cualquiera de las 68 lenguas originarias de México y sus 360 variantes</i></p>
         </div>
         """
     )
@@ -389,9 +366,9 @@ with gr.Blocks(theme=gr.themes.Ocean()) as demo:
             
         with gr.Column(scale=1):
             gr.Markdown(i18n("results_header"))
-            fam_table = gr.Dataframe(headers=[i18n("prediction_header"), i18n("confidence_header")], datatype=["str", "str"], label=i18n("pred_fam_label"), interactive=False, wrap=True)
-            super_table = gr.Dataframe(headers=[i18n("prediction_header"), i18n("confidence_header")], datatype=["str", "str"], label=i18n("pred_super_label"), interactive=False, wrap=True)
-            code_table = gr.Dataframe(headers=[i18n("prediction_header"), i18n("confidence_header")], datatype=["str", "str"], label=i18n("pred_code_label"), interactive=False, wrap=True)
+            fam_table = gr.Dataframe(headers=["Prediction / Predicción", "Confidence / Confianza"], datatype=["str", "str"], label=i18n("pred_fam_label"), interactive=False, wrap=True)
+            super_table = gr.Dataframe(headers=["Prediction / Predicción", "Confidence / Confianza"], datatype=["str", "str"], label=i18n("pred_super_label"), interactive=False, wrap=True)
+            code_table = gr.Dataframe(headers=["Prediction / Predicción", "Confidence / Confianza"], datatype=["str", "str"], label=i18n("pred_code_label"), interactive=False, wrap=True)
 
     submit_btn.click(
         fn=predict_language,
@@ -408,15 +385,16 @@ with gr.Blocks(theme=gr.themes.Ocean()) as demo:
     gr.Markdown(i18n("curated_header"))
     gr.Markdown(i18n("curated_desc"))
     
-    example_df = pd.DataFrame(curated_examples, columns=table_headers) if curated_examples else pd.DataFrame(columns=table_headers)
-    
-    sample_table = gr.Dataframe(value=example_df, headers=table_headers, interactive=False, wrap=True)
-
-    def load_sample_audio(evt: gr.SelectData):
-        row = evt.index[0]
-        return audio_paths[row] if row < len(audio_paths) else None
-
-    sample_table.select(fn=load_sample_audio, outputs=[audio_input])
+    with gr.Group():
+        columns = 3
+        for row_idx in range(0, len(curated_audio), columns):
+            with gr.Row():
+                for i in range(row_idx, min(row_idx+columns, len(curated_audio))):
+                    with gr.Column(scale=1):
+                        gr.Markdown(f"```text\n{curated_labels[i]}\n```")
+                        gr.Audio(value=curated_audio[i], interactive=False, show_label=False)
+                        btn = gr.Button("Load / Cargar", size="sm")
+                        btn.click(fn=lambda url=curated_audio[i]: url, inputs=[], outputs=[audio_input])
 
 if __name__ == "__main__":
     demo.launch(ssr_mode=False, i18n=i18n)