add phones

libriheavy.py

@@ -63,6 +63,13 @@ class Libriheavy(datasets.GeneratorBasedBuilder):
                             "word": datasets.Value("string"),
                         }
                     ),
+                    "phone_segments": datasets.Sequence(
+                        {
+                            "start": datasets.Value("int32"),
+                            "end": datasets.Value("int32"),
+                            "phone": datasets.Value("string"),
+                        }
+                    ),
                     "mel_spectrogram": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
                 }
             ),
@@ -154,6 +161,9 @@ class Libriheavy(datasets.GeneratorBasedBuilder):
                         "word_segments": [
                             {"start": segment[0], "end": segment[1], "word": segment[2]} for segment in utterance["word_segments"]
                         ],
+                        "phone_segments": [
+                            {"start": segment[0], "end": segment[1], "word": segment[2]} for segment in utterance["phone_segments"]
+                        ],
                         "mel_spectrogram": npz[str(utterance_id)].item()["mel"][0][0],
                     }
                     yield chunk["speaker_id"] + "_" + utterance_id, result