Metadata for python doesn't have confidence

transcription = ds.sttWithMetadata(audio, fs)
print(metadata_to_string(transcription))
print(transcription.num_items)
print(transcription.items[0].character)
print(transcription.items[0].timestep)
print(transcription.items[0].start_time)
print(transcription.confidence)



the moon santiago a seminarian in eeeeeeeeeeeeeeeeelllllll nineteen o nine in a eeeeeeeeeeeeeeeeelllllll nineteen o nine eeeeeeeeeeeeeeeeelllllll eeeeeeeeeeeeeeeeelllllll eeeeeeeeeeeeeeeeelllllll
195
t
15
0.29999998211860657
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-27-1b7a0387587d> in <module>
      4 print(transcription.items[0].timestep)
      5 print(transcription.items[0].start_time)
----> 6 print(transcription.confidence)

AttributeError: 'impl.Metadata' object has no attribute 'confidence'
typedef struct MetadataItem {
  /** The character generated for transcription */
  char* character;

  /** Position of the character in units of 20ms */
  int timestep;

  /** Position of the character in seconds */
  float start_time;
} MetadataItem;

/**
 * @brief Stores the entire CTC output as an array of character metadata objects
 */
typedef struct Metadata {
  /** List of items */
  MetadataItem* items;
  /** Size of the list of items */
  int num_items;
  /** Approximated confidence value for this transcription. This is roughly the
   * sum of the acoustic model logit values for each timestep/character that
   * contributed to the creation of this transcription.
   */
  double confidence;
} Metadata;

@Abhinav Care to share the version you are working ? confidence was named differently in previous releases.

TensorFlow: v1.13.1-10-g3e0cc53
DeepSpeech: v0.5.1-0-g4b29b78

So you should check v0.5.1 documentation: https://github.com/mozilla/DeepSpeech/blob/v0.5.1/native_client/deepspeech.h#L19-L31

Thanks lissyx, I had master checked out in my local machine.

1 Like