espnet2.train.dataset.AdapterForLabelScpReader

source

class espnet2.train.dataset.AdapterForLabelScpReader(loader: Dict[str, List[List[str | float | int]]])

Bases: Mapping

AdapterForLabelScpReader is a class that acts as an adapter for reading label

data in a structured format from a provided loader.

This class implements the collections.abc.Mapping interface, allowing it to be used in contexts where a mapping is required. It takes a loader, which is expected to return data in the format of a dictionary where the keys are utterance IDs and the values are lists of label sequences.

loader

A dictionary-like

• Type: Dict[str, List[List[Union[str, float, int]]]]

structure that holds the label data.

• Parameters:
  • loader (Dict *[*str , List *[*List *[*Union *[*str , float , int ] ] ] ]) – A mapping of
  • sequences (utterance IDs to their corresponding label)
  • label (where each)
  • time (end)
  • time
  • and
  • label.
• Returns: A tuple containing: : - sample_time (np.ndarray): A 2D numpy array where each row contains the start and end times of the labels.
  • sample_label (List): A list of labels corresponding to the times.
• Return type: Tuple[np.ndarray, List]

####### Examples

>>> label_data = {
...     "utt1": [[0.0, 1.0, "label1"], [1.0, 2.0, "label2"]],
...     "utt2": [[0.5, 1.5, "label3"]]
... }
>>> reader = AdapterForLabelScpReader(label_data)
>>> times, labels = reader["utt1"]
>>> print(times)
[[0.0, 1.0], [1.0, 2.0]]
>>> print(labels)
['label1', 'label2']

• Raises:
  • AssertionError – If the data format does not match the expected structure
  • or if the output is not as specified. –

keys()

Adapter class for loading label data from a dictionary.

This class provides a mapping interface to retrieve label data stored in a dictionary format. Each entry in the dictionary is expected to be a list of lists, where each inner list contains timing and label information.

loader

A dictionary that maps keys to a list of label data.

• Type: Dict[str, List[List[Union[str, float, int]]]]

• Parameters:loader (Dict *[*str , List *[*List *[*Union *[*str , float , int ] ] ] ]) – The dictionary containing the label data.

• Returns: A tuple containing the sample time as a 2D array and the sample label as a list.

• Return type: np.ndarray

####### Examples

>>> loader = {
...     "utt1": [[0.0, 1.0, "a"], [1.0, 2.0, "b"]],
...     "utt2": [[0.0, 1.5, "c"], [1.5, 3.0, "d"]]
... }
>>> adapter = AdapterForLabelScpReader(loader)
>>> time, labels = adapter["utt1"]
>>> print(time)
[[0.  1. ]
 [1.  2. ]]
>>> print(labels)
['a', 'b']

• Raises:AssertionError – If the retrieved value is not in the expected format.