nemo_curator.stages.audio.tagging.merge_alignment_diarization

Merge Alignment and Diarization Stage.

Module Contents

Classes

Name	Description
`MergeAlignmentDiarizationStage`	Stage that merges alignment and diarization information.

API

class nemo_curator.stages.audio.tagging.merge_alignment_diarization.MergeAlignmentDiarizationStage(
    text_key: str = 'text',
    words_key: str = 'words',
    name: str = 'MergeAlignmentDiarization'
)

Dataclass

Bases: ProcessingStage[AudioTask, AudioTask]

Stage that merges alignment and diarization information.

Takes a jsonl data containing both alignment and diarization information and merges the alignment info into the diarization segments.

Example: .. code-block:: yaml

target: nemo_curator.stages.audio.tagging.merge_alignment_diarization.MergeAlignmentDiarizationStage text_key: “text” words_key: “words”

Parameters:

text_key

strDefaults to 'text'

Key to add text to segments

words_key

strDefaults to 'words'

Key to add word alignments to segments

Returns:

The same data as in the input manifest, but with alignment information merged into

name

str = 'MergeAlignmentDiarization'

text_key

str = 'text'

words_key

str = 'words'

nemo_curator.stages.audio.tagging.merge_alignment_diarization.MergeAlignmentDiarizationStage.align_words_to_segments(
    alignment: list[dict],
    segments: list[dict],
    text_key: str,
    words_key: str
) -> None

staticmethod

Align words to segments based on timestamps.

Iterates through the alignment and finds words that belong in each segment, joining them together to form the text for the segment.

Alignment example: [ { “word”: “Hello”, “start”: 0.0, “end”: 1.0 },… ]

Segments example: [ { “speaker”: “speaker1”, “start”: 0.0, “end”: 3.0 },… ]

Output: [ { “speaker”: “speaker1”, “start”: 0.0, “end”: 3.0, “text”: “Hello there”, “words”: [ { “word”: “Hello”, “start”: 0.0, “end”: 1.0 }, { “word”: “there”, “start”: 1.0, “end”: 3.0 },… ] },… ]

Parameters:

alignment

list[dict]

List of words with start and end times

segments

list[dict]

List of segments with start and end times

text_key

str

Key to add text to segment

words_key

str

Key to add words to segment

Returns: None

None

nemo_curator.stages.audio.tagging.merge_alignment_diarization.MergeAlignmentDiarizationStage.inputs() -> tuple[list[str], list[str]]

nemo_curator.stages.audio.tagging.merge_alignment_diarization.MergeAlignmentDiarizationStage.outputs() -> tuple[list[str], list[str]]

nemo_curator.stages.audio.tagging.merge_alignment_diarization.MergeAlignmentDiarizationStage.process(
    task: nemo_curator.tasks.AudioTask
) -> nemo_curator.tasks.AudioTask

Process entry to merge alignment and diarization.