nemo_curator.stages.text.experimental.translation.evaluation.text_quality

Translation quality metrics for translated and backtranslated text.

Module Contents

Classes

Name	Description
`TextQualityMetricStage`	Compute translation quality metrics for two text columns.

Functions

Name	Description
`compute_text_quality_metric`	Compute one translation quality metric and its pass/fail flag.

API

class nemo_curator.stages.text.experimental.translation.evaluation.text_quality.TextQualityMetricStage(
    name: str = 'TextQualityMetricStage',
    reference_text_field: str = 'text',
    hypothesis_text_field: str = 'backtranslated_text',
    metrics: list[dict[str, typing.Any]] = list(),
    filter_enabled: bool = False,
    pass_column: str = 'is_quality_metric_passed'
)

Dataclass

Bases: ProcessingStage[DocumentBatch, DocumentBatch]

Compute translation quality metrics for two text columns.

filter_enabled

bool = False

hypothesis_text_field

str = 'backtranslated_text'

metrics

list[dict[str, Any]] = field(default_factory=list)

name

str = 'TextQualityMetricStage'

pass_column

str = 'is_quality_metric_passed'

reference_text_field

str = 'text'

nemo_curator.stages.text.experimental.translation.evaluation.text_quality.TextQualityMetricStage.inputs() -> tuple[list[str], list[str]]

nemo_curator.stages.text.experimental.translation.evaluation.text_quality.TextQualityMetricStage.outputs() -> tuple[list[str], list[str]]

nemo_curator.stages.text.experimental.translation.evaluation.text_quality.TextQualityMetricStage.process(
    batch: nemo_curator.tasks.DocumentBatch
) -> nemo_curator.tasks.DocumentBatch

nemo_curator.stages.text.experimental.translation.evaluation.text_quality.compute_text_quality_metric(
    hypothesis: str,
    reference: str,
    metric_type: str,
    threshold: float
) -> tuple[float, bool]

Compute one translation quality metric and its pass/fail flag.