nemo_curator.stages.text.experimental.translation.utils.metadata

View as Markdown

Output helpers for translation metadata and message reconstruction.

Module Contents

Functions

NameDescription
_parse_structured_messagesReturn translated messages when they are already structured.
_set_nestedSet a nested value when the full path already exists.
build_translation_metadataBuild translation metadata as JSON.
merge_faith_scores_into_metadataMerge FAITH scores into existing translation metadata.
reconstruct_messages_with_translationReturn a copy of messages with translated content inserted.

API

nemo_curator.stages.text.experimental.translation.utils.metadata._parse_structured_messages(
translated_text: object
) -> list[dict[str, object]] | None

Return translated messages when they are already structured.

nemo_curator.stages.text.experimental.translation.utils.metadata._set_nested(
obj: dict[str, object],
keys: list[str],
value: object
) -> None

Set a nested value when the full path already exists.

nemo_curator.stages.text.experimental.translation.utils.metadata.build_translation_metadata(
target_lang: str,
translated_text: str | None = None,
translation_map: dict[str, object] | None = None,
segmented_translation_map: dict[str, object] | None = None
) -> str

Build translation metadata as JSON.

nemo_curator.stages.text.experimental.translation.utils.metadata.merge_faith_scores_into_metadata(
metadata_json: str,
faith_scores: dict[str, object]
) -> str

Merge FAITH scores into existing translation metadata.

nemo_curator.stages.text.experimental.translation.utils.metadata.reconstruct_messages_with_translation(
original_messages: list[dict[str, object]],
translated_text: object,
field_path: str = 'content'
) -> list[dict[str, object]]

Return a copy of messages with translated content inserted.