File size: 3,381 Bytes
51e84d2 fc1dcbc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 | # License: Apache-2.0 License
# Created by: Patrick Lumbantobing, VertoX-AI
# Copyright (c) 2026 VertoX-AI. All rights reserved.
#
# This work is licensed under the Apache-2.0 License.
# To view a copy of this license, visit
# https://www.apache.org/licenses/LICENSE-2.0
"""Extract exactly the six contract boundary fields from validated alignments."""
from __future__ import annotations
from dataclasses import dataclass
from speech_text_data_aligners.core.errors import AlignmentValidationError
from speech_text_data_aligners.core.models import AlignmentRequest, AlignmentResult
from speech_text_data_aligners.core.validation import validate_alignment_result
BOUNDARY_COLUMNS = (
"first_word_text",
"first_word_start_s",
"first_word_end_s",
"last_word_text",
"last_word_start_s",
"last_word_end_s",
)
@dataclass(frozen=True, slots=True)
class BoundaryFields:
"""Hold original-facing first/last lexical surfaces and acoustic boundaries.
Parameters
----------
first_word_text: Original surface of the first lexical word.
first_word_start_s: Acoustic start of the first lexical word.
first_word_end_s: Acoustic end of the first lexical word.
last_word_text: Original surface of the last lexical word.
last_word_start_s: Acoustic start of the last lexical word.
last_word_end_s: Acoustic end of the last lexical word.
"""
first_word_text: str
first_word_start_s: float
first_word_end_s: float
last_word_text: str
last_word_start_s: float
last_word_end_s: float
def as_dict(self) -> dict[str, str | float]:
"""Return fields in the exact output-schema insertion order.
Returns:
A new ordered dictionary with the six contract column names.
"""
return {
"first_word_text": self.first_word_text,
"first_word_start_s": self.first_word_start_s,
"first_word_end_s": self.first_word_end_s,
"last_word_text": self.last_word_text,
"last_word_start_s": self.last_word_start_s,
"last_word_end_s": self.last_word_end_s,
}
def extract_boundary_fields(
request: AlignmentRequest,
result: AlignmentResult,
) -> BoundaryFields:
"""Validate a result and select its first/last lexical alignments.
A one-word utterance intentionally uses the same record for both endpoints.
Parameters
----------
request: Authoritative request used for semantic validation.
result: Candidate complete result.
Returns
-------
Exactly six backend-neutral boundary values using original word surfaces.
Raises
------
AlignmentValidationError: If validation fails or no aligned words exist.
TranscriptError: If the authoritative transcript has no lexical words.
"""
validated = validate_alignment_result(request, result)
if not validated.words:
raise AlignmentValidationError("cannot extract boundaries from no words")
first = validated.words[0]
last = validated.words[-1]
return BoundaryFields(
first_word_text=first.lexical_word.text,
first_word_start_s=first.start_s,
first_word_end_s=first.end_s,
last_word_text=last.lexical_word.text,
last_word_start_s=last.start_s,
last_word_end_s=last.end_s,
)
|