File size: 3,271 Bytes
b3b6115 | 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 | """TRCS Hugging Face Dataset Loader Utilities."""
from __future__ import annotations
import json
from pathlib import Path
from typing import Any, Dict, List, Optional
import datasets
PACKAGE_ROOT = Path(__file__).resolve().parent
def get_manifest_path(split: str = "telephone_grid") -> Path:
"""Return absolute path to dataset split manifest."""
split_norm = split.lower().strip()
if split_norm in ("telephone_grid", "e2"):
p = PACKAGE_ROOT / "manifests" / "e2_telephone_grid.jsonl"
if not p.exists():
p = PACKAGE_ROOT / "data" / "telephone_grid" / "test.jsonl"
return p
elif split_norm in ("hallucination", "e1"):
p = PACKAGE_ROOT / "manifests" / "e1_hallucination.jsonl"
if not p.exists():
p = PACKAGE_ROOT / "data" / "hallucination" / "test.jsonl"
return p
elif split_norm == "default":
p = PACKAGE_ROOT / "data" / "default" / "test.jsonl"
return p
else:
raise ValueError(f"Unknown split: {split}. Choose from ['telephone_grid', 'hallucination', 'default']")
def load_trcs_manifest(split: str = "telephone_grid") -> List[Dict[str, Any]]:
"""Load TRCS manifest records as a list of Python dictionaries."""
m_path = get_manifest_path(split)
if not m_path.exists():
raise FileNotFoundError(f"Manifest not found: {m_path}")
records = []
with open(m_path, "r", encoding="utf-8") as f:
for line in f:
line = line.strip()
if line:
rec = json.loads(line)
# Resolve relative audio path to absolute
if "audio_path" in rec:
p = Path(rec["audio_path"])
if not p.is_absolute():
rec["audio_path_abs"] = str(PACKAGE_ROOT / p)
else:
rec["audio_path_abs"] = str(p)
records.append(rec)
return records
def load_trcs_dataset(
split: str = "telephone_grid",
resolve_audio_paths: bool = True,
as_audio: bool = False,
) -> datasets.Dataset:
"""Load TRCS dataset as a Hugging Face `datasets.Dataset` object.
Args:
split: Split name ('telephone_grid'/'e2', 'hallucination'/'e1', or 'default').
resolve_audio_paths: If True, adds 'audio_path_abs' with absolute path to audio files.
as_audio: If True, adds an 'audio' column cast to `datasets.Audio(sampling_rate=16000)`.
"""
split_norm = split.lower().strip()
config_name = "telephone_grid" if split_norm in ("telephone_grid", "e2") else ("hallucination" if split_norm in ("hallucination", "e1") else "default")
ds = datasets.load_dataset(str(PACKAGE_ROOT), config_name, split="test", verification_mode="no_checks")
if resolve_audio_paths or as_audio:
def _add_abs(ex: Dict[str, Any]) -> Dict[str, Any]:
rel_p = Path(ex["audio_path"])
abs_p = str(rel_p if rel_p.is_absolute() else (PACKAGE_ROOT / rel_p))
out = {"audio_path_abs": abs_p}
if as_audio:
out["audio"] = abs_p
return out
ds = ds.map(_add_abs)
if as_audio:
ds = ds.cast_column("audio", datasets.Audio(sampling_rate=16000))
return ds
|