ancient-scripts-datasets / scripts /extract_ancient_language.py
Alvin
Add complete dataset: all sources, metadata, scripts, docs, and phylo enrichment
26786e3
#!/usr/bin/env python3
"""Reusable framework for extracting ancient language data from online sources.
Orchestrates multiple source-specific parsers (ASJP, Wiktionary, Oracc,
eDiAna, Avesta, LRC, DEDR) to build a unified TSV lexicon file for any
ancient or reconstructed language.
Pipeline:
1. For each source in the config, call the appropriate parser
2. Apply transliteration -> IPA mapping via transliteration_maps
3. Generate SCA sound-class codes via cognate_pipeline
4. Deduplicate by (word, ipa)
5. Write output TSV
Usage:
python extract_ancient_language.py \\
--iso hit --name Hittite --family anatolian \\
--source asjp:https://asjp.clld.org/languages/HITTITE \\
--source wiktionary:https://en.wiktionary.org/wiki/Appendix:Hittite_Swadesh_list
Output: data/training/lexicons/{iso}.tsv
"""
from __future__ import annotations
import argparse
import importlib
import logging
import sys
from pathlib import Path
# ---------------------------------------------------------------------------
# Path setup: make cognate_pipeline importable
# ---------------------------------------------------------------------------
ROOT = Path(__file__).resolve().parent.parent
sys.path.insert(0, str(ROOT / "cognate_pipeline" / "src"))
from cognate_pipeline.normalise.sound_class import ipa_to_sound_class # noqa: E402
# ---------------------------------------------------------------------------
# Transliteration maps (same scripts/ directory)
# ---------------------------------------------------------------------------
# transliteration_maps is expected to live alongside this script and
# expose a function: transliterate(translit: str, iso: str) -> str
# If it doesn't exist yet, we provide a passthrough fallback.
try:
sys.path.insert(0, str(Path(__file__).resolve().parent))
from transliteration_maps import transliterate as _transliterate # type: ignore
except ImportError:
logging.getLogger(__name__).warning(
"transliteration_maps not found in scripts/; "
"IPA output will use raw transliteration forms"
)
def _transliterate(translit: str, iso: str) -> str: # type: ignore[misc]
"""Passthrough: return transliteration as-is when no map is available."""
return translit
# ---------------------------------------------------------------------------
# Parser registry
# ---------------------------------------------------------------------------
PARSER_MODULES: dict[str, str] = {
"asjp": "scripts.parsers.parse_asjp",
"wiktionary": "scripts.parsers.parse_wiktionary",
"oracc": "scripts.parsers.parse_oracc",
"ediana": "scripts.parsers.parse_ediana",
"avesta": "scripts.parsers.parse_avesta",
"lrc": "scripts.parsers.parse_lrc",
"dedr": "scripts.parsers.parse_dedr",
}
# Also support direct import if running from the scripts/ directory
_PARSER_MODULES_ALT: dict[str, str] = {
"asjp": "parsers.parse_asjp",
"wiktionary": "parsers.parse_wiktionary",
"oracc": "parsers.parse_oracc",
"ediana": "parsers.parse_ediana",
"avesta": "parsers.parse_avesta",
"lrc": "parsers.parse_lrc",
"dedr": "parsers.parse_dedr",
}
logger = logging.getLogger(__name__)
def _load_parser(source_type: str):
"""Dynamically import and return the parse() function for a source type."""
if source_type not in PARSER_MODULES:
raise ValueError(
f"Unknown source type '{source_type}'. "
f"Available: {', '.join(sorted(PARSER_MODULES))}"
)
# Try the fully-qualified module path first, then the relative one
for module_map in (PARSER_MODULES, _PARSER_MODULES_ALT):
module_name = module_map[source_type]
try:
mod = importlib.import_module(module_name)
return mod.parse
except ImportError:
continue
raise ImportError(
f"Could not import parser module for '{source_type}'. "
f"Tried: {PARSER_MODULES[source_type]}, {_PARSER_MODULES_ALT[source_type]}"
)
def extract_language(config: dict) -> Path:
"""Extract language data from multiple sources and write a unified TSV.
Args:
config: Dictionary with keys:
iso (str): ISO 639-3 code (e.g. "hit" for Hittite)
name (str): Human-readable language name
family (str): Language family (e.g. "anatolian")
sources (list[dict]): List of source descriptors, each with:
type (str): Parser type (asjp, wiktionary, oracc, etc.)
url (str): URL to download from
... additional kwargs passed to the parser
target_dir (str | Path | None): Output directory. Defaults to
data/training/lexicons/ relative to repo root.
Returns:
Path to the output TSV file.
"""
iso = config["iso"]
name = config["name"]
family = config.get("family", "unknown")
sources = config.get("sources", [])
target_dir = config.get("target_dir")
if target_dir is None:
target_dir = ROOT / "data" / "training" / "lexicons"
else:
target_dir = Path(target_dir)
target_dir.mkdir(parents=True, exist_ok=True)
output_path = target_dir / f"{iso}.tsv"
logger.info(
"Extracting %s (%s, family=%s) from %d source(s)",
name, iso, family, len(sources),
)
# ------------------------------------------------------------------
# Step 1: Collect raw entries from all sources
# ------------------------------------------------------------------
all_entries: list[dict] = []
for source in sources:
source_type = source["type"]
source_url = source["url"]
# Pass any extra kwargs to the parser
extra_kwargs = {k: v for k, v in source.items() if k not in ("type", "url")}
logger.info(" Source: %s -> %s", source_type, source_url)
try:
parse_fn = _load_parser(source_type)
raw_entries = parse_fn(source_url, **extra_kwargs)
except Exception:
logger.exception(
" Failed to parse source %s (%s)", source_type, source_url
)
raw_entries = []
# Tag each entry with its source type
for entry in raw_entries:
entry["_source"] = source_type
all_entries.extend(raw_entries)
logger.info(" Total raw entries: %d", len(all_entries))
if not all_entries:
logger.warning(" No entries extracted for %s -- writing empty TSV", name)
output_path.write_text(
"Word\tIPA\tSCA\tSource\tConcept_ID\tCognate_Set_ID\n",
encoding="utf-8",
)
return output_path
# ------------------------------------------------------------------
# Step 2: Transliteration -> IPA -> SCA
# ------------------------------------------------------------------
processed: list[tuple[str, str, str, str, str, str]] = []
for entry in all_entries:
word = entry.get("word", "").strip()
translit = entry.get("transliteration", word).strip()
gloss = entry.get("gloss", "").strip()
source_tag = entry.get("_source", "unknown")
if not word:
continue
# Apply transliteration -> IPA mapping
ipa = _transliterate(translit, iso)
if not ipa:
ipa = translit # Fallback: use raw transliteration
# Generate SCA code
sca = ipa_to_sound_class(ipa)
# Use gloss as a concept ID (normalized)
concept_id = gloss.lower().replace(" ", "_")[:50] if gloss else "-"
cognate_set_id = "-"
processed.append((word, ipa, sca, source_tag, concept_id, cognate_set_id))
# ------------------------------------------------------------------
# Step 3: Deduplicate by (word, ipa)
# ------------------------------------------------------------------
seen: set[tuple[str, str]] = set()
deduplicated: list[tuple[str, str, str, str, str, str]] = []
for entry_tuple in processed:
key = (entry_tuple[0], entry_tuple[1]) # (word, ipa)
if key not in seen:
seen.add(key)
deduplicated.append(entry_tuple)
dupes_removed = len(processed) - len(deduplicated)
if dupes_removed > 0:
logger.info(" Removed %d duplicate entries", dupes_removed)
# ------------------------------------------------------------------
# Step 4: Write output TSV
# ------------------------------------------------------------------
with open(output_path, "w", encoding="utf-8", newline="") as f:
f.write("Word\tIPA\tSCA\tSource\tConcept_ID\tCognate_Set_ID\n")
for word, ipa, sca, source_tag, concept_id, cognate_set_id in sorted(
deduplicated, key=lambda x: x[0].lower()
):
f.write(
f"{word}\t{ipa}\t{sca}\t{source_tag}\t{concept_id}\t{cognate_set_id}\n"
)
logger.info(
" Wrote %d entries to %s", len(deduplicated), output_path
)
return output_path
def _parse_source_arg(source_str: str) -> dict:
"""Parse a --source argument of the form 'type:url'.
Returns:
Dict with keys 'type' and 'url'.
"""
if ":" not in source_str:
raise argparse.ArgumentTypeError(
f"Invalid source format: '{source_str}'. Expected 'type:url'"
)
# Split on first colon only (URLs contain colons)
source_type, url = source_str.split(":", 1)
source_type = source_type.strip().lower()
url = url.strip()
if source_type not in PARSER_MODULES:
raise argparse.ArgumentTypeError(
f"Unknown source type '{source_type}'. "
f"Available: {', '.join(sorted(PARSER_MODULES))}"
)
if not url:
raise argparse.ArgumentTypeError(
f"Empty URL for source type '{source_type}'"
)
return {"type": source_type, "url": url}
def main() -> None:
"""CLI entry point."""
parser = argparse.ArgumentParser(
description="Extract ancient language data from online sources into a TSV lexicon.",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog=(
"Examples:\n"
" python extract_ancient_language.py \\\n"
" --iso hit --name Hittite --family anatolian \\\n"
" --source asjp:https://asjp.clld.org/languages/HITTITE \\\n"
" --source wiktionary:https://en.wiktionary.org/wiki/Appendix:Hittite_Swadesh_list\n"
"\n"
" python extract_ancient_language.py \\\n"
" --iso xur --name Urartian --family hurro-urartian \\\n"
" --source oracc:http://oracc.museum.upenn.edu/ecut\n"
"\n"
f"Available source types: {', '.join(sorted(PARSER_MODULES))}\n"
"Source format: type:url\n"
),
)
parser.add_argument(
"--iso", required=True,
help="ISO 639-3 language code (e.g. hit, xur, xcl)",
)
parser.add_argument(
"--name", required=True,
help="Human-readable language name (e.g. Hittite, Urartian)",
)
parser.add_argument(
"--family", default="unknown",
help="Language family (e.g. anatolian, hurro-urartian, semitic)",
)
parser.add_argument(
"--source", action="append", dest="sources", default=[],
type=_parse_source_arg,
help="Source in format type:url (can be specified multiple times)",
)
parser.add_argument(
"--target-dir", default=None,
help="Output directory (default: data/training/lexicons/)",
)
parser.add_argument(
"--verbose", "-v", action="store_true",
help="Enable verbose logging",
)
args = parser.parse_args()
# Setup logging
logging.basicConfig(
level=logging.DEBUG if args.verbose else logging.INFO,
format="%(asctime)s %(levelname)-8s %(name)s: %(message)s",
datefmt="%H:%M:%S",
)
if not args.sources:
parser.error("At least one --source is required")
config = {
"iso": args.iso,
"name": args.name,
"family": args.family,
"sources": args.sources,
"target_dir": args.target_dir,
}
output_path = extract_language(config)
print(f"\nOutput written to: {output_path}")
if __name__ == "__main__":
main()