91 lines
3.1 KiB
Python
91 lines
3.1 KiB
Python
from pathlib import Path
|
|
from corpusparser import Parser
|
|
import argparse
|
|
import logging
|
|
import json
|
|
from pymongo import MongoClient
|
|
import sys
|
|
from multiprocessing import Pool
|
|
|
|
logging.basicConfig(level=logging.INFO)
|
|
logger = logging.getLogger(__name__)
|
|
|
|
n_kres_files = -1 # for logging
|
|
|
|
# handler for concurrency
|
|
def _handle_kres_file_tpl(kres_file_tpl):
|
|
kres_file_idx = kres_file_tpl[0]
|
|
kres_file = kres_file_tpl[1]
|
|
logging.info("Handling {} ({}/{})".format(
|
|
kres_file, kres_file_idx, n_kres_files))
|
|
kres_data = kres_parser.parse_xml_file(kres_file)
|
|
if args.output == "file":
|
|
kres_outdir = outdir / "kres_json"
|
|
kres_outdir.mkdir(parents=True, exist_ok=True)
|
|
kres_outfile = kres_outdir / Path(kres_file.name.split(".")[0]).with_suffix(".json")
|
|
with kres_outfile.open("w") as fp:
|
|
json.dump(kres_data, fp)
|
|
elif args.output == "db":
|
|
kres_col = valdb["kres"]
|
|
for sentence in kres_data:
|
|
kres_col.update({"sid": sentence["sid"]}, sentence, upsert=True)
|
|
|
|
if __name__ == "__main__":
|
|
parser = argparse.ArgumentParser(description="Parsing corpora kres and ssj500k.")
|
|
parser.add_argument('--kres-folder', required=True)
|
|
parser.add_argument('--kres-srl-folder', required=True)
|
|
parser.add_argument('--ssj-file', required=True)
|
|
parser.add_argument('--output', required=False, default=None)
|
|
parser.add_argument('--outdir', required=False, default=None)
|
|
parser.add_argument('--dbaddr', required=False, default=None)
|
|
parser.add_argument('--dbuser', required=False, default=None)
|
|
parser.add_argument('--dbpass', required=False, default=None)
|
|
parser.add_argument('--cores', required=False, default=1)
|
|
args = parser.parse_args()
|
|
|
|
outdir = None
|
|
valdb = None
|
|
if args.output == "file":
|
|
outdir = Path(args.outdir)
|
|
outdir.mkdir(parents=True, exist_ok=True)
|
|
elif args.output == "db":
|
|
dbclient = MongoClient(
|
|
"mongodb://{}".format(args.dbaddr),
|
|
username=args.dbuser,
|
|
password=args.dbpass,
|
|
authSource="valdb",
|
|
authMechanism='SCRAM-SHA-1'
|
|
)
|
|
valdb = dbclient.valdb
|
|
|
|
# SSJ
|
|
logger.info("Parsing Ssj: {}".format(args.ssj_file))
|
|
ssj_parser = Parser(corpus="ssj")
|
|
ssj_data = ssj_parser.parse_xml_file(Path(args.ssj_file))
|
|
if args.output == "file":
|
|
ssj_outfile = outdir / "ssj500k.json"
|
|
with ssj_outfile.open("w") as fp:
|
|
json.dump(ssj_data, fp)
|
|
elif args.output == "db":
|
|
ssj_col = valdb["ssj"]
|
|
for sentence in ssj_data:
|
|
ssj_col.update({"sid": sentence["sid"]}, sentence, upsert=True)
|
|
|
|
|
|
# Kres
|
|
logger.info("Parsing Kres: {}".format(args.kres_folder))
|
|
kres_parser = Parser(
|
|
corpus="kres",
|
|
kres_srl_folder=args.kres_srl_folder
|
|
)
|
|
|
|
# [(idx, filepath)]
|
|
kres_files = [x for x in Path(args.kres_folder).iterdir()]
|
|
kres_files = [x for x in enumerate(kres_files)]
|
|
n_kres_files = len(kres_files)
|
|
|
|
p = Pool(int(args.cores))
|
|
p.map(_handle_kres_file_tpl, kres_files)
|
|
|
|
logger.info("Finished parsing.")
|