You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
422 lines
15 KiB
422 lines
15 KiB
import os
|
|
from xml.etree import ElementTree
|
|
import logging
|
|
import re
|
|
import sys
|
|
import gzip
|
|
import pathlib
|
|
from io import StringIO
|
|
|
|
from luscenje_struktur.progress_bar import progress
|
|
from luscenje_struktur.word import Word
|
|
|
|
|
|
def is_root_id(id_):
|
|
return len(id_.split('.')) == 3
|
|
|
|
|
|
def load_files(args, database, w_collection=None, input_corpus=None):
|
|
filenames = input_corpus if input_corpus is not None else args.input
|
|
skip_id_check = args.skip_id_check
|
|
do_msd_translate = not args.no_msd_translate
|
|
|
|
if len(filenames) == 1 and os.path.isdir(filenames[0]):
|
|
filenames = [os.path.join(filenames[0], file) for file in os.listdir(filenames[0]) if file[-5:] != '.zstd']
|
|
|
|
if len(filenames) > 1:
|
|
filenames = [filename for filename in filenames if filename[-5:] != '.zstd']
|
|
filenames = sorted(filenames, key=lambda x: int(x.split('.')[-1]))
|
|
|
|
database.init("CREATE TABLE Files ( filename varchar(2048) )")
|
|
|
|
for idx, fname in enumerate(filenames):
|
|
logging.info("FILE " + fname + "{}/{}".format(idx, len(filenames)))
|
|
extension = pathlib.Path(fname).suffix
|
|
|
|
# check if file with the same name already loaded...
|
|
loaded = database.execute("SELECT * FROM Files WHERE filename=?", (fname,)).fetchone()
|
|
if loaded is not None:
|
|
logging.info("ALREADY LOADED")
|
|
continue
|
|
|
|
if extension == ".xml":
|
|
et = load_xml(fname)
|
|
if input_corpus is None:
|
|
yield file_sentence_generator(et, args)
|
|
else:
|
|
sentence_generator = file_sentence_generator_valency(et, skip_id_check, do_msd_translate, args.pc_tag, w_collection)
|
|
for sent_id, sentence, othr_attributes in sentence_generator:
|
|
yield sent_id, sentence, othr_attributes
|
|
elif extension == ".gz":
|
|
if input_corpus is None:
|
|
yield load_csv(fname, True)
|
|
else:
|
|
sentences = load_csv_valency(fname, True, w_collection)
|
|
for sentence in sentences:
|
|
yield sentence
|
|
elif extension == ".conllu":
|
|
if input_corpus is None:
|
|
yield load_conllu(fname)
|
|
else:
|
|
raise Exception('conllu with input_corpus is not supported!')
|
|
else:
|
|
if input_corpus is None:
|
|
yield load_csv(fname, False)
|
|
else:
|
|
sentences = load_csv_valency(fname, False, w_collection)
|
|
for sentence in sentences:
|
|
yield sentence
|
|
|
|
database.execute("INSERT INTO Files (filename) VALUES (?)", (fname,))
|
|
database.commit()
|
|
|
|
|
|
def lines_gz(filename):
|
|
with gzip.open(filename, 'r') as fp:
|
|
for line in progress(fp, 'load-gz'):
|
|
yield line.decode('utf8')
|
|
|
|
|
|
def lines_csv(filename):
|
|
with open(filename, 'r') as fp:
|
|
for line in progress(fp, 'load-csv'):
|
|
yield line
|
|
|
|
|
|
def load_conllu(filename):
|
|
import conllu
|
|
result = []
|
|
bad_sentence = False
|
|
|
|
words = {}
|
|
links = []
|
|
|
|
def sentence_end(bad_sentence, sent_id):
|
|
if bad_sentence:
|
|
return
|
|
|
|
for lfrom, ldest, ana in links:
|
|
if lfrom not in words or ldest not in words:
|
|
logging.warning("Bad link in sentence: " + sent_id)
|
|
continue
|
|
words[lfrom].add_link(ana, words[ldest])
|
|
result.extend(words.values())
|
|
|
|
with open(filename, 'r') as f:
|
|
data = f.read()
|
|
# conlls = conllu.parse_incr(StringIO(data))
|
|
# for sent in conlls:
|
|
# try:
|
|
# for word in sent:
|
|
# full_id = "{}.{}".format(sent.metadata['sent_id'], str(word['id']))
|
|
# words[str(word['id'])] = Word(word['id'], word['xpos'], full_id, word['form'], False)
|
|
# except:
|
|
# logging.error(f"Error while reading file {filename} in sentence {sent.metadata['sent_id']}. Check if required data is available!")
|
|
|
|
conlls = conllu.parse_incr(StringIO(data))
|
|
# build dep parse
|
|
for sent in conlls:
|
|
try:
|
|
# adding fake word
|
|
words['0'] = Word('', '', '0', '', False, True)
|
|
for word in sent:
|
|
if type(word['id']) == tuple:
|
|
continue
|
|
full_id = "{}.{}".format(sent.metadata['sent_id'], str(word['id']))
|
|
words[str(word['id'])] = Word(word['lemma'], word['upos'], full_id, word['form'], False)
|
|
links.append((str(word['head']), str(word['id']), word['deprel']))
|
|
sentence_end(False, sent.metadata['sent_id'])
|
|
links = []
|
|
words = {}
|
|
except:
|
|
links = []
|
|
words = {}
|
|
logging.error(f"Error while reading file {filename} in sentence {sent.metadata['sent_id']}. Check if required data is available!")
|
|
|
|
return result
|
|
|
|
|
|
def load_csv(filename, compressed):
|
|
result = []
|
|
bad_sentence = False
|
|
|
|
words = {}
|
|
links = []
|
|
|
|
def sentence_end(bad_sentence):
|
|
if bad_sentence:
|
|
return
|
|
|
|
for lfrom, ldest, ana in links:
|
|
if lfrom not in words or ldest not in words:
|
|
logging.warning("Bad link in sentence: " + line_split[0])
|
|
continue
|
|
words[lfrom].add_link(ana, words[ldest])
|
|
result.extend(words.values())
|
|
|
|
line_gen = lines_gz if compressed else lines_csv
|
|
for line in line_gen(filename):
|
|
line_str = line.strip()
|
|
line_fixed = line_str.replace(',', '\t').replace('\t\t\t', '\t,\t')
|
|
line_split = line_fixed.split("\t")
|
|
|
|
if line_split[1] == "1" and len(words) > 0:
|
|
# adding fake word
|
|
words['0'] = Word('', '', '0', '', False, True)
|
|
sentence_end(bad_sentence)
|
|
bad_sentence = False
|
|
links = []
|
|
words = {}
|
|
|
|
try:
|
|
sid, wid, text, msd, lemma, link_src, link_type = line_split
|
|
except ValueError:
|
|
bad_sentence = True
|
|
full_id = "{}.{}".format(sid, wid)
|
|
|
|
words[wid] = Word(lemma, msd, full_id, text, True)
|
|
# if link_src != '0':
|
|
links.append((link_src, wid, link_type))
|
|
|
|
# adding fake word
|
|
words['0'] = Word('', '', '0', '', False, True)
|
|
sentence_end(bad_sentence)
|
|
return result
|
|
|
|
|
|
def load_csv_valency(filename, compressed, w_collection):
|
|
# TODO skip sentences that are not in sentences of interest!!!
|
|
result = {}
|
|
bad_sentence = False
|
|
|
|
words = {}
|
|
links = []
|
|
idi = 0
|
|
|
|
def sentence_end(bad_sentence, sid):
|
|
if bad_sentence:
|
|
return
|
|
|
|
for lfrom, ldest, ana in links:
|
|
if lfrom not in words or ldest not in words:
|
|
logging.warning("Bad link in sentence: " + line_split[0])
|
|
continue
|
|
words[lfrom].add_link(ana, words[ldest])
|
|
result[sid] = list(words.values())
|
|
|
|
line_gen = lines_gz if compressed else lines_csv
|
|
for line in line_gen(filename):
|
|
line_str = line.strip()
|
|
line_fixed = line_str.replace('\t\t\t', '\t,\t')
|
|
line_split = line_fixed.split("\t")
|
|
|
|
if line_split[1] == "1" and len(words) > 0:
|
|
sentence_end(bad_sentence, sid)
|
|
bad_sentence = False
|
|
links = []
|
|
words = {}
|
|
idi = 0
|
|
|
|
try:
|
|
sid, wid, text, msd, lemma, link_src, link_type = line_split
|
|
except ValueError:
|
|
bad_sentence = True
|
|
full_id = "{}.{}".format(sid, wid)
|
|
|
|
words[wid] = Word(lemma, msd, full_id, text, True)
|
|
if not (len(text[0]) == 1 and re.match('^[\w]+$', text[0]) is None):
|
|
words[wid].idi = str(idi)
|
|
idi += 1
|
|
|
|
if link_src != '0':
|
|
links.append((link_src, wid, link_type))
|
|
|
|
sentence_end(bad_sentence, sid)
|
|
|
|
sentence_ids = list(result.keys())
|
|
cur = w_collection.find({'_id': {'$in': sentence_ids}})
|
|
cur = [c for c in cur]
|
|
unsorted_result = [(c['_id'], result[c['_id']], {k: v for k, v in c.items() if k != '_id'}) for c in cur]
|
|
return sorted(unsorted_result, key=lambda x: (x[0].split('.')[0], int(x[0].split('.')[1]), int(x[0].split('.')[2])))
|
|
|
|
def load_xml(filename):
|
|
with open(filename, 'r') as fp:
|
|
content = fp.read()
|
|
|
|
xmlstring = re.sub(' xmlns="[^"]+"', '', content, count=1)
|
|
xmlstring = xmlstring.replace(' xml:', ' ')
|
|
return ElementTree.XML(xmlstring)
|
|
|
|
|
|
def file_sentence_generator(et, args):
|
|
skip_id_check = args.skip_id_check
|
|
do_msd_translate = not args.no_msd_translate
|
|
pc_tag = args.pc_tag
|
|
use_punctuations = not args.ignore_punctuations
|
|
previous_pc = False
|
|
|
|
words = {}
|
|
paragraphs = list(et.iter('p'))
|
|
for paragraph in progress(paragraphs, "load-text"):
|
|
previous_glue = ''
|
|
sentences = list(paragraph.iter('s'))
|
|
for sentence in sentences:
|
|
# create fake root word
|
|
words[sentence.get('id')] = Word.fake_root_word(sentence.get('id'))
|
|
last_word_id = None
|
|
|
|
if args.new_tei:
|
|
for w in sentence.iter():
|
|
if w.tag == 'w':
|
|
words[w.get('id')] = Word.from_xml(w, do_msd_translate)
|
|
if use_punctuations:
|
|
previous_glue = '' if 'join' in w.attrib and w.get('join') == 'right' else ' '
|
|
elif w.tag == pc_tag:
|
|
words[w.get('id')] = Word.pc_word(w, do_msd_translate)
|
|
if use_punctuations:
|
|
words[w.get('id')].previous_glue = previous_glue
|
|
words[w.get('id')].glue = '' if 'join' in w.attrib and w.get('join') == 'right' else ' '
|
|
previous_glue = '' if 'join' in w.attrib and w.get('join') == 'right' else ' '
|
|
else:
|
|
for w in sentence.iter():
|
|
if w.tag == 'w':
|
|
words[w.get('id')] = Word.from_xml(w, do_msd_translate)
|
|
if use_punctuations:
|
|
previous_glue = ''
|
|
last_word_id = None
|
|
elif w.tag == pc_tag:
|
|
words[w.get('id')] = Word.pc_word(w, do_msd_translate)
|
|
if use_punctuations:
|
|
last_word_id = w.get('id')
|
|
words[w.get('id')].previous_glue = previous_glue
|
|
previous_glue = ''
|
|
elif use_punctuations and w.tag == 'c':
|
|
# always save previous glue
|
|
previous_glue = w.text
|
|
if last_word_id:
|
|
words[last_word_id].glue += w.text
|
|
|
|
for l in sentence.iter("link"):
|
|
if 'dep' in l.keys():
|
|
ana = l.get('afun')
|
|
lfrom = l.get('from')
|
|
dest = l.get('dep')
|
|
else:
|
|
ana = l.get('ana')
|
|
if ana[:8] != 'jos-syn:': # dont bother...
|
|
continue
|
|
ana = ana[8:]
|
|
lfrom, dest = l.get('target').replace('#', '').split()
|
|
|
|
if lfrom in words:
|
|
if not skip_id_check and is_root_id(lfrom):
|
|
logging.error("Id {} is not fine, you might want to try with tag --skip-id-check".format(lfrom))
|
|
sys.exit(1)
|
|
|
|
if dest in words:
|
|
next_word = words[dest]
|
|
words[lfrom].add_link(ana, next_word)
|
|
else:
|
|
logging.error("Unknown id: {}".format(dest))
|
|
sys.exit(1)
|
|
|
|
else:
|
|
# strange errors, just skip...
|
|
pass
|
|
|
|
a = list(words.values())
|
|
return list(words.values())
|
|
|
|
|
|
|
|
def file_sentence_generator_valency(et, skip_id_check, do_msd_translate, pc_tag, w_collection):
|
|
words = {}
|
|
sentences = list(et.iter('s'))
|
|
sentence_ids = [s.attrib['id'] for s in sentences]
|
|
cur = w_collection.find({'_id': {'$in': sentence_ids}})
|
|
sentences_of_interest = {c['_id']: {k: v for k, v in c.items() if k != '_id'} for c in cur}
|
|
|
|
for sentence in progress(sentences, "load-text"):
|
|
if sentence.attrib['id'] not in sentences_of_interest:
|
|
continue
|
|
idi = 0
|
|
last_word_id = None
|
|
for w in sentence.iter():
|
|
if w.tag == 'w':
|
|
last_word_id = w.get('id')
|
|
words[last_word_id] = Word.from_xml(w, do_msd_translate)
|
|
words[last_word_id].idi = str(idi)
|
|
idi += 1
|
|
elif w.tag == pc_tag:
|
|
last_word_id = w.get('id')
|
|
words[last_word_id] = Word.pc_word(w, do_msd_translate)
|
|
elif w.tag == 'c':
|
|
if last_word_id:
|
|
words[last_word_id].glue += w.text
|
|
|
|
for l in sentence.iter("link"):
|
|
if 'dep' in l.keys():
|
|
ana = l.get('afun')
|
|
lfrom = l.get('from')
|
|
dest = l.get('dep')
|
|
else:
|
|
ana = l.get('ana')
|
|
if ana[:8] != 'jos-syn:': # dont bother...
|
|
continue
|
|
ana = ana[8:]
|
|
lfrom, dest = l.get('target').replace('#', '').split()
|
|
|
|
if lfrom in words:
|
|
if not skip_id_check and is_root_id(lfrom):
|
|
logging.error("NOO: {}".format(lfrom))
|
|
sys.exit(1)
|
|
|
|
if dest in words:
|
|
next_word = words[dest]
|
|
words[lfrom].add_link(ana, next_word)
|
|
else:
|
|
logging.error("Unknown id: {}".format(dest))
|
|
sys.exit(1)
|
|
|
|
else:
|
|
# strange errors, just skip...
|
|
pass
|
|
yield sentence.attrib['id'], list(words.values()), sentences_of_interest[sentence.attrib['id']]
|
|
words = {}
|
|
|
|
|
|
def file_sentence_glue_generator(files, pc_tag, w_collection):
|
|
for fname in files:
|
|
et = load_xml(fname)
|
|
|
|
words = {}
|
|
sentences = list(et.iter('s'))
|
|
|
|
sentence_ids = [s.attrib['id'] for s in sentences]
|
|
cur = w_collection.find({'_id': {'$in': sentence_ids}})
|
|
sentences_of_interest = {c['_id']: {k: v for k, v in c.items() if k != '_id'} for c in cur}
|
|
|
|
for sentence in progress(sentences, "load-text"):
|
|
if sentence.attrib['id'] not in sentences_of_interest:
|
|
continue
|
|
w_id = 1
|
|
last_word_id = None
|
|
sentence_id = None
|
|
for w in sentence.iter():
|
|
if w.tag == 'w':
|
|
last_word_id = w_id
|
|
words[last_word_id] = [w.text, last_word_id, '']
|
|
w_id += 1
|
|
elif w.tag == pc_tag:
|
|
last_word_id = w_id
|
|
words[last_word_id] = [w.text, last_word_id, '']
|
|
w_id += 1
|
|
elif w.tag == 'c':
|
|
if last_word_id:
|
|
words[last_word_id][2] += w.text
|
|
elif w.tag == 's':
|
|
sentence_id = w.attrib['id']
|
|
|
|
yield (sentence_id, list(words.values()))
|
|
words = {}
|