def get_words(txt):
words = nltk.word_tokenize(txt)
words_1 = []
for w in words:
if analyse(w)[0][0][1] != None:
words_1.append(analyse(w)[0][0][1])
else:
words_1.append(w)
words_1 = [w for w in words_1 if len(w) > 2]
fdist = FreqDist(words_1)
most = fdist.most_common(5)
most = [m[0] for m in most]
return most
def test1(self):
text = 'Mama ma.'
interps = morfeusz.analyse(text)
if sgjp:
self.assertEqual(interps.pop(),
[(u('Mama'), u('mama'), 'subst:sg:nom:f'), (u('ma'), u('mój'), 'adj:sg:voc:f:pos'), (u('.'), u('.'), 'interp')]
)
self.assertEqual(interps, [
[(u('Mama'), u('mama'), 'subst:sg:nom:f'), (u('ma'), u('mieć'), 'fin:sg:ter:imperf'), (u('.'), u('.'), 'interp')],
[(u('Mama'), u('mama'), 'subst:sg:nom:f'), (u('ma'), u('mój'), 'adj:sg:nom:f:pos'), (u('.'), u('.'), 'interp')]
])
def test2(self):
text = u('Miałem miał.')
interps = morfeusz.analyse(text, dag=True)
self.assertEqual(interps, [
(0, 1, (u('Miał'), u('mieć'), u('praet:sg:m1:imperf'))),
(0, 1, (u('Miał'), u('mieć'), u('praet:sg:m2:imperf'))),
(0, 1, (u('Miał'), u('mieć'), u('praet:sg:m3:imperf'))),
(1, 2, (u('em'), u('być'), u('aglt:sg:pri:imperf:wok'))),
(0, 2, (u('Miałem'), u('miał'), u('subst:sg:inst:m3'))),
(2, 3, (u('miał'), u('miał'), u('subst:sg:nom:m3'))),
(2, 3, (u('miał'), u('miał'), u('subst:sg:acc:m3'))),
(2, 3, (u('miał'), u('mieć'), u('praet:sg:m1:imperf'))),
(2, 3, (u('miał'), u('mieć'), u('praet:sg:m2:imperf'))),
(2, 3, (u('miał'), u('mieć'), u('praet:sg:m3:imperf'))),
(3, 4, (u('.'), u('.'), u('interp'))),
])
def get_word_counts(text):
words = defaultdict(int)
tokens = nltk.wordpunct_tokenize(text)
for word in tokens: # for every word in given text -> lemmatize & count
word = re.sub(r'[_+=:;"\'\?/>.<,\\]',' ',word)
if len(word) > 1:
#print word.encode('utf8')
res = morfeusz.analyse(word,expand_tags=False,dag=True) # morfological analyzer for Polish
try:
base = res[0][2][1]
except IndexError:
base = None
#list.append(tup[1])
if base is not None:
words[base] += 1 # increment the word count
else:
pass
return words
def single_word_lemma(word):
# analyzing sometimes fails
try:
interpretations = morfeusz.analyse(word)
except:
print('Parsing failed:', word, sys.exc_info(), file=sys.stderr)
return word
lemmas = set([i[0][1] for i in interpretations])
if len(interpretations) == 1 and interpretations[0][0][2] == 'ign':
# OOD - out of dictionary: maybe we can find this word in a different database
# cases:
# [inflected] acronyms
# inflected names like Murphy'iego
apos_agluts = "'a 'ego 'em 'er 'o 's 'u".split()
lemma = interpretations[0][0][1]
for ending in apos_agluts:
if ending in lemma:
return lemma.replace(ending, '')
if word in lemmas:
# print('Leaving:', word)
return word
if len(lemmas) == 1:
lemma = list(lemmas)[0]
# print('Single candidate:', lemma)
splitted = lemma.split(
':') # there are some flags in Morpheus, like Polska:s2
return splitted[0].capitalize() if word[0].isupper() else splitted[0]
# next step - cleans lemmas of markers, select only same-capitalization
matching_lemmas = set(
lemma.split(':')[0] for lemma in lemmas
if lemma[0].isupper() == word[0].isupper())
if len(matching_lemmas) == 1:
return list(matching_lemmas)[0]
# fallback
return word
def orphaned_examples(test_word=None,
hashtable=None,
online=False,
complete_overwrite=False,
onepage_testmode=False):
buffer_size = 20 #how many words will be printed on one page
if online:
active_words = fetch_active_words(
) # prepare only as many pages as we need at the moment
else:
active_words = {'active': [], 'inactive': [], 'under_review': []}
edit_history = read_edit_history()
excluded_words = active_words['active'] + edit_history['added']
with open('output/empty_sections.txt', 'r') as g:
empty_sections = g.readlines()
random.shuffle(empty_sections)
if not complete_overwrite:
excluded_words += active_words['inactive']
else:
excluded_words += active_words['under_review']
if not hashtable:
authors_hashtable = read_author_hashtable()
else:
authors_hashtable = hashtable
site = pwb.Site()
# this is a dirty trick, because morfAnalyse() and wikilink() don't
# really work as they should. The following regex extracts the first part
# of [[these|links]]
re_base_form = re.compile(r'\[\[(.*?)(?:\||\]\])')
words_count = 0
with open('input/orphans.txt') as f,\
open('output/empty_sections.txt', 'r') as g:
# list of pages with no examples (obtained by empty_section.py)
orphans = f.read()
# for testing purposes
if test_word:
empty_sections = [test_word]
pages_count = 666 if onepage_testmode else 0 #loop helper
output = [] #list-container for examples
for input_word in empty_sections:
if complete_overwrite == False and words_count > 2 * len(
active_words['active']):
with open('output/example_queue.json', 'w') as o:
formatted_output = json.dumps(ordermydict(output),
ensure_ascii=False,
indent=4)
o.write(formatted_output)
return 2
if (pages_count == 101) or (pages_count == 667
and onepage_testmode):
return 0
# dealing with various list formats, e.g. *[[word]]
input_word = input_word.strip('*[]\n')
if len(input_word) < 4 or input_word.upper == input_word:
continue
if input_word in excluded_words:
continue
print(input_word)
if complete_overwrite:
# write to file/page every N words
if len(output) == buffer_size:
formatted_output = json.dumps(ordermydict(output),
ensure_ascii=False,
indent=4)
if online:
while (True):
output_page = pwb.Page(
site,
'Wikisłownik:Dodawanie przykładów/dane/{0:03d}'
.format(pages_count - 1))
if pages_count == 666 or output_page.userName(
) == 'AlkamidBot':
output_page.text = formatted_output
output_page.save(
comment=
'Pobranie nowych przykładów z NKJP.pl')
break
else:
pages_count += 1
if pages_count == 100:
return 0
with open(
'output/json_examples_{0}.json'.format(
pages_count), 'w') as o:
o.write(formatted_output)
pages_count += 1
output = []
if input_word[0] == '-' or input_word[-1] == '-' or input_word[
0].isupper():
continue # let's skip prefixes and sufixes for now, also whatever starts with a capital leter
query = '{0}**'.format(input_word).replace(' ', '** ')
result = nkjp_lookup(query)
root = etree.parse(result).getroot()
#print(xml.dom.minidom.parseString(etree.tostring(root)).toprettyxml())
#return -1
if root.find('concordance') is not None:
found = 0
found_orphan = 0
defs = get_definitions(input_word)
if defs == 0:
continue
new_word = ExampleDict()
new_word['title'] = input_word
new_word['fetch_time'] = str(defs[1])
new_word['definitions'] = defs[0]
for line in root.find('concordance').findall('line'):
sentence = extract_one_sentence(line, input_word)
# NKJP treats gerunds as verb forms. We don't
if '\'\'czasownik' in new_word['definitions'] and\
all(('ger:' in analysed[2] or 'subst:' in analysed[2]) for analysed in morfeusz.analyse(sentence[1])[0]):
continue
if check_sentence_quality(sentence) == 0:
continue
ref = get_reference(line, authors_hashtable)
if ref == '':
break
if len(new_word['examples']) < 2:
temp_example = {
'verificator': 'None',
'correct_num': 'None',
'good_example': False,
'bad_example': False
}
#temp_example['left'] = line.find('left').text
#temp_example['right'] = line.find('right').text
temp_example['example'] = wikitext_one_sentence(
sentence, input_word)
temp_example['left_extra'] = phrases_wikilink(
wikilink(sentence[3]))
temp_example['right_extra'] = phrases_wikilink(
wikilink(sentence[4]))
temp_example['source'] = ref
orphan_switch = check_if_includes_orphan(
sentence, orphans, edit_history['orphans'])
temp_example['orphan'] = orphan_switch
new_word['examples'].append(temp_example)
else:
found_new = 0
wikified_example = wikitext_one_sentence(
sentence, input_word)
for ex_ix, ex in enumerate(new_word['examples']):
neworphan = check_if_includes_orphan(
sentence, orphans, edit_history['orphans'])
if neworphan:
if ex['orphan']:
if wikified_proportion(
ex['example']
) < wikified_proportion(wikified_example):
new_example = new_word['examples'][
ex_ix]
found_new = 1
orphan_switch = neworphan
break
elif not orphan_switch:
new_example = new_word['examples'][ex_ix]
found_new = 1
break
else:
if not ex['orphan']:
if wikified_proportion(
ex['example']
) < wikified_proportion(wikified_example):
new_example = new_word['examples'][
ex_ix]
found_new = 1
break
if found_new:
new_example['orphan'] = neworphan
#new_example['left'] = line.find('left').text
#new_example['right'] = line.find('right').text
new_example['example'] = wikitext_one_sentence(
sentence, input_word)
new_example['left_extra'] = phrases_wikilink(
wikilink(sentence[3]))
new_example['right_extra'] = phrases_wikilink(
wikilink(sentence[4]))
new_example['source'] = ref
if new_word and len(new_word['examples']) > 0:
output.append(new_word)
words_count += 1
def orphaned_examples(test_word=None, hashtable=None, online=False, complete_overwrite=False, onepage_testmode=False):
buffer_size = 20 #how many words will be printed on one page
if online:
active_words = fetch_active_words() # prepare only as many pages as we need at the moment
else:
active_words = {'active': [], 'inactive': [], 'under_review': []}
edit_history = read_edit_history()
excluded_words = active_words['active'] + edit_history['added']
with open('output/empty_sections.txt', 'r') as g:
empty_sections = g.readlines()
random.shuffle(empty_sections)
if not complete_overwrite:
excluded_words += active_words['inactive']
else:
excluded_words += active_words['under_review']
if not hashtable:
authors_hashtable = read_author_hashtable()
else:
authors_hashtable = hashtable
site = pwb.Site()
# this is a dirty trick, because morfAnalyse() and wikilink() don't
# really work as they should. The following regex extracts the first part
# of [[these|links]]
re_base_form = re.compile(r'\[\[(.*?)(?:\||\]\])')
words_count = 0
with open('input/orphans.txt') as f,\
open('output/empty_sections.txt', 'r') as g:
# list of pages with no examples (obtained by empty_section.py)
orphans = f.read()
# for testing purposes
if test_word:
empty_sections = [test_word]
pages_count = 666 if onepage_testmode else 0 #loop helper
output = [] #list-container for examples
for input_word in empty_sections:
if complete_overwrite == False and words_count > 2*len(active_words['active']):
with open('output/example_queue.json', 'w') as o:
formatted_output = json.dumps(ordermydict(output), ensure_ascii=False, indent=4)
o.write(formatted_output)
return 2
if (pages_count == 101) or (pages_count == 667 and onepage_testmode):
return 0
# dealing with various list formats, e.g. *[[word]]
input_word = input_word.strip('*[]\n')
if len(input_word) < 4 or input_word.upper == input_word:
continue
if input_word in excluded_words:
continue
print(input_word)
if complete_overwrite:
# write to file/page every N words
if len(output) == buffer_size:
formatted_output = json.dumps(ordermydict(output), ensure_ascii=False, indent=4)
if online:
while(True):
output_page = pwb.Page(site, 'Wikisłownik:Dodawanie przykładów/dane/{0:03d}'.format(pages_count-1))
if pages_count == 666 or output_page.userName() == 'AlkamidBot':
output_page.text = formatted_output
output_page.save(comment='Pobranie nowych przykładów z NKJP.pl')
break
else:
pages_count += 1
if pages_count == 100:
return 0
with open('output/json_examples_{0}.json'.format(pages_count), 'w') as o:
o.write(formatted_output)
pages_count += 1
output = []
if input_word[0] == '-' or input_word[-1] == '-' or input_word[0].isupper():
continue # let's skip prefixes and sufixes for now, also whatever starts with a capital leter
query = '{0}**'.format(input_word).replace(' ', '** ')
result = nkjp_lookup(query)
root = etree.parse(result).getroot()
#print(xml.dom.minidom.parseString(etree.tostring(root)).toprettyxml())
#return -1
if root.find('concordance') is not None:
found = 0
found_orphan = 0
defs = get_definitions(input_word)
if defs == 0:
continue
new_word = ExampleDict()
new_word['title'] = input_word
new_word['fetch_time'] = str(defs[1])
new_word['definitions'] = defs[0]
for line in root.find('concordance').findall('line'):
sentence = extract_one_sentence(line, input_word)
# NKJP treats gerunds as verb forms. We don't
if '\'\'czasownik' in new_word['definitions'] and\
all(('ger:' in analysed[2] or 'subst:' in analysed[2]) for analysed in morfeusz.analyse(sentence[1])[0]):
continue
if check_sentence_quality(sentence) == 0:
continue
ref = get_reference(line, authors_hashtable)
if ref == '':
break
if len(new_word['examples']) < 2:
temp_example = {'verificator': 'None', 'correct_num': 'None', 'good_example': False, 'bad_example': False}
#temp_example['left'] = line.find('left').text
#temp_example['right'] = line.find('right').text
temp_example['example'] = wikitext_one_sentence(sentence, input_word)
temp_example['left_extra'] = phrases_wikilink(wikilink(sentence[3]))
temp_example['right_extra'] = phrases_wikilink(wikilink(sentence[4]))
temp_example['source'] = ref
orphan_switch = check_if_includes_orphan(sentence, orphans, edit_history['orphans'])
temp_example['orphan'] = orphan_switch
new_word['examples'].append(temp_example)
else:
found_new = 0
wikified_example = wikitext_one_sentence(sentence, input_word)
for ex_ix, ex in enumerate(new_word['examples']):
neworphan = check_if_includes_orphan(sentence, orphans, edit_history['orphans'])
if neworphan:
if ex['orphan']:
if wikified_proportion(ex['example']) < wikified_proportion(wikified_example):
new_example = new_word['examples'][ex_ix]
found_new = 1
orphan_switch = neworphan
break
elif not orphan_switch:
new_example = new_word['examples'][ex_ix]
found_new = 1
break
else:
if not ex['orphan']:
if wikified_proportion(ex['example']) < wikified_proportion(wikified_example):
new_example = new_word['examples'][ex_ix]
found_new = 1
break
if found_new:
new_example['orphan'] = neworphan
#new_example['left'] = line.find('left').text
#new_example['right'] = line.find('right').text
new_example['example'] = wikitext_one_sentence(sentence, input_word)
new_example['left_extra'] = phrases_wikilink(wikilink(sentence[3]))
new_example['right_extra'] = phrases_wikilink(wikilink(sentence[4]))
new_example['source'] = ref
if new_word and len(new_word['examples']) > 0:
output.append(new_word)
words_count += 1
# else every exception is "during handling of OSError..."
restart = True
if restart:
# keyboard interrupts jump out here if raised in wrapper
env['LD_LIBRARY_PATH'] = os.path.realpath(os.path.dirname(__file__))
print('Restarting with LD_LIBRARY_PATH =', env['LD_LIBRARY_PATH'])
subprocess.call([sys.executable] + sys.argv, env=env)
exit()
parser = argparse.ArgumentParser()
parser.add_argument('--dag', action='store_true', help='use dag=True with morfeusz.analyze()')
args = parser.parse_args()
while True:
try:
data = input('morfeusz> ')
except EOFError:
print()
break
except KeyboardInterrupt:
print()
break
if not data:
break
pprint.pprint(morfeusz.analyse(data, dag=args.dag))
print('bye')
