def process_text(corpus):
import hu_core_ud_lg
nlp = hu_core_ud_lg.load()
doc = nlp(corpus)
return {
str(token): str(token.lemma_)
for token in doc
if token.tag_ not in ['PUNCT', 'NUM'] and not token.is_stop
}
import pickle
import spacy
from spacy.pipeline import EntityRuler
import hu_core_ud_lg
with open("data/interim/entities.p", "rb") as f:
patterns = pickle.load(f)
patterns = [e for e in patterns if len(e["pattern"]) > 0]
nlp = hu_core_ud_lg.load()
ruler = EntityRuler(nlp)
ruler.add_patterns(patterns)
nlp.add_pipe(ruler)
ruler.to_disk("data/processed/patterns.jsonl")
# test it
doc = nlp(
"Magyarország miniszterelnöke Orbán Viktor nem határos az Amerikai Egyesült Államokkal."
)
print([(ent.text, ent.label_) for ent in doc.ents])
doc = nlp("Gyurcsány Ferenc, Orbán Viktor és Antal József miniszterelnökök.")
print([(ent.text, ent.label_) for ent in doc.ents])
doc = nlp(
"Nagy port kavart Márki-Zay Péter hódmezővásárhelyi polgármester, független ellenzéki politikus egy múlt heti interjúja, amiben hosszasan beszélt arról, hogy szerinte elfogadható (sőt szükséges), ha a szülő testi fenyítést alkalmaz gyerekeivel szemben fegyelmezésként."
)
print([(ent.text, ent.label_) for ent in doc.ents])
word.lemma_, # LEMMA
word.pos_, # UPOSTAG
"_", # XPOSTAG
"_", # FEATS
str(head_idx), # HEAD
word.dep_, # DEPREL
"_", # DEPS
"_", # MISC
)
res.append("\t".join(linetuple))
res.append("")
return "\n".join(res) + "\n"
nlp = hu_core_ud_lg.load() # Load models
print(nlp.pipeline) # Print loaded modules
nlp.remove_pipe("tagger")
nlp.remove_pipe("lemmatizer")
#nlp.remove_pipe("parser") # Remove modules on demand
print(nlp.pipeline)
doc = nlp.make_doc('Ennek kell lennie a példamondatnak. Ez egy másik.'
) # Create Document from text.
for name, proc in nlp.pipeline: # iterate over components in order
doc = proc(doc)
print(name)
print(format_as_conllu(doc)) # Retrieve the result in CoNLL-U
----------
Cleaned docs as list of lists
'''
all_tokens = sum(df[tokens_col], [])
tokens_once = set(word for word in set(all_tokens)
if all_tokens.count(word) == 1)
tokenized_no_single = [[word for word in text if word not in tokens_once]
for text in df[tokens_col]]
return tokenized_no_single
if __name__ == '__main__':
# Load hu model
nlp = hu.load()
# Load corpus
path_data = Path('../../data')
df = pd.read_csv(path_data / 'processed' / 'input_data.csv')
# Run tokenizer
df['tokenized_raw'] = df['text'].apply(
lambda x: tokenize_1gram(clean(x), model=nlp, ents=['PER', 'LOC']))
df['tokenized_raw_cnt'] = df['tokenized_raw'].apply(lambda x: len(x))
df['tokenized_mults'] = remove_single(df, 'tokenized_raw')
df['tokenized_mults_cnt'] = df['tokenized_mults'].apply(lambda x: len(x))
# Drop extra stops if available
try:
extra_stops = pd.read_csv(path_data / 'processed' / 'extra_stops.csv')
def __init__(self, vector_output=False):
self.vector_output = vector_output
self.nlp = hu_core_ud_lg.load()
