def get_and_save_ner(unprocessed_text: str, lang_code: str, doc_title: str):
if lang_code == 'en':
nlp = en_core_web_sm.load()
else:
nlp = xx_ent_wiki_sm.load()
doc = nlp(unprocessed_text)
entity_label_pairs = [(x.text, x.label_) for x in doc.ents]
label_counts = Counter([x.label_ for x in doc.ents])
html_title = re.sub(r'[\W\s]', '_', doc_title)
if html_title[-1] == '_':
html_title = html_title[:-1]
html_title = f'static/labelled_texts/html_{html_title}.html'
# displacy.serve(doc, style='ent')
ner_text = displacy.render(doc, style='ent')
with open(html_title, 'w', encoding='utf-8') as html_file:
html_file.write(ner_text)
return entity_label_pairs, label_counts, ner_text, html_title
def run_english(sent):
import spacy
import xx_ent_wiki_sm
nlp = xx_ent_wiki_sm.load()
sentence = nlp (sent)
l = [(X.text, X.label_) for X in sentence.ents]
# print(l)
to_return = []
# print(l)
for item in l:
if len(item[0].split(' ')) == 1:
if item[1] == 'PER':
new_tuple = (item[0], 'PERSON')
elif item[1] == 'ORG':
new_tuple = (item[0], 'ORGANIZATION')
elif item[1] == 'GPE':
new_tuple = (item[0], 'GPE')
else:
new_tuple = (item[0], 'MISC')
to_return.append(new_tuple)
else:
for tab in item[0].split(' '):
if item[1] == 'PER':
new_tuple = (tab, 'PERSON')
to_return.append(new_tuple)
elif item[1] == 'ORG':
new_tuple = (tab, 'ORGANIZATION')
to_return.append(new_tuple)
elif item[1] == 'GPE':
new_tuple = (tab, 'GPE')
to_return.append(new_tuple)
else:
new_tuple = (tab, 'MISC')
to_return.append(new_tuple)
return to_return
#print(run_english("Alex prezinta licenta."))
import json
import re
import pandas as pd
import numpy as np
from post import Post
from polyglot.detect import Detector
import spacy
import xx_ent_wiki_sm
import multiprocessing as mp
from pathlib import Path
spacy.require_gpu()
nlp = xx_ent_wiki_sm.load()
false_langs = {"kn", "un", "or", "chr", "xx"}
translation_words = np.loadtxt("translation_prob.csv", usecols=0, dtype="str")
translation_words = set(translation_words)
def code_switch_polyglot(country, translation=True):
"""
Function to find code switch posts given a country
:param translation: bool
if we should remove translation posts
:param country: str country to find code switching posts in
:return: array
array of Post objects
"""
global valid_countries
comments = []
input_folder = Path("/ais/hal9000/masih/codeswitch/allposts/")
import spacy
import pandas as pd
import xx_ent_wiki_sm #Multi-language
# import en_core_web_sm
en_model = xx_ent_wiki_sm.load()
# en_core_web_sm.load()
import csv
import json
df_2 = pd.DataFrame([])
sample_text = "WTF, die grüne Grenze zwischen Österreich und Deutschland ist 800 km lang. Kein Problem für #Flüchtlinge. #Grenzkontrollen ist keine Lösung!"
nlp = spacy.load('xx_ent_wiki_sm') #('en_core_web_sm')
# df= pd.read_csv(r"C:\Users\insan\OneDrive\Desktop\task_initial\Tweet_data\raw_tweet.csv")
# for index, row in df.iterrows():
# #print(row['Tweet'])
# #sample_text+=(row['Tweet'])
def ner(text):
ent_docs = en_model(text)
return [(ent.text, ent.label_) for ent in ent_docs.ents]
print(ner(sample_text))
text_list, ne_list = map(list, zip(*ner(sample_text)))
d = {'Tweet': text_list, 'Named Entity': ne_list}
df_2 = pd.DataFrame(data=d, columns=['Tweet', 'Named Entity'])
print(df_2)
doc = nlp(sample_text)
import re
import time
from urllib.request import urlopen, Request
import en_core_web_sm # download via python -m spacy download en_core_web_sm
import nltk
import xx_ent_wiki_sm # download via python -m spacy download xx_ent_wiki_sm
nlpmultilang = xx_ent_wiki_sm.load()
nlpeng = en_core_web_sm.load()
nltk.download('punkt', quiet=True)
nltk.download('averaged_perceptron_tagger', quiet=True)
TIMEOUT_LIMIT = 32
def tokenize_title(title):
"""split title into initial tokens"""
tokens = nltk.word_tokenize(title)
tagged = nltk.pos_tag(tokens)
# remove tokens that are tagged with CC (coordinating conjunction), DT (Determiner), EX (existential there),
# IN (preposition of subordinating conjunction) PDT, PRP, PRP$, RP, TO, UH, WP, WP$, WRB, WDT
tagged = list(
filter(
lambda x: x[1] != 'CC' and x[1] != 'DT' and x[1] != 'EX' and x[1]
!= 'IN' and x[1] != 'PDT' and x[1] != 'PRP' and x[1] != 'PRP$' and
x[1] != 'RP' and x[1] != 'TO' and x[1] != 'UH' and x[1] != 'WP' and
x[1] != 'WP$' and x[1] != 'WRB' and x[1] != 'WDT', tagged))
# TODO: potentially develop this to gather by clauses
