def load_final_pipe():
import string
import en_core_sci_lg
import en_core_web_md
from collections import Counter
from tqdm import tqdm
en_core_sci_lg.load()
en_core_web_md.load()
return load_object(FINAL_PIPE_FILE_PATH)
def __init__(self, model_size='en_core_web_md'):
try:
self.nlp = spacy.load(model_size)
except:
self.nlp = en_core_web_md.load()
finally:
self.nlp.tokenizer = self.custom_tokenizer(self.nlp)
def entity_analyzer(my_text):
nlp = en_core_web_md.load()
docx = nlp(my_text)
tokens = [token.text for token in docx]
entities = [(entity.text, entity.label_) for entity in docx.ents]
allData = ['"Token":{},\n"Entities":{}'.format(tokens, entities)]
return allData
def text_analyzer(my_text):
nlp = en_core_web_md.load()
docx = nlp(my_text)
allData = [('"Token":{},\n"Lemma":{}'.format(token.text, token.lemma_))
for token in docx]
return allData
def do_eval(test_data_path, shuffle=False):
if FLAGS.load_model is None:
raise ValueError("You need to specify the model location by --load_model=[location]")
# Load Testing Data
question_1, question_2, labels = get_input_from_csv(test_data_path)
if shuffle:
question_1, question_2, labels = shuffle_data(question_1, question_2, labels)
# Load Pre-trained Model
if FLAGS.best_glove:
import en_core_web_md
nlp = en_core_web_md.load() # load best-matching version for Glove
else:
nlp = spacy.load('en')
embedding_matrix = load_glove_embeddings(nlp.vocab, n_unknown=FLAGS.num_unknown) # shape=(1071074, 300)
tf.logging.info('Build model ...')
esim = ESIM(embedding_matrix, FLAGS.max_length, FLAGS.num_hidden, FLAGS.num_classes, FLAGS.keep_prob, FLAGS.learning_rate)
if FLAGS.load_model:
model = esim.build_model(FLAGS.load_model)
else:
raise ValueError("You need to specify the model location by --load_model=[location]")
# Convert the "raw data" to word-ids format && convert "labels" to one-hot vectors
q1_test, q2_test = convert_questions_to_word_ids(question_1, question_2, nlp, max_length=FLAGS.max_length, tree_truncate=FLAGS.tree_truncate)
labels = to_categorical(np.asarray(labels, dtype='int32'))
scores = model.evaluate([q1_test, q2_test], labels, batch_size=FLAGS.batch_size, verbose=1)
print("=================== RESULTS =====================")
print("[*] LOSS OF TEST DATA: %.4f" % scores[0])
print("[*] ACCURACY OF TEST DATA: %.4f" % scores[1])
def test_predict(self):
nlp = en_core_web_md.load()
classifier = TextClassifier(nlp.vocab, n_classes=2)
text = "Sample text to test the classifier"
output = classifier.predict(text)
self.assertTrue(isinstance(output.data, torch.Tensor))
def test_get_ids(self):
nlp = en_core_web_md.load()
classifier = TextClassifier(nlp.vocab, n_classes=2)
text = "Sample text to test the classifier"
ids = classifier.get_ids(text)
self.assertTrue(isinstance(ids, list))
def get_spacy_model():
global _spacy_cache
if _spacy_cache is None:
print("Initializing spaCy model...")
_spacy_cache = en_core_web_md.load()
return _spacy_cache
def __init__(self,
dictionary_loader: DictionaryLoader,
language="en",
filters: List[AnnotationFilter] = None):
"""
Parameters
----------
dictionary_loader: DictionaryLoader
The dictionary loader that will provide the dictionary contents
language: str
The language of the text that will processed (affects the choice of tokenner and stemmer).
filters: List[AnnotationFilter]
A list of filters to apply post recognition
"""
super().__init__(dictionary_loader, language=language, filters=filters)
self.punctuation_remove = regex.compile(
r'[\p{C}|\p{M}|\p{P}|\p{S}|\p{Z}]+', regex.UNICODE)
self.label_concept_index = {}
self.label_token_counts = {}
self.label_lengths = {}
self.trie = Trie()
if language == 'en':
import en_core_web_md
self.spacy = en_core_web_md.load()
elif language == 'fr':
import fr_core_web_md
self.spacy = fr_core_web_md.load()
else:
raise ValueError(f"Unsupported language: {language}")
def train(train_data, val_data, batch_size, n_epochs, save_dir=None):
# Stage 1: Read training data (csv) && Preprocessing them
tf.logging.info('Loading training and validataion data ...')
train_question_1, train_question_2, train_labels = get_input_from_csv(train_data)
# val_question_1, val_question_2, val_labels = get_input_from_csv(val_data)
# Stage 2: Load Pre-trained embedding matrix (Using GLOVE here)
tf.logging.info('Loading pre-trained embedding matrix ...')
if FLAGS.best_glove:
import en_core_web_md
nlp = en_core_web_md.load() # load best-matching version for Glove
else:
nlp = spacy.load('en')
embedding_matrix = load_glove_embeddings(nlp.vocab, n_unknown=FLAGS.num_unknown) # shape=(1071074, 300)
# Stage 3: Build Model
tf.logging.info('Build model ...')
esim = ESIM(embedding_matrix, FLAGS.max_length, FLAGS.num_hidden, FLAGS.num_classes, FLAGS.keep_prob, FLAGS.learning_rate)
if FLAGS.load_model:
model = esim.build_model(FLAGS.load_model)
else:
model = esim.build_model()
# Stage 4: Convert the "raw data" to word-ids format && convert "labels" to one-hot vectors
tf.logging.info('Converting questions into ids ...')
q1_train, q2_train = convert_questions_to_word_ids(train_question_1, train_question_2, nlp, max_length=FLAGS.max_length, tree_truncate=FLAGS.tree_truncate)
train_labels = to_categorical(np.asarray(train_labels, dtype='int32'))
# q1_val, q2_val = convert_questions_to_word_ids(val_question_1, val_question_2, nlp, max_length=FLAGS.max_length, tree_truncate=FLAGS.tree_truncate)
# val_labels = to_categorical(np.asarray(val_labels, dtype='int32'))
# Stage 5: Training
tf.logging.info('Start training ...')
callbacks = []
save_dir = save_dir if save_dir is not None else 'checkpoints'
filepath = os.path.join(save_dir, "weights-{epoch:02d}-{val_acc:.2f}.hdf5")
checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')
callbacks.append(checkpoint)
if FLAGS.tensorboard:
graph_dir = os.path.join('.', 'GRAPHs')
if not os.path.exists(graph_dir):
os.makedirs(graph_dir)
tb = TensorBoard(log_dir=graph_dir, histogram_freq=0, write_graph=True, write_images=True)
callbacks.append(tb)
model.fit(
x=[q1_train, q2_train],
y=train_labels,
batch_size=batch_size,
epochs=n_epochs,
# validation_data=([q1_val, q2_val], val_labels),
validation_split=0.2,
callbacks=callbacks,
shuffle=True,
verbose=FLAGS.verbose
)
def __init__(self):
ifninstall(self.__class__.md_pkg)
print('initiating NLP language pipeline...', end='', flush=True)
import en_core_web_md
self._nlp = en_core_web_md.load()
print('done')
self._corpus = None
def _get_spacy_wmd_model():
global _spacy_wmd_cache
if _spacy_wmd_cache is None:
_spacy_wmd_cache = en_core_web_md.load()
_spacy_wmd_cache.add_pipe(
wmd.WMD.SpacySimilarityHook(_spacy_wmd_cache), last=True)
return _spacy_wmd_cache
def __init__(self):
# init model and disabling unuse module(s)
try:
self.model = en_core_web_md.load()
print("Set SpaCy model to medium model")
except:
self.model = en_core_web_sm.load()
print("Set SpaCy model to small model")
def __init__(self, lang):
if lang == LANG.EN:
self.nlp = en_core_web_md.load()
else:
self.nlp = de_core_news_sm.load()
self.stanford_ner = StanfordNERTagger(model,
'../models/stanford-ner.jar',
encoding='utf-8')
def embbeding(self):
pretrained_emb = []
spacy_tool = en_core_web_md.load()
pretrained_emb.append(spacy_tool('PAD').vector)
pretrained_emb.append(spacy_tool('UNK').vector)
pretrained_emb.append(spacy_tool('CLS').vector)
for i in range(len(self.dictionary.idx2word)):
word = self.dictionary.idx2word[i]
pretrained_emb.append(spacy_tool(word).vector)
return np.array(pretrained_emb)
def create_corpus(text_list, target_path):
nlp = en_core_web_md.load(disable=['tagger', 'ner', 'textcat'])
with open(target_path, 'w') as f:
# Split sentences for each document
for text in tqdm(text_list):
text_lines = [str(sent)
for i, sent in enumerate(nlp(str(text)).sents)]
f.write('\n'.join(text_lines))
f.write("\n \n")
def en_lang(cls):
me_list = ['i', 'my']
embeddings_model = FlairEmbeddingModels().en_lang()
nlp = en_core_web_md.load()
relationship_list = [
'father', 'mother', 'sister', 'brother', 'son', 'daughter',
'husband', 'wife', 'grandson', 'granddaughter', 'grandmother',
'grandfather', 'uncle', 'aunt', 'friend'
]
return cls(me_list, embeddings_model, nlp, relationship_list)
def __init__(self, n_classes, arch="cnn", vocab=None, hparams={}):
"""
Constructor class for TextClassifier
Args:
vocab (spacy.vocab): Spacy vocabulary class
arch (string/pl.Lightningmodule): The underlying text classifier model architecture
hparams (dict): Dictionary of hyper parameters for the underlying model
Returns:
None
"""
if isinstance(vocab, Vocab):
self._vocab = vocab
else:
self._vocab = en_core_web_md.load().vocab
self._vocab.set_vector("@pad@",
vector=np.zeros(self.vocab.vectors.shape[1]))
self._vocab.set_vector("@oov@",
vector=np.zeros(self.vocab.vectors.shape[1]))
oov_orth = self._vocab["@oov@"].orth
self.oov_id = self._vocab.vectors.key2row[oov_orth]
self.tokenizer = Tokenizer(self.vocab)
input_dim, embedding_dim = self.vocab.vectors.shape
output_dim = n_classes
pad_idx = self.vocab.vectors.key2row[self.vocab["@pad@"].orth]
hparams["pad_idx"] = pad_idx
hparams["input_dim"] = input_dim
hparams["embedding_dim"] = embedding_dim
if isinstance(arch, BaseModel):
self._model = arch
elif isinstance(arch, str):
if arch == "cnn":
self._model = CNN2D(output_dim, hparams)
elif arch == "bilstm":
self._model = BiLSTM(output_dim, hparams)
else:
print("No such architecture exists")
else:
print("arch should be string or a torch file duh")
def spacy_keyList(q):
nlp = en_core_web_md.load()
doc = nlp(q)
# for w in doc:
# print(f'{w.text:15s}[{w.tag_:5s}|{w.lemma_:20s}|{w.pos_:6s}]')
# for ent in doc.ents:
# print(f'{ent.text:15s}[{ent.label_}]')
pattern1 = [{'ENT_TYPE': 'ORG', "OP": "+"}]
pattern2 = [{'ENT_TYPE': 'PERSON', "OP": "+"}]
pattern3 = [{'ENT_TYPE': 'FAC', "OP": "+"}]
pattern4 = [{'ENT_TYPE': 'LOC', "OP": "+"}]
pattern5 = [{'ENT_TYPE': 'GPE', "OP": "+"}]
pattern6 = [{
'ENT_TYPE': 'CARDINAL',
"OP": "*"
}, {
'ENT_TYPE': 'LOC',
"OP": "+"
}, {
"IS_PUNCT": True
}, {
'ENT_TYPE': 'GPE',
"OP": "+"
}, {
"IS_PUNCT": True
}, {
'ENT_TYPE': 'GPE',
"OP": "+"
}]
pattern7 = [{
'ENT_TYPE': 'FAC',
"OP": "+"
}, {
"IS_LOWER": True
}, {
'ENT_TYPE': 'CARDINAL',
"OP": "?"
}, {
'ENT_TYPE': 'GPE',
"OP": "+"
}]
matcher = Matcher(nlp.vocab)
matcher.add("name1", None, pattern1, pattern2, pattern3, pattern4,
pattern5, pattern6, pattern7)
matches = matcher(doc)
keyList = []
for m_id, s, e in matches:
keyword = doc[s:e].text
if len(keyword) > 1 and len(doc[s:e].text.split()) > 1:
keyList.append(keyword)
return keyList
def __init__(self, title, path, verbose = False, testing = 0): self.verbose = verbose self.testing = testing self.home = path self.title = title self.date = self.get_date() self.today = date.today() self.pdf_path = os.path.join(self.home, self.title) self.img_path = os.path.join(self.pdf_path + 'convertedimages') self.words = [] self.num_pages = self.get_pdf_images() self.text = self.get_text() self.nlp = en_core_web_md.load() self.doc = self.nlp(self.text)
def get_entities(self, document: str, language: str = 'english'):
''' Takes a document and returns a list of extracted entities '''
if language == 'spanish':
try:
import es_core_news_md
logger.info("Success importing en_core_web_md")
except ImportError:
logger.error("Error importing es_core_news_md")
sys.exit(-1)
else:
try:
import en_core_web_md
logger.info("Success importing en_core_web_md")
except ImportError:
logger.error("Error importing en_core_web_md")
sys.exit(-1)
if isinstance(document, List):
document = " ".join(document)
# if language given is not the name of a spacy parser, try to convert it to one
parser_name = language if language in LANG_TO_PARSER.values(
) else LANG_TO_PARSER.get(language.lower())
if not parser_name:
raise Exception('language not supported')
# if requested parser is not already in memory, try to load from spacy
if parser_name not in PARSERS:
try:
logger.info("Trying to load parser.")
if language == 'spanish':
PARSERS[parser_name] = es_core_news_md.load()
else:
PARSERS[parser_name] = en_core_web_md.load()
#PARSERS[parser_name] = spacy.load(parser_name)
except Exception:
logger.exception("Error loading parser")
sys.exit(-1)
else:
logger.info("Found parser %s in memory" % parser_name)
def get_ents(doc):
''' prep, parse, then extract entities from doc text '''
doc = prep_text(doc) # preprocess string
doc = PARSERS[parser_name](doc) # parse prepped doc
ents = set(ent.text for ent in doc.ents
if not self.filter_entity(ent)) # extract entities
return list(ents)
return get_ents(document)
def __init__(self, name="", motorcycles=[], closest_track=None, **kwargs):
print("Loading in NLP models... (this could take a minute or two)")
self.nlp = en_core_web_md.load()
self.sbert_model = SentenceTransformer('stsb-roberta-base')
# For NER (at least for detecting a person's name) it seems to perform a lot better with small, it wouldn't detect my name on medium
self.ner = en_core_web_sm.load()
# user data
self.name = name
self.motorcycles = motorcycles
self.closest_track = closest_track
self.conn = sqlite3.connect("knowledge_base.db").cursor()
self.motorcycle_finder = MotorcycleFinder(ner=self.ner)
self.new_user = False if name else True # will eventually be something like True if name else False
def tokenize_string_details(s):
global nlp
if nlp is None:
# not doing it here causes some errors about en_models.vector not being able to be loaded
# essentially this causes the models to be loaded at most once in the executors
# not very good, but it will do for this sie
nlp = en_core_web_md.load()
doc = nlp(s)
tokens = []
for token in doc:
if token.ent_type_ in ['GPE', 'LOC', 'DATE', 'TIME']:
tokens.append(f'-{token.ent_type}-')
elif token.like_url:
tokens.append(f'-URL-')
else:
tokens.append(token.lower_)
return tokens
def main():
logging.getLogger().setLevel('INFO')
nlp = en_core_web_md.load()
split = 'val'
sent = sentences(json.load(open(coco_path + '/dataset.json')), split=split)
writer = csv.writer(open(data_path + '/depparse_coco_val.csv', "w"))
writer.writerow(
["sentid", "position", "word", "postag", "postag_c", "dep", "head"])
for sent_i in sent:
#logging.info("Parsing: {} {}".format(sent_i['sentid'], ' '.join(sent_i['tokens'])))
logging.info("Parsing: {} {}".format(sent_i['sentid'], sent_i['raw']))
postag, postag_c, label, head = parse(nlp, sent_i['tokens'])
for i in range(len(sent_i['tokens'])):
writer.writerow([
sent_i['sentid'], i, sent_i['tokens'][i], postag[i],
postag_c[i], label[i], head[i]
])
def get_spacy_tokenizer(default_lingo, supported_languages, bigmodel_required):
'''returns the spacy nlp function corresponding to the language of a document'''
if default_lingo in supported_languages:
if bigmodel_required == False:
if default_lingo == "German":
import de_core_news_sm
nlp = de_core_news_sm.load()
elif default_lingo == "English":
import en_core_web_sm
nlp = en_core_web_sm.load()
elif default_lingo == "Spanish":
import es_core_news_sm
nlp = es_core_news_sm.load()
elif default_lingo == "French":
import fr_core_news_sm
nlp = fr_core_news_sm.load()
elif default_lingo == "Portuguese":
import pt_core_news_sm
nlp = pt_core_news_sm.load()
else:
import it_core_news_sm
nlp = it_core_news_sm.load()
else:
if default_lingo == "German":
import de_core_news_md
nlp = de_core_news_md.load()
elif default_lingo == "English":
import en_core_web_md
nlp = en_core_web_md.load()
elif default_lingo == "Spanish":
import es_core_news_md
nlp = es_core_news_md.load()
elif default_lingo == "French":
import fr_core_news_md
nlp = fr_core_news_md.load()
elif default_lingo == "Portuguese":
# there is no pt_md model
import pt_core_news_sm
nlp = pt_core_news_sm.load()
else:
# there is no it_md model
import it_core_news_sm
nlp = it_core_news_sm.load()
else:
print("NOT A SUPPORTED LANGUAGE!")
return nlp
def en_lang(cls):
nlp = en_core_web_md.load()
embeddings_model = FlairEmbeddingModels().en_lang()
# PP: e.g. 'I have a son', 'I have a smaller brother', 'I have a 9 year old son'
# NP: e.g. 'My (little) sister'
grammar = r"""
PP: {<PRON><VERB><NUM>?<DET>?<ADJ>?<NOUN>}
NP: {<ADJ><ADJ>?<NOUN>}
REL: {<PP>|<NP>}"""
relationship_list = [
'father', 'mother', 'sister', 'brother', 'son', 'daughter',
'husband', 'wife', 'grandson', 'granddaughter', 'grandmother',
'grandfather', 'uncle', 'aunt', 'friend'
]
me_list = ['i', 'my', 'me']
return cls(nlp, grammar, relationship_list, me_list, embeddings_model)
def process_page(row):
nlp = en_core_web_md.load()
warc_record = WarcRecord(row)
warc_payload = warc_record.payload
if warc_record.broken or not warc_payload or warc_payload is None or warc_record.id is None:
return (yield "")
ents = []
soup = None
try:
soup = BeautifulSoup(warc_payload, "html.parser")
except:
raise Exception("SOUP FAILED: %s %s\n" % (warc_payload, warc_record.id) )
for script in soup(['style', 'script', 'head', 'title', 'meta', '[document]', 'code' 'blockquote', 'cite']):
script.extract()
text = " ".join(re.findall(r'\w+', soup.get_text()))
article = nlp(text)
raw_words = []
for ent in article.ents:
if(ent.label_ not in ["CARDINAL", "DATE", "QUANTITY", "TIME", "ORDINAL", "MONEY", "PERCENT", "QUANTITY"]) and len(ent.text.split(" ")) < 4:
raw_words.append(ent.text)
ents.append(ent)
processed_results = {}
output = []
for ent in ents:
stripped_ent = ent.text.strip()
if text == '':
continue
es_results = es.search(stripped_ent)
sims = []
abstracts = []
processed_results = {}
for es_result in es_results:
if es_result.label in processed_results:
continue
else:
processed_results[es_result.label] = True
abstract = sparql.search(es_result.id, es_result.label, es_result.score)
if abstract:
abstracts.append([es_result.id, abstract])
if len(abstracts) > 0:
output.append(stringify_reply(warc_record.id, stripped_ent, get_sim(text, abstracts)))
yield output
def load_model(model_str):
"""
Loads a model given an input string (could work differently)
"""
if model_str == "en_core_web_md":
import en_core_web_md
model_func = en_core_web_md.load()
if model_str == "en_resume_model":
import en_resume_model
model_func = en_resume_model.load()
if model_str == "seed_data":
import spacy
from spacy.matcher import PhraseMatcher
from spacy.tokens import Span
class EntityMatcher(object):
"""
Matcher with multi-case
"""
def __init__(self, nlp, terms, label):
patterns = [nlp(text.lower()) for text in terms]
patterns += [nlp(text.upper()) for text in terms]
patterns += [nlp(text.title()) for text in terms]
self.matcher = PhraseMatcher(nlp.vocab)
self.matcher.add(label, None, *patterns)
self.name = '{}_matcher'.format(label)
def __call__(self, doc):
matches = self.matcher(doc)
for match_id, start, end in matches:
span = Span(doc, start, end, label=match_id)
doc.ents = list(doc.ents) + [span]
return doc
nlp = spacy.load('en_core_web_md', disable=['ner'])
with open("../data/seeds.json", 'r') as f:
seed_data_json = json.loads(f.read())
for key in seed_data_json.keys():
terms = seed_data_json[key]
#terms = spacy_similarity(terms, nlp, num_similar=2)
entity_matcher = EntityMatcher(nlp, terms, key)
nlp.add_pipe(entity_matcher)
model_func = nlp
return model_func
def eventLister(text, subject):
# Load the large English NLP model
nlp = en_core_web_md.load()
# Parse the document with spaCy
doc = nlp(text)
# Extract semi-structured statements
statements = textacy.extract.semistructured_statements(doc, subject)
# Return the results
str = "Here are the things I know about " + subject + ":"
for statement in statements:
subject, verb, fact = statement
str += f"\n - {fact}"
return str
def get_people_orgs_from_text(text, model="large"):
"""
Loads and runs spaCy NLP object on the text provided.
returns two lists:
(1) tuples of people (person, number of appearances)
(2) tuples of organizations (organization, number of appearances)
"""
# Step 1: Check model input and load nlp object
if model.lower() == "large": nlp = en_core_web_lg.load()
elif model.lower() == "medium": nlp = en_core_web_md.load()
elif model.lower() == "small": nlp = en_core_web_sm.load()
else:
print("Invalid NLP model.")
return [], []
# Step 2: Process the text using spaCy and get list of people and
# orgnization objects
doc = nlp(text)
orgs = [x.text for x in doc.ents if x.label_ == 'ORG']
people = [x.text for x in doc.ents if x.label_ == 'PERSON']
#pprint([(x.text, x.label_) for x in doc.ents])
# Step 3: Process and filter the list of organizations
orgdict = dict()
for item in orgs:
if item in orgdict.keys(): orgdict[item] += 1
else: orgdict[item] = 1
orgs = list(orgdict.items())
orgs.sort(key=lambda x: x[1], reverse=True)
orgs = filter_people_orgs(orgs)
# Step 4: Process and filter the list of people
peopledict = dict()
for item in people:
if item in peopledict.keys(): peopledict[item] += 1
else: peopledict[item] = 1
people = list(peopledict.items())
people.sort(key=lambda x: x[1], reverse=True)
people = filter_people_orgs(people)
return people, orgs
