def main():
st.title("Tweet Screener")
st.subheader("*Guaranteeing 2020 proof tweets to the masses*")
st.sidebar.image("image_resources/DSI-logo.jpg", use_column_width = True)
st.sidebar.subheader("Africa DSI NLP Project by Team 2")
st.sidebar.write("Catherine, Fanamby, Malcolm, and Martin")
section = st.sidebar.radio('Sections to Visit',('Swear Word Analyser', 'Sentiment Analyser', 'Topic Identifier'))
publish = st.sidebar.button(label = "Publish Tweet!")
if section == 'Swear Word Analyser':
st.subheader('Swear Word Analyser')
blacklist = load_screener()
sentence = st.text_area('Input your message/tweet here:')
if sentence:
# Pre-process tweet
answer = PS.profanityscreen(sentence, blacklist, True)
st.subheader("Swear Analysis Results:")
# Show predictions
st.write('Swear Words Found:')
st.dataframe(pd.DataFrame(answer[1], columns=["Swear Words"]))
st.write('Your Censored Tweet:')
st.write(answer[0])
if section == "Sentiment Analyser":
st.subheader('Sentiment Analyser')
sentSentence = st.text_area('Input your message/tweet here:')
if sentSentence:
sentimentTweet = Sentence(preprocess(sentSentence))
emoteTweet = Sentence(preprocess(sentSentence))
#Sentiment Dictionaries
sentiment_dict = {'0': 'Negative', '4': 'Positive'}
emote_dict = {'0': 'Anger', '1': 'Fear', '2': 'Joy', '3': 'Love', '4': 'Sadness', '5': 'Surprise'}
emoji_dict = {'0': ':rage:', '1': ':fearful:', '2': ':joy:', '3': ':heart_eyes:', '4': ':cry:', '5': ':astonished:'}
basic_emo_dict = {"0": ':rage:', "4": ":smile:"}
with st.spinner("Weeeeeee......."):
SentClassifier = TextClassifier.load('twitter_sentiment/model-saves/final-model.pt')
EmoteClassifier = TextClassifier.load('twitter_sentiment/model-saves/emotion-model.pt')
SentClassifier.predict(sentimentTweet)
EmoteClassifier.predict(emoteTweet)
st.subheader("Sentiment Analysis Results:")
predSent = sentimentTweet.labels[0]
predSText = sentiment_dict[predSent.value[0]]
st.markdown('Your sentence is ' + str(predSText) + " " + basic_emo_dict[predSent.value[0]] + ' with '+ "{:.2f}".format(predSent.score*100)+ '% confidence')
predEmote = emoteTweet.labels[0]
predEText = emote_dict[predEmote.value[0]]
st.markdown('Your sentence is predicted to portray ' + predEText + " " +emoji_dict[predEmote.value[0]] +' with '+ "{:.2f}".format(predEmote.score*100)+ ' % confidence')
if section == "Topic Identifier":
st.subheader("Topic Identifier")
topicSentence = st.text_area('Input your message/tweet here:')
st.subheader("Sensitivity Analysis Results:")
if topicSentence:
topicTweet = preprocess_test(topicSentence)
topic_dict = {0: "obscenity", 1: "violence", 2: "verbal abuse", 3: "identity hate speech", 4: "hate speech", 5: "offense", 6: "neither"}
policies_dict = {0 : "https://help.twitter.com/en/safety-and-security/offensive-tweets-and-content",
1 : "https://help.twitter.com/en/rules-and-policies/violent-threats-glorification",
2 : "https://help.twitter.com/en/rules-and-policies/abusive-behavior",
3 : "https://help.twitter.com/en/rules-and-policies/hateful-conduct-policy",
4 : "https://help.twitter.com/en/rules-and-policies/hateful-conduct-policy",
5 : "https://help.twitter.com/en/safety-and-security/offensive-tweets-and-content"}
twitter_rules = "https://help.twitter.com/en/rules-and-policies#general-policies"
with st.spinner("Predicting..."):
TopicClassifier = tf.keras.models.load_model('Topic Identifier/model_saves/topic_identifier_model.h5')
topic_pred = TopicClassifier.predict(topicTweet)
topTopic = topic_pred.argmax(1)[0]
topTopicText = topic_dict[topic_pred.argmax(1)[0]]
graph_pred = pd.DataFrame(topic_pred, columns = ["obscenity", "violence", "verbal abuse", "identity hate crime", "hate crime", "offense", "neither"])
columns = ["obscenity", "violence", "verbal abuse", "identity hate speech", "hate speech", "offense", "neither"]
if topic_pred.argmax(1)[0]!=6 :
st.write("Your tweet may contain sentences that promote " + topTopicText+ " with "+str(topic_pred[0][topTopic]*100) +" % confidence")
st.write("Please review Twitter Rules and policies: "+ twitter_rules)
st.write("And Twiiter's "+ topTopicText + " policy: "+ policies_dict[topic_pred.argmax(1)[0]])
else:
st.write("Your tweet is fine in terms of policy.")
plt.bar(height = topic_pred.flatten(),x=columns, width = 1)
plt.title('Policy Breaking Likelihood')
plt.xticks(rotation=45)
plt.xlabel('Twitter Policies (Topics)')
plt.ylabel('Probability of Violation')
st.pyplot()
st.write(topic_pred)
if publish:
if sentence:
sentiment = "None"
sentence = sentence
publish_tweet(sentiment, sentence)
elif sentSentence:
sentiment = predEText
sentence = sentSentence
publish_tweet(sentiment, sentence)
elif topicSentence:
sentiment = predEText
sentence = topicSentence
publish_tweet(sentiment, sentence)
else:
st.sidebar.write("You haven't written or analysed your tweet yet.")
st.sidebar.markdown("This application helps determine how problematic your tweet is before publishing it."
+ " We utilise three main tools to achieve this."
+" A swear word analyser that checks your tweet for profanity and delivers a censored tweet."
+" A sentiment analyser that predicts the emotion in your tweet, to check if you were really being positive."
+" Finally a topic identifier which determines if you broke one of Twitter's policies with out knowing it!"
+" Once you have thouroughly scrubbed you tweet you may store your results for further analyses.")
from flair.embeddings import WordEmbeddings, DocumentPoolEmbeddings, Sentence
from flair.data_fetcher import NLPTaskDataFetcher
from flair.embeddings import BertEmbeddings
from flair.models import TextClassifier
from flair.trainers import ModelTrainer
from pathlib import Path
glove_embedding = WordEmbeddings('glove')
bert_embedding = BertEmbeddings('bert-base-uncased')
corpus = NLPTaskDataFetcher.load_classification_corpus(Path('./'),
test_file='test.csv',
dev_file='dev.csv',
train_file='train.csv')
document_embeddings = DocumentPoolEmbeddings([bert_embedding, glove_embedding])
classifier = TextClassifier(document_embeddings,
label_dictionary=corpus.make_label_dictionary(),
multi_label=True)
trainer = ModelTrainer(classifier, corpus)
trainer.train('./', max_epochs=10)
)
stats = corpus.obtain_statistics()
print(stats)
# create the label dictionary
label_dict = corpus.make_label_dictionary()
print(label_dict)
# make a list of word embeddings
embeddings = [OneHotEmbeddings(corpus=corpus)]
# initialize document embedding by passing list of word embeddings
# Can choose between many RNN types (GRU by default, to change use rnn_type parameter)
document_embeddings = DocumentRNNEmbeddings(embeddings,
bidirectional=True,
hidden_size=256)
# create the text classifier
classifier = TextClassifier(document_embeddings, label_dictionary=label_dict)
# initialize the text classifier trainer
trainer = ModelTrainer(classifier, corpus)
# start the training
trainer.train('resources/',
learning_rate=0.1,
mini_batch_size=128,
anneal_factor=0.5,
patience=5,
max_epochs=20)
#from bson.objectid import ObjectId
from flask_cors import CORS
from flair.models import TextClassifier
from flair.data import Sentence
from flask import session
app = Flask(__name__)
app.secret_key = "super_secret_key"
# # app.config['SECRET_KEY'] = 'oh_so_secret'
# app.config['MONGO_DBNAME'] = 'exposeModel'
# app.config['MONGO_URI'] = 'mongodb://*****:*****@app.route('/', methods=['GET'])
def index():
return jsonify("welcome to Arafa API")
@app.route('/api/tasks', methods=['GET'])
def get_result():
result = []
try:
data_result = session['my_result']
result.append({
'title': data_result['title'],
'tag': data_result['tag']
def upload_file():
uploaded_file = request.files['file']
if uploaded_file.filename != '':
## uploaded_file jako parametr pandas
## cala logika tworzenia positive i negative
df = load_dataset_from(uploaded_file)
# To verify if data loaded correctly:
# print(df.head(10))
print(df.head(10))
classifier = TextClassifier.load('sentiment')
tmp_negatives = {}
tmp_positives = {}
print(f'Number of players: {len(df["Player"].unique())}')
for player_name in df["Player"].unique():
tmp_negatives[player_name] = list()
tmp_positives[player_name] = list()
for dp in df.values:
l = dp.tolist()
sentence = Sentence(l[-1])
classifier.predict(sentence)
if sentence.labels[0].value == "NEGATIVE":
tmp_negatives[l[-2]].append(sentence.labels[0].score)
elif sentence.labels[0].value == "POSITIVE":
tmp_positives[l[-2]].append(sentence.labels[0].score)
negative = {}
positive = {}
for player in tmp_negatives.keys():
if len(tmp_negatives[player]) > 10:
negative[player] = sum(tmp_negatives[player]) / len(
tmp_negatives[player])
for player in tmp_positives.keys():
if len(tmp_positives[player]) > 10:
positive[player] = sum(tmp_positives[player]) / len(
tmp_positives[player])
print(negative)
print(positive)
# for player in tmp.keys():
# tmp[player] = Sentence(tmp[player])
# classifier.predict(tmp[player])
# if tmp[player].labels[0].value == "NEGATIVE":
# negative[player] = tmp[player].labels[0].score
# elif tmp[player].labels[0].value == "POSITIVE":
# positive[player] = tmp[player].labels[0].score
#
positive_json = json.dumps(positive, indent=4)
negative_json = json.dumps(negative, indent=4)
list = {}
list[0] = positive_json
list[1] = negative_json
return list
'Buenos dias, vamos a hacer algunos recados y a empezar el dia con energia!!', # 3
'@mireiaescribano justo cuando se terminan las fiestas de verano, me viene genial', # 3
'No sabes cuantas horas, minutos y segundos espero para volver a ver esa sonrisa que tanto me gusta ver salir de ti', # 0
'@cinthiaszc jajajaja me vas a decir a mi mi abuela cocina tan rico que mando al tacho la dieta :v', # 0
'te adoroVen a Perú pls' # 3
]
label_dictionary = ['0', '1', '2', '3']
fasttext_path = '../fasttext/models/{}_{}'.format('intertass', FT_MODEL_NAME)
fasttext_model = fasttext.load_model(path=fasttext_path)
dev_fasttext_probabilities, dev_fasttext_predictions = fasttext_embedding.predict_with_fasttext_model(
fasttext_model, tweets_to_test, label_dictionary)
print(dev_fasttext_probabilities)
print(dev_fasttext_predictions)
print("BERT MODEL")
bert_path = '{}/{}/{}/best-model.pt'.format(BERT_MODEL_PATH, BERT_MODEL_NAME, 'intertass')
bert_model = TextClassifier.load(bert_path)
dev_bert_probabilities, dev_bert_predictions = bert_embeddings.predict_with_bert_model(
bert_model, tweets_to_test, label_dictionary)
print(dev_bert_probabilities)
print(dev_bert_predictions)
# sentences = ['horrible']
#
# analyzer = vaderSentiment.vaderSentiment.SentimentIntensityAnalyzer()
# for sentence in sentences:
# vs = analyzer.polarity_scores(sentence)
# print("{:-<65} {}".format(sentence, str(vs)))
def train_model(self,
model_name="text_classification_model",
custom_word_embeddings=None,
rnn_type="GRU",
use_pool_embedding=False,
hidden_size=16,
reproject_words=True,
reproject_words_dimension=128,
learning_rate=1e-3,
batch_size=8,
anneal_factor=0.5,
patience=2,
max_epochs=30,
**kwargs):
"""
Train flair model and save it in your data folder
Parameters
----------
model_name: str
Name of your model
custom_word_embeddings: list<embedding>
Use custom flair embedding
See more in flair documentation: https://github.com/zalandoresearch/flair/tree/master/resources/docs
Return
-------
None
"""
self.model_name = model_name
corpus = CSVClassificationCorpus(self.data_folder,
self.column_name_map,
skip_header=True)
label_dict = corpus.make_label_dictionary()
# Word embedding selection
if custom_word_embeddings is None:
word_embeddings = [WordEmbeddings('fr')]
else:
word_embeddings = custom_word_embeddings
# initialize document embedding by passing list of word embeddings and parameters
if use_pool_embedding:
document_embeddings = DocumentPoolEmbeddings(
word_embeddings, pooling='max', fine_tune_mode='nonlinear')
else:
document_embeddings = DocumentRNNEmbeddings(
word_embeddings,
hidden_size=hidden_size,
reproject_words=reproject_words,
reproject_words_dimension=reproject_words_dimension,
rnn_type=rnn_type)
# create the text classifier and initialize trainer
classifier = TextClassifier(document_embeddings,
label_dictionary=label_dict)
trainer = ModelTrainer(classifier, corpus, optimizer=Adam)
# let's train !
num_workers = cpu_count()
trainer.train("{0}\\{1}".format(self.data_folder, self.model_name),
learning_rate=learning_rate,
num_workers=num_workers,
mini_batch_size=batch_size,
anneal_factor=anneal_factor,
patience=patience,
max_epochs=max_epochs,
**kwargs)
def train(self, X, y):
X_text = X[:, self.args.TEXT_COL]
y = y.flatten()
#corpus: TaggedCorpus = NLPTaskDataFetcher.load_corpus(NLPTask.CONLL_03)
train: List[Sentence] = []
for tweet, label in zip(X_text, y):
if tweet == '':
tweet = 'dummy'
s: Sentence = Sentence(tweet)
s.add_label(str(label))
train.append(s)
corpus: TaggedCorpus = TaggedCorpus(train, train, train)
# 2. create the label dictionary
label_dict = corpus.make_label_dictionary()
# 3. make a list of word embeddings
word_embeddings = [
glove_embeddings,
#twitter_embeddings,
# comment in this line to use character embeddings
#CharacterEmbeddings(),
# comment in flair embeddings for state-of-the-art results
# FlairEmbeddings('news-forward'),
fflair,
# FlairEmbeddings('news-backward'),
bflair
]
# 4. initialize document embedding by passing list of word embeddings
document_embeddings: DocumentLSTMEmbeddings = DocumentLSTMEmbeddings(
word_embeddings,
hidden_size=512,
reproject_words=True,
reproject_words_dimension=256,
)
# 5. create the text classifier
classifier = TextClassifier(document_embeddings,
label_dictionary=label_dict,
multi_label=False)
# 6. initialize the text classifier trainer
trainer = ModelTrainer(classifier, corpus)
self.model = trainer.model
self.model.save = self.save
self.model.save_checkpoint = self.save_checkpoint
# 7. start the training
trainer.train('../data/ecuador_earthquake_2016/models',
learning_rate=0.1,
mini_batch_size=32,
anneal_factor=0.5,
patience=5,
max_epochs=5)
self.clf = classifier
def test():
classifier = TextClassifier.load(model_file)
sentence = Sentence('Awesome stuff!')
classifier.predict(sentence)
print(sentence.labels)
# coding utf-8
# docker run --name flask -p 5000:5000 -v G:\workspace:/flask -it python bash
# export FLASK_APP="server.py"
# flask run --host=0.0.0.0
# or python -m forever.run -t 1000000 python server.py &
from flask import Flask, jsonify
import feedparser
from flair.data import build_japanese_tokenizer, Sentence
from flair.models import TextClassifier
app = Flask(__name__)
classifier = TextClassifier.load('resources/best-model.pt')
japanese_tokenizer = build_japanese_tokenizer()
def get_score(text):
# create example sentence
sentence = Sentence(text, use_tokenizer=japanese_tokenizer)
# predict class and print
classifier.predict(sentence)
label_dict = sentence.to_dict()["labels"][0]
return label_dict["confidence"] if label_dict[
"value"] == "__label__O" else 0
def get_feed():
RSS_URL = "https://www.lifehacker.jp/feed/index.xml"
feed = feedparser.parse(RSS_URL)
#############################################################################################
#############################################################################################
print("#####################################")
print("########## EMOTIONS ##############")
print("#####################################")
from flair.models import TextClassifier
# load sentiment model
classifier = TextClassifier.load('en-sentiment')
# example sentence
sentence = Sentence('Porto wins. I am happy')
# predict NER tags
classifier.predict(sentence)
# print sentence with predicted labels
print(sentence.labels)
print("#####################################")
print("########## EMOTIONS ##############")
print("#####################################")
def train(self, intent_fst) -> None:
from flair.data import Sentence, Token
from flair.models import SequenceTagger, TextClassifier
from flair.embeddings import (
FlairEmbeddings,
StackedEmbeddings,
DocumentRNNEmbeddings,
)
from flair.data import TaggedCorpus
from flair.trainers import ModelTrainer
# Directory to look for downloaded embeddings
cache_dir = self.profile.read_path(
self.profile.get("intent.flair.cache_dir", "flair/cache")
)
os.makedirs(cache_dir, exist_ok=True)
# Directory to store generated models
data_dir = self.profile.write_path(
self.profile.get("intent.flair.data_dir", "flair/data")
)
if os.path.exists(data_dir):
shutil.rmtree(data_dir)
self.embeddings = self.profile.get("intent.flair.embeddings", [])
assert len(self.embeddings) > 0, "No word embeddings"
# Create directories to write training data to
class_data_dir = os.path.join(data_dir, "classification")
ner_data_dir = os.path.join(data_dir, "ner")
os.makedirs(class_data_dir, exist_ok=True)
os.makedirs(ner_data_dir, exist_ok=True)
# Convert FST to training data
class_data_path = os.path.join(class_data_dir, "train.txt")
ner_data_path = os.path.join(ner_data_dir, "train.txt")
# { intent: [ { 'text': ..., 'entities': { ... } }, ... ] }
sentences_by_intent: Dict[str, Any] = {}
# Get sentences for training
do_sampling = self.profile.get("intent.flair.do_sampling", True)
start_time = time.time()
if do_sampling:
# Sample from each intent FST
num_samples = int(self.profile.get("intent.flair.num_samples", 10000))
intent_map_path = self.profile.read_path(
self.profile.get("training.intent.intent_map", "intent_map.json")
)
with open(intent_map_path, "r") as intent_map_file:
intent_map = json.load(intent_map_file)
# Gather FSTs for all known intents
fsts_dir = self.profile.write_dir(
self.profile.get("speech_to_text.fsts_dir")
)
intent_fst_paths = {
intent_id: os.path.join(fsts_dir, f"{intent_id}.fst")
for intent_id in intent_map.keys()
}
# Generate samples
self._logger.debug(
f"Generating {num_samples} sample(s) from {len(intent_fst_paths)} intent(s)"
)
sentences_by_intent = sample_sentences_by_intent(
intent_fst_paths, num_samples
)
else:
# Exhaustively generate all sentences
self._logger.debug(
"Generating all possible sentences (may take a long time)"
)
sentences_by_intent = make_sentences_by_intent(intent_fst)
sentence_time = time.time() - start_time
self._logger.debug(f"Generated sentences in {sentence_time} second(s)")
# Get least common multiple in order to balance sentences by intent
lcm_sentences = lcm(*(len(sents) for sents in sentences_by_intent.values()))
# Generate examples
class_sentences = []
ner_sentences: Dict[str, List[Sentence]] = defaultdict(list)
for intent_name, intent_sents in sentences_by_intent.items():
num_repeats = max(1, lcm_sentences // len(intent_sents))
for intent_sent in intent_sents:
# Only train an intent classifier if there's more than one intent
if len(sentences_by_intent) > 1:
# Add balanced copies
for i in range(num_repeats):
class_sent = Sentence(labels=[intent_name])
for word in intent_sent["tokens"]:
class_sent.add_token(Token(word))
class_sentences.append(class_sent)
if len(intent_sent["entities"]) == 0:
continue # no entities, no sequence tagger
# Named entity recognition (NER) example
token_idx = 0
entity_start = {ev["start"]: ev for ev in intent_sent["entities"]}
entity_end = {ev["end"]: ev for ev in intent_sent["entities"]}
entity = None
word_tags = []
for word in intent_sent["tokens"]:
# Determine tag label
tag = "O" if not entity else f"I-{entity}"
if token_idx in entity_start:
entity = entity_start[token_idx]["entity"]
tag = f"B-{entity}"
word_tags.append((word, tag))
# word ner
token_idx += len(word) + 1
if (token_idx - 1) in entity_end:
entity = None
# Add balanced copies
for i in range(num_repeats):
ner_sent = Sentence()
for word, tag in word_tags:
token = Token(word)
token.add_tag("ner", tag)
ner_sent.add_token(token)
ner_sentences[intent_name].append(ner_sent)
# Start training
max_epochs = int(self.profile.get("intent.flair.max_epochs", 100))
# Load word embeddings
self._logger.debug(f"Loading word embeddings from {cache_dir}")
word_embeddings = [
FlairEmbeddings(os.path.join(cache_dir, "embeddings", e))
for e in self.embeddings
]
if len(class_sentences) > 0:
self._logger.debug("Training intent classifier")
# Random 80/10/10 split
class_train, class_dev, class_test = self._split_data(class_sentences)
class_corpus = TaggedCorpus(class_train, class_dev, class_test)
# Intent classification
doc_embeddings = DocumentRNNEmbeddings(
word_embeddings,
hidden_size=512,
reproject_words=True,
reproject_words_dimension=256,
)
classifier = TextClassifier(
doc_embeddings,
label_dictionary=class_corpus.make_label_dictionary(),
multi_label=False,
)
self._logger.debug(
f"Intent classifier has {len(class_sentences)} example(s)"
)
trainer = ModelTrainer(classifier, class_corpus)
trainer.train(class_data_dir, max_epochs=max_epochs)
else:
self._logger.info("Skipping intent classifier training")
if len(ner_sentences) > 0:
self._logger.debug(f"Training {len(ner_sentences)} NER sequence tagger(s)")
# Named entity recognition
stacked_embeddings = StackedEmbeddings(word_embeddings)
for intent_name, intent_ner_sents in ner_sentences.items():
ner_train, ner_dev, ner_test = self._split_data(intent_ner_sents)
ner_corpus = TaggedCorpus(ner_train, ner_dev, ner_test)
tagger = SequenceTagger(
hidden_size=256,
embeddings=stacked_embeddings,
tag_dictionary=ner_corpus.make_tag_dictionary(tag_type="ner"),
tag_type="ner",
use_crf=True,
)
ner_intent_dir = os.path.join(ner_data_dir, intent_name)
os.makedirs(ner_intent_dir, exist_ok=True)
self._logger.debug(
f"NER tagger for {intent_name} has {len(intent_ner_sents)} example(s)"
)
trainer = ModelTrainer(tagger, ner_corpus)
trainer.train(ner_intent_dir, max_epochs=max_epochs)
else:
self._logger.info("Skipping NER sequence tagger training")
elif (val.value == 'b'):
row.pred_neg = 0
row.pred_pos = 1
elif (val.value == 'a'):
row.pred_neg = 0
row.pred_pos = 0
elif (val.value == 'd'):
row.pred_neg = 1
row.pred_pos = 0
else:
print("incorrectVal {}".format(val))
return row
if __name__ == "__main__":
classifier = TextClassifier.load('./data/POLARITY_MULTI/best-model.pt')
data_to_pred = pd.read_csv('./data/test_gold.csv',sep='\t')
data_to_pred['pred_pos'] = 0
data_to_pred['pred_neg'] = 0
data_to_pred = data_to_pred.apply(lambda row: addPredictionForPositive(row), axis=1)
data_to_pred.head()
data_to_save = data_to_pred[['idtwitter','subj','pred_pos','pred_neg','iro','lpos','lneg','top']]
data_to_save.idtwitter = data_to_save.idtwitter.astype(str)
data_to_save.subj = data_to_save.subj.astype(str)
data_to_save.pred_pos = data_to_save.pred_neg.astype(str)
data_to_save.pred_neg = data_to_save.pred_neg.astype(str)
data_to_save.iro = data_to_save.iro.astype(str)
data_to_save.lpos = data_to_save.lpos.astype(str)
data_to_save.lneg = data_to_save.lneg.astype(str)
data_to_save.top = data_to_save.top.astype(str)
print(label_dictionary)
flat_labels = [item for sublist in labels for item in sublist]
class_weights = compute_class_weight('balanced', np.unique(flat_labels),
flat_labels)
unique_labels = np.unique(flat_labels)
weights = {}
for i in range(len(unique_labels)):
weights[unique_labels[i]] = class_weights[i]
document_embeddings = TransformerDocumentEmbeddings(
params['version_model'], fine_tune=True)
classifier = TextClassifier(document_embeddings,
label_dictionary=label_dictionary,
loss_weights=weights,
multi_label=False)
trainer = ModelTrainer(classifier, corpus)
trainer.train(params['model_dir'],
learning_rate=params['learning_rate'],
mini_batch_size=params['batch_size'],
anneal_factor=params['anneal_factor'],
patience=params['patience'],
max_epochs=params['epochs'],
embeddings_storage_mode=params['embeddings_storage_mode'])
# print_predictions(trainer, tokens_test, params['results']+'gloveSentence')
test_run = args.test_run
nrows = args.nrows
model_type = args.model
# read data
df = datatable.fread(f"./data/from_USNavy_for_flair.csv").to_pandas()
print(len(df))
df = df.head(nrows)
data = df.copy()
data = data[['sentence', "row"]]
# load classifier
from flair.models import TextClassifier
from flair.data import Sentence
model = TextClassifier.load(
f'./data/model_{model_type}_{fold}/_{test_run}_best-model.pt')
document_embeddings = model.document_embeddings
# prepare df for output to csv
# batch size for embedding tweet instances
bs = 32
tweets_to_embed = data['sentence'].copy()
print("beginning embedding")
# prepare mini batches
low_limits = list()
for x in range(0, len(tweets_to_embed), bs):
low_limits.append(x)
up_limits = [x + bs for x in low_limits[:-1]]
up_limits.append(len(tweets_to_embed))
# FlairEmbeddings('news-forward'),
# FlairEmbeddings('news-backward'),
]
# 4. initialize document embedding by passing list of word embeddings
# Can choose between many RNN types (GRU by default, to change use rnn_type parameter)
document_embeddings: DocumentRNNEmbeddings = DocumentRNNEmbeddings(
word_embeddings,
hidden_size=512,
reproject_words=True,
reproject_words_dimension=256,
)
# 5. create the text classifier
classifier = TextClassifier(document_embeddings,
label_dictionary=label_dict,
multi_label=False)
# 6. initialize the text classifier trainer
trainer = ModelTrainer(classifier, corpus)
# 7. start the training
trainer.train('resources/taggers/ag_news',
learning_rate=0.1,
mini_batch_size=32,
anneal_factor=0.5,
patience=5,
max_epochs=150)
# 8. plot training curves (optional)
from flair.visual.training_curves import Plotter
import numpy as np
from nlppreprocess import NLP
app = Flask(__name__)
app.secret_key = "super_secret_key"
APP_ROOT = os.path.dirname(os.path.abspath(__file__))
UPLOAD_FOLDER = os.path.join(APP_ROOT, 'uploads')
app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
# # app.config['SECRET_KEY'] = 'oh_so_secret'
# app.config['MONGO_DBNAME'] = 'exposeModel'
# app.config['MONGO_URI'] = 'mongodb://*****:*****@app.route('/', methods=['GET'])
def index():
return jsonify("welcome to Sadeeq API")
@app.route('/api/tasks', methods=['GET'])
def get_result():
result = []
try:
# result = session['my_result']
result = modelResult
def trainer(file_path: Path, filenames: Tuple[str, str, str], checkpoint: str,
stack: str, n_epochs: int) -> None:
"""Train sentiment model using Flair NLP library:
https://github.com/zalandoresearch/flair/blob/master/resources/docs/TUTORIAL_7_TRAINING_A_MODEL.md
To help provide added context, we can stack Glove, Bert or ELMo embeddings along with Flair embeddings.
"""
# pip install flair allennlp
from flair.datasets import ClassificationCorpus
from flair.embeddings import FlairEmbeddings, DocumentRNNEmbeddings
from flair.models import TextClassifier
from flair.trainers import ModelTrainer
from flair.training_utils import EvaluationMetric
from flair.visual.training_curves import Plotter
if stack == "glove":
from flair.embeddings import WordEmbeddings
stacked_embedding = WordEmbeddings('glove')
elif stack == "elmo":
from flair.embeddings import ELMoEmbeddings
stacked_embedding = ELMoEmbeddings('original')
elif stack == "bert":
from flair.embeddings import BertEmbeddings
stacked_embedding = BertEmbeddings('bert-base-cased')
else:
stacked_embedding = None
# Define and Load corpus from the provided dataset
train, dev, test = filenames
corpus = ClassificationCorpus(
file_path,
train_file=train,
dev_file=dev,
test_file=test,
)
# Create label dictionary from provided labels in data
label_dict = corpus.make_label_dictionary()
# Stack Flair string-embeddings with optional embeddings
word_embeddings = list(
filter(None, [
stacked_embedding,
FlairEmbeddings('news-forward'),
FlairEmbeddings('news-backward'),
]))
# Initialize document embedding by passing list of word embeddings
document_embeddings = DocumentRNNEmbeddings(
word_embeddings,
hidden_size=512,
reproject_words=True,
reproject_words_dimension=256,
)
# Define classifier
classifier = TextClassifier(document_embeddings,
label_dictionary=label_dict,
multi_label=False)
if not checkpoint:
trainer = ModelTrainer(classifier, corpus)
else:
# If checkpoint file is defined, resume training
checkpoint = classifier.load_checkpoint(Path(checkpoint))
trainer = ModelTrainer.load_from_checkpoint(checkpoint, corpus)
# Begin training (enable checkpointing to continue training at a later time, if desired)
trainer.train(file_path,
EvaluationMetric.MACRO_F1_SCORE,
max_epochs=n_epochs,
checkpoint=True)
# Plot curves and store weights and losses
plotter = Plotter()
plotter.plot_training_curves(file_path / 'loss.tsv')
plotter.plot_weights(file_path / 'weights.txt')
WordEmbeddings('glove'),
#FlairEmbeddings('news-forward'),
#FlairEmbeddings('news-backward'),
]
# initialize document embedding by passing list of word embeddings
# Can choose between many RNN types (GRU by default, to change use rnn_type parameter)
document_embeddings: DocumentRNNEmbeddings = DocumentRNNEmbeddings(
word_embeddings,
hidden_size=512,
reproject_words=True,
reproject_words_dimension=256)
# create the text classifier
classifier = TextClassifier(document_embeddings,
label_dictionary=label_dict,
multi_label=False)
# initialize the text classifier trainer
trainer = ModelTrainer(classifier, corpus)
trainer.train('resources/clf',
learning_rate=0.1,
mini_batch_size=64,
anneal_factor=0.5,
patience=1,
max_epochs=60)
classifier = TextClassifier.load('resources/clf/final-model.pt')
# create example sentence
f1_list = []
mcc_list = []
f1sub_list = []
f1perf_list = []
f1rec_list = []
f1tec_list = []
f1time_list = []
f1ll_list = []
bas_list = []
# main loop
for fold in range(k_folds):
# load pre-trained classifier
try:
subject_classifier = TextClassifier.load(
'./data/{}/model_subject_category_{}/{}_best-model.pt'.format(
dataset, fold, model_type))
except FileNotFoundError:
print("----- Does such test run exist? Try another name. -----")
quit()
# placeholder for results of predictions
flair_subject = []
# add timestamp before all classification takes place
time_schedule.append(time.perf_counter())
# Main Analysis Loop: start analysis
for i in range(len(data)):
# get the sentence to be analyzed
sent = [str(data.iloc[i, 2])]
class TARSClassifier(FewshotClassifier):
"""
TARS model for text classification. In the backend, the model uses a BERT based binary
text classifier which given a <label, text> pair predicts the probability of two classes
"True", and "False". The input data is a usual Sentence object which is inflated
by the model internally before pushing it through the transformer stack of BERT.
"""
static_label_type = "tars_label"
def __init__(
self,
task_name: str,
label_dictionary: Dictionary,
label_type: str,
embeddings: str = 'bert-base-uncased',
num_negative_labels_to_sample: int = 2,
prefix: bool = True,
**tagger_args,
):
"""
Initializes a TextClassifier
:param task_name: a string depicting the name of the task
:param label_dictionary: dictionary of labels you want to predict
:param embeddings: name of the pre-trained transformer model e.g.,
'bert-base-uncased' etc
:param num_negative_labels_to_sample: number of negative labels to sample for each
positive labels against a sentence during training. Defaults to 2 negative
labels for each positive label. The model would sample all the negative labels
if None is passed. That slows down the training considerably.
:param multi_label: auto-detected by default, but you can set this to True
to force multi-label predictionor False to force single-label prediction
:param multi_label_threshold: If multi-label you can set the threshold to make predictions
:param beta: Parameter for F-beta score for evaluation and training annealing
"""
super(TARSClassifier, self).__init__()
from flair.embeddings import TransformerDocumentEmbeddings
if not isinstance(embeddings, TransformerDocumentEmbeddings):
embeddings = TransformerDocumentEmbeddings(model=embeddings,
fine_tune=True,
layers='-1',
layer_mean=False,
)
# prepare TARS dictionary
tars_dictionary = Dictionary(add_unk=False)
tars_dictionary.add_item('False')
tars_dictionary.add_item('True')
# initialize a bare-bones sequence tagger
self.tars_model = TextClassifier(document_embeddings=embeddings,
label_dictionary=tars_dictionary,
label_type=self.static_label_type,
**tagger_args,
)
# transformer separator
self.separator = str(self.tars_embeddings.tokenizer.sep_token)
if self.tars_embeddings.tokenizer._bos_token:
self.separator += str(self.tars_embeddings.tokenizer.bos_token)
self.prefix = prefix
self.num_negative_labels_to_sample = num_negative_labels_to_sample
# Store task specific labels since TARS can handle multiple tasks
self.add_and_switch_to_new_task(task_name, label_dictionary, label_type)
def _get_tars_formatted_sentence(self, label, sentence):
original_text = sentence.to_tokenized_string()
label_text_pair = f"{label} {self.separator} {original_text}" if self.prefix \
else f"{original_text} {self.separator} {label}"
sentence_labels = [label.value for label in sentence.get_labels(self.get_current_label_type())]
tars_label = "True" if label in sentence_labels else "False"
tars_sentence = Sentence(label_text_pair, use_tokenizer=False).add_label(self.static_label_type, tars_label)
return tars_sentence
def _get_state_dict(self):
model_state = {
"state_dict": self.state_dict(),
"current_task": self._current_task,
"label_type": self.get_current_label_type(),
"label_dictionary": self.get_current_label_dictionary(),
"tars_model": self.tars_model,
"num_negative_labels_to_sample": self.num_negative_labels_to_sample,
"task_specific_attributes": self._task_specific_attributes,
}
return model_state
@staticmethod
def _init_model_with_state_dict(state):
print("init TARS")
# init new TARS classifier
label_dictionary = state["label_dictionary"]
model: TARSClassifier = TARSClassifier(
task_name=state["current_task"],
label_dictionary=label_dictionary,
label_type=state["label_type"],
embeddings=state["tars_model"].document_embeddings,
num_negative_labels_to_sample=state["num_negative_labels_to_sample"],
)
# set all task information
model.task_specific_attributes = state["task_specific_attributes"]
# linear layers of internal classifier
model.load_state_dict(state["state_dict"])
return model
@staticmethod
def _fetch_model(model_name) -> str:
model_map = {}
hu_path: str = "https://nlp.informatik.hu-berlin.de/resources/models"
model_map["tars-base"] = "/".join([hu_path, "tars-base", "tars-base-v8.pt"])
cache_dir = Path("models")
if model_name in model_map:
model_name = cached_path(model_map[model_name], cache_dir=cache_dir)
return model_name
@property
def tars_embeddings(self):
return self.tars_model.document_embeddings
def predict(
self,
sentences: Union[List[Sentence], Sentence],
mini_batch_size=32,
verbose: bool = False,
label_name: Optional[str] = None,
return_loss=False,
embedding_storage_mode="none",
):
# return
"""
Predict sequence tags for Named Entity Recognition task
:param sentences: a Sentence or a List of Sentence
:param mini_batch_size: size of the minibatch, usually bigger is more rapid but consume more memory,
up to a point when it has no more effect.
:param all_tag_prob: True to compute the score for each tag on each token,
otherwise only the score of the best tag is returned
:param verbose: set to True to display a progress bar
:param return_loss: set to True to return loss
:param label_name: set this to change the name of the label type that is predicted
:param embedding_storage_mode: default is 'none' which is always best. Only set to 'cpu' or 'gpu' if
you wish to not only predict, but also keep the generated embeddings in CPU or GPU memory respectively.
'gpu' to store embeddings in GPU memory.
"""
if label_name == None:
label_name = self.get_current_label_type()
# with torch.no_grad():
if not sentences:
return sentences
if isinstance(sentences, Sentence):
sentences = [sentences]
# set context if not set already
previous_sentence = None
for sentence in sentences:
if sentence.is_context_set(): continue
sentence._previous_sentence = previous_sentence
sentence._next_sentence = None
if previous_sentence: previous_sentence._next_sentence = sentence
previous_sentence = sentence
# reverse sort all sequences by their length
rev_order_len_index = sorted(range(len(sentences)), key=lambda k: len(sentences[k]), reverse=True)
reordered_sentences: List[Union[Sentence, str]] = [sentences[index] for index in rev_order_len_index]
dataloader = DataLoader(dataset=SentenceDataset(reordered_sentences), batch_size=mini_batch_size)
# progress bar for verbosity
if verbose:
dataloader = tqdm(dataloader)
overall_loss = 0
overall_count = 0
batch_no = 0
with torch.no_grad():
for batch in dataloader:
batch_no += 1
if verbose:
dataloader.set_description(f"Inferencing on batch {batch_no}")
batch = self._filter_empty_sentences(batch)
# stop if all sentences are empty
if not batch:
continue
# go through each sentence in the batch
for sentence in batch:
# always remove tags first
sentence.remove_labels(label_name)
all_labels = [label.decode("utf-8") for label in self.get_current_label_dictionary().idx2item]
all_detected = {}
for label in all_labels:
tars_sentence = self._get_tars_formatted_sentence(label, sentence)
loss_and_count = self.tars_model.predict(tars_sentence,
label_name=label_name,
return_loss=True)
overall_loss += loss_and_count[0].item()
overall_count += loss_and_count[1]
predicted_tars_label = tars_sentence.get_labels(label_name)[0]
if predicted_tars_label.value == "True":
sentence.add_label(label_name, label, predicted_tars_label.score)
# clearing token embeddings to save memory
store_embeddings(batch, storage_mode=embedding_storage_mode)
if return_loss:
return overall_loss, overall_count
#!/usr/bin/env python
# coding: utf-8
# In[1]:
from flair.models import TextClassifier
from flair.data import Sentence
classifier = TextClassifier.load('en-sentiment')
sentence = Sentence('Flair is pretty neat!')
classifier.predict(sentence)
# print sentence with predicted labels
print('Sentence above is: ', sentence.labels)
# In[1]:
import pandas as pd
data = pd.read_csv("./spam.csv", encoding='latin-1').sample(frac=1).drop_duplicates()
data = data[['v1', 'v2']].rename(columns={"v1":"label", "v2":"text"})
data['label'] = '__label__' + data['label'].astype(str)
data.iloc[0:int(len(data)*0.8)].to_csv('train.csv', sep='\t', index = False, header = False)
data.iloc[int(len(data)*0.8):int(len(data)*0.9)].to_csv('test.csv', sep='\t', index = False, header = False)
data.iloc[int(len(data)*0.9):].to_csv('dev.csv', sep='\t', index = False, header = False)
# In[2]:
except FileNotFoundError:
print(
"File couldn't be found. Verify if '{f_path}' is a correct file path!"
.format(f_path=csv_file))
exit(1)
if __name__ == '__main__':
print(">>> STARTING PROGRAM")
df = load_dataset_from(DATASET_FILE)
# To verify if data loaded correctly:
# print(df.head(10))
classifier = TextClassifier.load('sentiment')
tmp_negatives = {}
tmp_positives = {}
print(f'Number of players: {len(df["Player"].unique())}')
for player_name in df["Player"].unique():
tmp_negatives[player_name] = list()
tmp_positives[player_name] = list()
for dp in df.values:
l = dp.tolist()
sentence = Sentence(l[-1])
classifier.predict(sentence)
if sentence.labels[0].value == "NEGATIVE":
def __init__(self, model_name_or_path: str):
self.classifier = TextClassifier.load(model_name_or_path)
def __init__(self, c_classifier_file, p_classifier_file):
self.classifier_c = TextClassifier.load_from_file(c_classifier_file)
self.classifier_p = TextClassifier.load_from_file(p_classifier_file)
# if not universal sentence encoder
if not _use:
# load Flair
import torch
import flair
from flair.models import TextClassifier
# load various embeddings
from flair.embeddings import WordEmbeddings, DocumentPoolEmbeddings, RoBERTaEmbeddings
from flair.data import Sentence
# if trained embeddings
if not pool:
# embeddings trained to the "downstream task"
model = TextClassifier.load(
f'./data/model_sentiment_{fold}/{test_run}_best-model.pt')
document_embeddings = model.document_embeddings
# if simple pooled embeddings
else:
if test_run == "fasttext":
document_embeddings = DocumentPoolEmbeddings(
[WordEmbeddings('en-twitter')])
elif test_run == "roberta":
document_embeddings = DocumentPoolEmbeddings([
RoBERTaEmbeddings(
pretrained_model_name_or_path="roberta-large",
layers="21,22,23,24",
pooling_operation="first",
use_scalar_mix=True)
])
def initialize_training(text_column_index,
label_column_index,
delimiter=';',
model_type=None,
model=None,
max_epochs=10,
patience=3,
use_amp=0,
calc_class_weights=0):
"""
Create a text classification model using FLAIR, SentenceTransformers and
Huggingface Transformers.
Params:
data_folder_path: Folder path with each file titled appropriately i.e.
train.csv test.csv dev.csv.
Will create a 80/10/10 split if only train is supplied.
output_folder_path: Folder path for storing the best model & checkpoints.
text_column_index: In which index (starting from 0) the input column is located.
label_column_index: In which index (starting from 0) the label column is located.
delimiter: type of delimiter used in the .csv file.
model_type: SentenceTransformerDocumentEmbeddings or TransformerDocumentEmbeddings
model: Which model to use.
max_epochs: Number of epochs to train the model for.
patience: Number of epochs without improvement before adjusting learning rate.
use_amp: Whether to enable automatic mixed precisions (AMP).
calc_class_weights: Whether to create a dictionary with class weights to deal
with imbalanced datasets.
Output:
best-model.pt
final-model.pt
training.log
"""
# 1. Column format indicating which columns hold the text and label(s)
column_name_map = {
text_column_index: "text",
label_column_index: "label_topic"
}
# 2. Load corpus containing training, test and dev data.
corpus: Corpus = CSVClassificationCorpus("/root/text-classification/data/",
column_name_map,
skip_header=True,
delimiter=delimiter)
# Print statistics about the corpus.
training_data_statistics = corpus.obtain_statistics()
print(training_data_statistics)
# 3A. Create a label dictionary.
label_dict = corpus.make_label_dictionary()
# 3B. Calculate class weights.
if bool(calc_class_weights):
weight_dict = create_weight_dict(delimiter=delimiter,
label_index=label_column_index)
else:
weight_dict = None
# 4. Initialize the sentence_transformers model.
if model_type == "SentenceTransformerDocumentEmbeddings":
document_embeddings = SentenceTransformerDocumentEmbeddings(model)
elif model_type == "TransformerDocumentEmbeddings":
document_embeddings = TransformerDocumentEmbeddings(model,
fine_tune=True)
elif model_type == "WordEmbeddings":
word_embeddings = [WordEmbeddings(model)]
document_embeddings = DocumentRNNEmbeddings(word_embeddings,
hidden_size=256)
elif model_type == "StackedEmbeddings":
document_embeddings = DocumentRNNEmbeddings([
WordEmbeddings('glove'),
FlairEmbeddings(model + '-backward'),
FlairEmbeddings(model + '-forward')
])
else:
raise Exception(
"Pick SentenceTransformerDocumentEmbeddings, StackedEmbeddings, WordEmbeddings or TransformerDocumentEmbeddings."
)
# 5. create the text classifier
classifier = TextClassifier(document_embeddings,
label_dictionary=label_dict,
loss_weights=weight_dict)
# 6. initialize the text classifier trainer with Adam optimizer
trainer = ModelTrainer(classifier,
corpus,
optimizer=Adam,
use_tensorboard=False)
# 7. start the training
trainer.train("/root/text-classification/checkpoint/",
learning_rate=3e-5,
max_epochs=max_epochs,
patience=patience,
use_amp=bool(use_amp),
checkpoint=True,
mini_batch_size=16,
mini_batch_chunk_size=4)
# In[125]:
# initialize document embedding by passing list of word embeddings
document_embeddings: DocumentRNNEmbeddings = DocumentRNNEmbeddings(stacked_embeddings,
hidden_size=512,
reproject_words=True,
reproject_words_dimension=256,
)
# In[126]:
# create text classifier
classifier = TextClassifier(document_embeddings, label_dictionary=label_dict, multi_label=False)
# In[127]:
# initialize the text classifier trainer
trainer = ModelTrainer(classifier, corpus)
# In[ ]:
trainer.train('FLAIR_data\\BERT',
learning_rate=0.1,
mini_batch_size=16,
def main(data_folder, benchmark_classifier_folder, new_data_folder,
finetuned_classifier_folder):
from flair.embeddings import FlairEmbeddings, DocumentLSTMEmbeddings, BertEmbeddings, DocumentRNNEmbeddings, TransformerDocumentEmbeddings
from flair.models import TextClassifier
from flair.trainers import ModelTrainer
from flair.datasets import CSVClassificationCorpus
from flair.data import Corpus
import pandas as pd
import os
### First Stage (Train on benchmark dataset)
benchmark = pd.read_csv(data_folder + "combined_benchmark.csv")
benchmark = benchmark[['label', 'text']]
#### Create train, dev and test set
#benchmark = benchmark.sample(frac=1) # if not set random state, everytime has different training result
benchmark = benchmark.sample(frac=1, random_state=42)
benchmark.iloc[0:int(len(benchmark) * 0.8)].to_csv(data_folder +
'train.csv',
sep='\t',
index=False,
header=False)
benchmark.iloc[int(len(benchmark) * 0.8):int(len(benchmark) * 0.9)].to_csv(
data_folder + 'test.csv', sep='\t', index=False, header=False)
benchmark.iloc[int(len(benchmark) * 0.9):].to_csv(data_folder + 'dev.csv',
sep='\t',
index=False,
header=False)
#### Build corpus
column_name_map = {1: "text", 0: "label_topic"}
corpus: Corpus = CSVClassificationCorpus(
data_folder,
column_name_map,
skip_header=False, #no header in kaggle data
delimiter='\t', # comma separated rows
#train_file='train.csv', ## passing in file names manually when it can't auto detect
#dev_file='dev.csv',
#test_file='test.csv'
)
#### Create word embeddings
word_embeddings = [
BertEmbeddings(),
FlairEmbeddings('news-forward-fast'),
FlairEmbeddings('news-backward-fast')
]
## caveat: issue of deprecation. BertEmbeddings and DocumentLSTMEmbeddings existed in version 0.4.5, and became legacy embeddings(still available) in version 0.5
#### First Stage Fine-tuning
document_embeddings = DocumentLSTMEmbeddings(word_embeddings,
hidden_size=512,
reproject_words=True,
reproject_words_dimension=256)
classifier = TextClassifier(
document_embeddings,
label_dictionary=corpus.make_label_dictionary(),
multi_label=False)
trainer = ModelTrainer(classifier, corpus)
#trainer.train(benchmark_classifier_folder, max_epochs=1) #offline test use epoch=1
trainer.train(benchmark_classifier_folder, max_epochs=10)
### every finetuning results in different scores
### accuracy at phase1 finetuning does not matter too much, phase2 scores more important in biasing the models towards learning indicator-specific keywords
### Second Stage (train on hand annotated datasets)
#### Build corpus
### this column_name_map must be updated to reflect which column stores the X(text features) and y(golden labels) for training use
### in the csv file contained in new_data_folder, 2nd column is 'title_desc',
### 4th column is 'title_desc_sent_1' (where we stored agreed sentiment annotations)
new_column_name_map = {1: "text", 3: "label_topic"}
print(new_column_name_map)
corpus: Corpus = CSVClassificationCorpus(
new_data_folder,
new_column_name_map,
skip_header=True,
delimiter=',' # comma separated rows
)
#### Second Stage fine-tuning
benchmark_classifier = TextClassifier.load(
os.path.join(benchmark_classifier_folder, 'best-model.pt'))
trainer = ModelTrainer(benchmark_classifier, corpus)
#trainer.train(finetuned_classifier_folder, max_epochs=1) #offline test use
trainer.train(finetuned_classifier_folder, max_epochs=10)
def spamcleaner(text):
classifier = TextClassifier.load('./utils/model/best-model.pt')
sentence = Sentence(text)
classifier.predict(sentence)
print(sentence.labels)
return (str(sentence.labels))