def aggregate_file(interval, file_name, start=None):
""" Aggregate the data within the windows of time
interval: time in seconds to aggregate data
file_name: which file to record
start: start time to record data, if none given then it starts
from te beginning.
returns: array of the aggregated data in each interval
"""
if start is None:
start = get_start_time_for(file_name)
start = datetime.strptime(start, TIME_FORMAT)
summaries = [Summarizer() for _ in range(10)]
with open(file_name, 'r+') as data:
headers = data.readline().strip().lower().split(',')
for line in data:
args = line.strip().split(',')
time = datetime.strptime(args[0], TIME_FORMAT)
window = int((time - start).total_seconds() / interval)
if window < 0:
continue
if window >= len(summaries):
for i in range(window + 1):
summaries.append(Summarizer())
item = dict(zip(headers, args))
summaries[window].add(item)
return [s for s in summaries if s.used]
def summarization():
try:
json_data = request.get_json(force=True)
text = json_data.get('text')
method = json_data.get('method', 'T5')
pretrained = json_data.get('pretrained', 't5-large')
# initialize Summarizer
s = Summarizer(method=method, pretrained=pretrained)
pred = s.summarize(text)
if isinstance(pred, dict):
summary = pred['summary']
data = dict()
data['text'] = text
data['summary'] = summary
data['summary_len'] = Utility.get_doc_length(summary)
data['text_len'] = Utility.get_doc_length(text)
return json.dumps(data)
else:
return json.dumps(pred)
except Exception as e:
return {"Error": str(e)}
def Summ(spacy_model, body, ratio, coreference, greedyness, min_lenght,
max_lenght, bert_model):
# Loading Spanish BERT custom model and tokenizer (output_hidden_states must be set to True).
custom_config = AutoConfig.from_pretrained(
pretrained_model_name_or_path=bert_model)
custom_config.output_hidden_states = True
custom_tokenizer = AutoTokenizer.from_pretrained(bert_model)
custom_model = AutoModel.from_pretrained(bert_model, config=custom_config)
#Summarization
if coreference:
# With Coreference
handler = CoreferenceHandler(spacy_model=spacy_model,
greedyness=greedyness)
model = Summarizer(custom_model=custom_model,
custom_tokenizer=custom_tokenizer,
sentence_handler=handler)
result = model(body, min_length=min_lenght)
else:
# Without Coreference
model = Summarizer(custom_model=custom_model,
custom_tokenizer=custom_tokenizer)
result = model(body,
ratio=ratio,
min_length=min_lenght,
max_lenght=max_lenght)
summarized_text = ''.join(result)
return summarized_text
def test_run():
if path.exists("keys"):
api = Summarizer("Manchester", "manchester", "keys")
assert api.keyToVideo() == 0
else:
with open("data.json", "r") as f:
return json.load(f)
def get_weight_sentence(document):
text = document.text
theme = document.theme
lang = document.language
nlp = nlp_it
if lang == 'italian':
model = Summarizer(sentence_handler=SentenceHandler(language=Italian))
args = get_words(text, nlp)
else:
model = Summarizer(sentence_handler=SentenceHandler(language=Russian))
args = get_words_for_ru(text)
document.args = str(args)
document.save()
sentences = sent_tokenize(text)
if lang == 'italian':
sorted_sentences = get_sorted_sentence(sentences, nlp, text, args,
theme, lang)
else:
sorted_sentences = get_sorted_sentence_for_ru(sentences, text, args,
theme, lang)
note = generate(sentences, sorted_sentences)
note_with_ml = model(text)
note_item = Note(document_id=document,
text=note,
text_for_algo="",
text_for_ml=note_with_ml)
note_item.save()
def post():
try:
url = request.args.get('url')
summarizer = Summarizer(url)
return render_template('post.html', text=Markup(summarizer.summarize()), img= summarizer.article.top_image, title=summarizer.title), 200
except:
raise InvalidUsage('Oops something got wrong. Please try again', status_code=404)
def process_text(text):
"""
Tries to get the summary, intro, keywords
Return is_valid indicating if text was okay
:param text:
:param title:
:return:
"""
is_valid = True
keywords = []
summarizer = Summarizer(text)
try:
summary = summarizer.summarize()
keywords = summarizer.keywords(number_keywords=number_keywords)
phrases = summarizer.key_noun_phrases(
number_noun_phrases=number_noun_phrase)
phrases = [phrase.replace(" ", "_") for phrase in phrases]
keywords = keywords + phrases
except SummaryException:
summary = [""]
is_valid = False
summary = " ".join(summary)
return is_valid, summary, keywords
def inference(_id, corrector, segmenter):
res = getBodyTopic(_id)
body = res['text']
topic = res['topic']
username = res['username']
length = res['length']
body = preProcessTranscript(body, corrector)
entities = entityRecog(_id)
# topic = preProcessTranscript(topic)
meeting_sentiment = sentiment(_id)
# body = correct_sentence__test(body, corrector, segmenter)
model = Summarizer()
data = model.get_summary(body, topic)
size = len(data)
result = ""
if size < length:
result = [result + data[i]['sentence'] for i in range(len(data))]
else:
result = [result + data[i]['sentence'] for i in range(length)]
result = " ".join(result)
res = saveSummary(_id, result, _id,
username, length, meeting_sentiment, entities, topic)
return {"_id": str(res.inserted_id)}
def summarize():
url = request.args.get('url')
summarizer = Summarizer(url)
if url != False:
summary = summarizer.summarize()
response = {'title': summarizer.title, 'summary': summary, 'img': summarizer.article.top_image}
return jsonify(response)
raise InvalidUsage('Oops something got wrong. Please try again', status_code=404)
class TextTeaser(object):
def __init__(self):
self.summarizer = Summarizer()
def summarize(self, title, text, category = "Undefined", source = "Undefined", count = 5):
result = self.summarizer.summarize(text, title, source, category)
result = self.summarizer.sortSentences(result[:count])
result = [res['sentence'] for res in result]
return result
def test_generate_summary(self):
"""Various test cases for summarizer.generate_summary."""
summarizer_simple = Summarizer(self.simple_dataframe, self.config)
output_df_simple = summarizer_simple.generate_summary()
self.assertEqual(len(output_df_simple), 3)
self.assertIn(1, output_df_simple['Size'].values)
self.assertIn('test', output_df_simple['Text'].values)
self.assertIn('test2', output_df_simple['Text'].values)
self.assertIn('test3', output_df_simple['Text'].values)
self.assertIn('', output_df_simple['ClassLines'].values)
summarizer_stack_lines = Summarizer(self.stack_lines_dataframe,
self.config)
output_df_stack_lines = summarizer_stack_lines.generate_summary()
self.assertIn('some.class.java',
output_df_stack_lines['ClassLines'].values)
self.assertIn('some.class2.java',
output_df_stack_lines['ClassLines'].values)
self.assertIn('some.class3.java',
output_df_stack_lines['ClassLines'].values)
self.assertIn('', output_df_stack_lines['Text'].values)
summarizer_multi_cluster = Summarizer(self.multi_cluster_dataframe,
self.config)
output_df_multi_cluster = summarizer_multi_cluster.generate_summary()
self.assertEqual(len(output_df_multi_cluster), 2)
self.assertIn(2, output_df_multi_cluster['Size'].values)
self.assertIn(1, output_df_multi_cluster['Size'].values)
def pre_processing(list_lineInDoc,
caseIdx,
output_path="./output/train_data_preprocessed.txt",
separator=DEFAULT_SEPARATOR,
write_valid=True,
logPrint=False):
'''
To Jeff:
Implement here.
'''
if (logPrint == True):
print("")
print("\tPre-processing law case [%d]" % caseIdx)
if (caseIdx == 0 and logPrint == True):
print("")
print("\t\t1.Remove \\n, \\t")
print("\t\t2.Remove Punctuations.")
print("\t\t3.Split Words.")
print("\t\t4.Remove StopWords.")
print("\t\t5.Lemmatize Words.")
## 1. Remove '\n','\t', etc and make it readable.
for idx in range(len(list_lineInDoc)):
list_lineInDoc[idx] = " ".join(list_lineInDoc[idx].replace(
"\n", " ").split())
## 2. Analyse catchphrase and sentences.
catchphrase = list_lineInDoc[0]
sentences = list_lineInDoc[1:]
summarizer = Summarizer()
result = summarizer.summarize(sentences, catchphrase)
## 3. Save result.
if (write_valid == True):
f = open(output_path, 'a')
f.write(separator + str(caseIdx))
f.write('\n')
for i in range(0, len(result)):
word_stat_line = result[i][0]
for j in range(0, len(word_stat_line)):
f.write(str(word_stat_line[j]['word']))
f.write(":")
f.write(str(word_stat_line[j]['count']))
f.write(" ")
f.write('\n')
f.close()
return result
class TextTeaser(object):
def __init__(self):
self.summarizer = Summarizer()
def summarize(self, text, count):
result = self.summarizer.summarize(text)
#print(result)
result = self.summarizer.sortSentences(result[:count])
result = [res['sentence'] for res in result]
return result
class nlps(object):
def __init__(self):
self.summarizer = Summarizer()
def summarize(self, title, text, category="Undefined", source="Undefined", count=5):
result = self.summarizer.summarize(text, title, source, category)
result = self.summarizer.sortSentences(result[:count])
result = [res['sentence'] for res in result]
return result
def worker():
i = 0
while True:
item = q.get()
if item is None:
print("Break ! cuz item is None")
break
api = Summarizer(item[0], item[1], "keys")
api.keyToVideo()
i += 1
print("-----Task{0}----".format(i))
q.task_done()
def summary_of_detection(filename,
model,
use_bots=False,
use_attack=False,
sample=False,
use_ahead=False,
steps_ahead=1,
trees=50,
norm_and_standardize=False):
"""General call to train and test any model under any features.
Args:
filename: Aggregated file to get features and labels from
model: Abbreviated model name to use (rf: Random Forest, dl: Deep Learning, dt: Decision Trees)
use_bots: Use bot type as the label (Only general-*.featureset.csv files have this information)
use_attack: Use attack type as the label (only all-*.featureset.csv files have this information)
sample: Sample the featureset so that there is an equal number of attack and normal labels
use_ahead: Works only on files aggregated with `--use_separator` flag turned on.
steps_ahead: Only used if use_ahead is true. The amount if time windows ahead to skip to get the new label
for the current feature set.
trees: The number of trees to use on Random Forest.
norm_and_standardize: The normalize and standardize the numerical features.
Returns:
The accuracy, precision, recall, and f1_score of the model.
"""
if use_ahead:
xtrain, xtest, ytrain, ytest = get_ahead_feature_labels(
filename,
Summarizer().features, steps_ahead)
else:
xtrain, xtest, ytrain, ytest = get_features_labels_from(
filename,
Summarizer().features,
use_bots,
use_attack,
sample=sample,
norm_and_standardize=norm_and_standardize)
if model == 'rf':
clf = rf_train(xtrain, ytrain, use_attack, use_ahead, trees=trees)
elif model == 'dt':
clf = dt_train(xtrain, ytrain)
elif model == 'dl':
if use_bots:
# Multi-label needs to be converted into a sparse array for Tensorflow.
ytrain = to_tf_labels(ytrain)
ytest = to_tf_labels(ytest)
clf = dl_train(xtrain, ytrain, use_bots)
return dl_test(clf, xtrain, ytrain, use_bots)
if use_attack:
return [
x['micro'] for x in test(clf, xtest, ytest, use_bots, use_attack)
]
return test(clf, xtest, ytest, use_bots, use_ahead=use_ahead)
class ReviewSummaryAction(Action):
summarizer = Summarizer()
@staticmethod
def review_summarization(reviews):
message = "\n"
for review in reviews:
result = ReviewSummaryAction.summarizer(review["text"])
full = ''.join(result)
if full != '':
message += "\n-----\n" + full + " - " + review[
"author_name"] + ", " + review["relative_time_description"]
return message
def name(self) -> Text:
return "review_summary_action"
def run(self, dispatcher: CollectingDispatcher, tracker: Tracker,
domain: Dict[Text, Any]) -> List:
facility_details = tracker.get_slot("facility_details")
if facility_details is None:
error_message = "You have to choose a facility first"
dispatcher.utter_message(error_message)
return []
if "reviews" not in facility_details:
review_message = ATMOSPHERE_DICT.get("review_messages")[-1]
dispatcher.utter_message(review_message)
else:
summary_message = ReviewSummaryAction.review_summarization(
facility_details["reviews"])
print(summary_message)
dispatcher.utter_message(summary_message)
def main():
#ask user for how many summaries they want
print(
"Welcome to CFS's Summary generator. This pulls articles from NLTK's reuter's corpus and summarizers them\n"
)
try:
numarts = eval(
input("How many articles do you wish us to summarize?: "))
except Exception:
print("Uh oh, looks like that wasn't a number... \n Please try again")
return
#prevents users from asking for negative numbers or more than 5 at a time.
if numarts < 1:
print(
"I understand why you wouldn't want to read summaries of financial articles, but come on, at least one? =[ "
)
return
if numarts > 5:
print(
"Okay, do you really want to read more than five bland reuters articles summaries at a time?"
)
print("Please try a more sensible number")
return
print()
# create our summarizer
cfs = Summarizer()
getSummaries(cfs, numarts)
def summarize(body):
model = Summarizer()
result = model(body, min_length=20)
full = ''.join(result)
return full
def load_summarizer():
try:
summ = pickle.load(open('./models/summarizer.pkl', 'rb'))
except:
summ = Summarizer()
pickle.dump(summ, open('./models/summarizer.pkl', 'wb'))
return summ
def text_summarization(stt_file_path):
""" this method take the speech to text output file path
==> apply bert extractive text summarization
==> return the text summary file path
"""
try:
#Read Data From input File (sumtext.txt)
inputfile = open(stt_file_path, "r")
datainput = (inputfile.read())
print("Summarization processing ......")
model = Summarizer()
result = model(datainput, min_length=10)
output_summary = ''.join(result)
#Write Data in output File (bert-extractive-summarizer.txt)
filename = 'summary-bert' + str(
uuid.uuid4()) + '.txt' #genere a unique file name
out_path = 'upload/text_summarization/' + filename
path = os.path.join(MEDIA_ROOT, out_path)
file = open(path, "a")
file.write(output_summary)
file.close()
print('sucess summary >>>>>>')
return out_path
# pylint: disable=bare-except
except:
print("An exception occurred in text summarization")
def load_summarizer():
try:
model = pickle.load(open('./models/summarizer.pkl', 'rb'))
except:
model = Summarizer()
pickle.dump(model, open('./models/summarizer.pkl', 'wb'))
return model
def load(self):
try:
logging.info('Loading BERTExtractiveSummarizer...')
BERTExtractiveSummarizer.model = Summarizer()
except Exception as e:
raise ModelFailedLoadException(
f'Failed to load BERTExtractiveSummarizer with {str(e)}')
def get_summarizer_model(self):
if self.summarizer_model is None:
nlp = spacy.load('en_core_web_md',
disable=['ner', 'parser', 'tagger'])
nlp.add_pipe(nlp.create_pipe('sentencizer'))
self.summarizer_model = Summarizer()
return self.summarizer_model
def bert_summarizer(data):
summarizer_bert = Summarizer()
summary_bert = summarizer_bert(data, min_length=30, max_length=140)
print('Bert for Text - Summarization')
summary = ''.join(summary_bert)
rouge_scores = rouge.get_scores(summary, data)
return summary, rouge_scores
def test_multi_hidden_concat(summarizer_multi_hidden: Summarizer, passage):
summarizer_multi_hidden.hidden_concat = True
res = summarizer_multi_hidden(passage,
num_sentences=5,
min_length=40,
max_length=500)
assert len(res) > 10
def texte_summarizer(input_texte, ratio=0.5):
"""
INPUT:
input_texte : str - il s'agit du texte que l'on souhaite résumer
ratio : float (compris entre 0 et 1) - pourcentage de réduction du texte, si ratio = 0.5 on va garder 50% des phrases du texte d'origine
OUTPUT:
dico : dict - {'summary_texte' : str result} avec result le texte résumé
BODY:
On charge depuis le dossier static/model le modele de NLP à utiliser puis on l'applique au texte à résumer
"""
try:
custom_config = AutoConfig.from_pretrained(
"static/model/camembert-base")
custom_config.output_hidden_states = True
custom_tokenizer = AutoTokenizer.from_pretrained(
"static/model/camembert-base")
custom_model = AutoModel.from_pretrained("static/model/camembert-base",
config=custom_config)
model = Summarizer(
custom_model=custom_model,
custom_tokenizer=custom_tokenizer,
)
result = model(input_texte, ratio=ratio)
dico = {'summary_text': result}
except:
import_bert()
dico = texte_summarizer(input_texte, ratio)
return dico
def load(model: str, tokenizer: BertTokenizer, device=None) -> Summarizer:
config = BertConfig.from_pretrained(model)
config.output_hidden_states = True
bert_model = BertModel.from_pretrained(model, config=config)
if device is not None:
bert_model = bert_model.to(device)
return Summarizer(custom_model=bert_model, custom_tokenizer=tokenizer)
def main():
# Parse the JSON arguments
try:
config_args = parse_args()
except:
print("Add a config file using \'--config file_name.json\'")
exit(1)
# Create the experiment directories
_, config_args.summary_dir, config_args.checkpoint_dir = create_experiment_dirs(
config_args.experiment_dir)
# Reset the default Tensorflow graph
tf.reset_default_graph()
# Tensorflow specific configuration
config = tf.ConfigProto(allow_soft_placement=True)
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
# Data loading
# The batch size is equal to 1 when testing to simulate the real experiment.
data_batch_size = config_args.batch_size if config_args.train_or_test == "train" else 1
data = DataLoader(data_batch_size, config_args.shuffle)
print("Loading Data...")
config_args.img_height, config_args.img_width, config_args.num_channels, \
config_args.train_data_size, config_args.test_data_size = data.load_data()
print("Data loaded\n\n")
# Model creation
print("Building the model...")
model = ShuffleNet(config_args)
print("Model is built successfully\n\n")
# Summarizer creation
summarizer = Summarizer(sess, config_args.summary_dir)
# Train class
trainer = Train(sess, model, data, summarizer)
if config_args.train_or_test == 'train':
try:
print("FLOPs for batch size = " + str(config_args.batch_size) +
"\n")
calculate_flops()
print("Training...")
trainer.train()
print("Training Finished\n\n")
except KeyboardInterrupt:
trainer.save_model()
elif config_args.train_or_test == 'test':
print("FLOPs for single inference \n")
calculate_flops()
# This can be 'val' or 'test' or even 'train' according to the needs.
print("Testing...")
trainer.test('val')
print("Testing Finished\n\n")
else:
raise ValueError("Train or Test options only are allowed")
def extSumm(full):
orgi = full
model = Summarizer(
model='distilbert-base-uncased') # can adjust parameters
#summa = model(orgi, ratio=0.3, min_length=60) # can adjust parameters
summa = model(orgi, ratio=0.2, min_length=60)
return summa
def run():
parser = argparse.ArgumentParser(
description='Process and summarize lectures')
parser.add_argument('-path',
dest='path',
default=None,
help='File path of lecture')
parser.add_argument('-lang',
dest='lang',
default='en',
help='Language supported : en,fr')
parser.add_argument('-model',
dest='model',
default='bert-large-uncased',
help='')
parser.add_argument('-num-sentences',
dest='num_sentences',
default=-1,
help='Will return X sentences')
parser.add_argument(
'-ratio',
dest='ratio',
default=-1,
help=
'Will return a ratio of sentences from the text length (0.2 is a good value)'
)
parser.add_argument('-hidden',
dest='hidden',
default=-2,
help='Which hidden layer to use from Bert')
parser.add_argument('-reduce-option',
dest='reduce_option',
default='mean',
help='How to reduce the hidden layer from bert')
parser.add_argument('-greedyness',
dest='greedyness',
help='Greedyness of the NeuralCoref model',
default=0.45)
args = parser.parse_args()
if not args.path:
raise RuntimeError("Must supply text path.")
with open(args.path) as d:
text_data = d.read()
spacy_model = sentence_handlers[args.lang]
model = Summarizer(
model=args.model,
hidden=args.hidden,
reduce_option=args.reduce_option,
sentence_handler=CoreferenceHandler(spacy_model=spacy_model))
if int(args.num_sentences) >= 0:
result = model(text_data, num_sentences=int(args.num_sentences))
elif float(args.ratio) >= 0.0:
result = model(text_data, ratio=float(args.num_sentences))
else:
result = model(text_data)
print(result)
def Summary(self):
summ = dict.fromkeys(self.clust[0].unique())
# Will store the name of Cluster based on article name
subtitles = []
# Will store the url of main article summarizing the cluster
urls = []
model = Summarizer()
# Create the summary by aggregating all the bodys of the articles
for c in summ.keys():
df2 = self.clust[self.clust[0] == c]
subtitles = subtitles + [self.results.loc[df2.index[0]]['title']]
urls = urls + [self.results.loc[df2.index[0]]['url']]
full_text = ' '.join(self.results.loc[df2.index]['body'].unique())
full_text = full_text.replace('\n', ' ')
length = len(full_text)
if length < 500:
target = 200
ratiol = min(1, target / length)
else:
target = min(self.max_len + 150 * (len(df2)),
length / len(df2) * (0.15 + 0.05 * len(df2)))
ratiol = min(1, target / length)
summ[c] = model(full_text, min_length=60, ratio=ratiol)
self.summary = summ
self.subtitles = subtitles
self.urls = urls
async def process(ctx):
file_path = Formatter(ctx.channel.id).getInputFilePath(
int(open("counter.txt", 'r').readline()))
input_text = ''.join(open(file_path, 'r').readlines())
output = await Summarizer().summarize(ctx, input_text)
Summarizer().save_summaries(output,
int(open("counter.txt", 'r').readline()))
#!-*- coding: utf-8 -*-
from goose import Goose
from goose.text import StopWordsChinese
from summarizer import Summarizer
import json
g = Goose({'stopwords_class': StopWordsChinese})
url = 'http://sports.sina.com.cn/nba/2013-10-29/00086855748.shtml'
article = g.extract(url=url)
title = article.title
print title
text = article.cleaned_text
print text
summary = Summarizer()
summary_list = summary.summarize(title, text)
print 'summary is below:'
for sentence in summary_list:
print sentence
summary_json = json.dumps(summary_list)
print summary_json
def getInput():
with open('input.txt') as file:
content = file.readlines()
# remove unnecessary \n
content = [c.replace('\n', '') for c in content if c != '\n']
title = content[0]
text = content[-(len(content) - 1):]
return {'title': title, 'text': ' '.join(text)}
# #####################
input = getInput()
input['text'] = input['text'].decode("ascii", "ignore")
input['text'] = " ".join(input['text'].replace("\n", " ").split())
summarizer = Summarizer()
result = summarizer.summarize(input['text'], input['title'], 'Undefined', 'Undefined')
result = summarizer.sortScore(result)
result = summarizer.sortSentences(result[:30])
print 'Summary:'
for r in result:
print r['sentence']
# print r['totalScore']
# print r['order']
def __init__(self):
self.summarizer = Summarizer()
if __name__ == '__main__':
mongo_client = MongoClient()
db = mongo_client.g_project_data
coll = db.test_data
summary_list, article_list = get_summaries_and_articles(coll)
mongo_client.close()
idf = unpickle('idf')
vocab = unpickle('vocab')
count = CountVectorizer(vocabulary=vocab, stop_words='english')
summarizer_multi = Summarizer(vocab=vocab, idf=idf, vectorizer=count, scoring='multi_Tfidf')
summarizer_single = Summarizer(vocab=vocab, idf=idf, vectorizer=count, scoring='single_Tfidf')
summarizer_sig = Summarizer(vocab=vocab, idf=idf, vectorizer=count, scoring='significance')
summarizer_sim = Summarizer(vocab=vocab, idf=idf, vectorizer=count, scoring='similarity')
summarizer_rand = Summarizer(vocab=vocab, idf=idf, vectorizer=count, scoring='random')
multi_r2 = []
multi_reduction2 = []
single_r2 = []
single_reduction2 = []
sig_r2 = []
sig_reduction2 = []
sim_r2 = []
sim_reduction2 = []
rand_r2 = []
rand_reduction2 = []
