class ProcessingScheduler:
def __init__(self):
self.processing_queue = deque()
self.actively_processing = {}
self.mongo = MongoDatabase()
def add_video(self, request: dict = {}):
'''Add a video to the processing queue.
@request: dictionary containing the video filename and specified options.
'''
self.processing_queue.append(request)
if not self.actively_processing:
self.actively_processing = self.processing_queue.popleft()
self.process()
def get_status(self):
'''Get the status of each video in the processing pipeline.'''
status = {el['filename']: 'Waiting...' for el in self.processing_queue}
status[self.actively_processing['filename']] = 'Processing'
return status
def process(self):
'''Process a video.'''
print(f'Processing video {self.actively_processing["filename"]}...')
print('Add processing code here.......')
self.mongo.add_processed_video(self.actively_processing)
self.actively_processing = {}
def __init__(self):
super().__init__()
self.ui = mainwindow.Ui_Frec()
self.ui.setupUi(self)
self.db = MongoDatabase()
self.init_frec()
def __init__(self):
logging.info('Initialized {}'.format(self.__class__.__name__))
# instantiate database
self.db = MongoDatabase()
# set utf8 encoding
reload(sys)
sys.setdefaultencoding('utf8')
def __init__(self):
logging.info('Initialized {}'.format(self.__class__.__name__))
# instantiate database
self.db = MongoDatabase()
# location to store plots
self.plot_save_folder = os.path.join('files', 'plots')
# location to store tables to
self.table_save_folder = os.path.join('files', 'tables')
def __init__(self):
super().__init__()
self.ui = mainwindow.Ui_Frec()
self.ui.setupUi(self)
self.db = MongoDatabase()
self.init_frec()
self.is_local_save = False
self.dirs = config.Dirs()
self.setWindowIcon(QIcon(self.dirs.appicon))
def confirmDB(self): print 'In confirmDB' # global path, thumbPath, fullImgPath, self.running #print '\n\n' + str(path) + '\n\n' self.thumbPath=self.path + 'imgThumb' self.fullImgPath=self.path + 'imgFull' if not os.path.exists(self.path): getPath(self) if not os.path.exists(self.thumbPath): os.makedirs(self.thumbPath) if not os.path.exists(self.fullImgPath): os.makedirs(self.fullImgPath) ### Initialize slide db ### # print db.collection_names() if u'slideColl' not in MongoDatabase.db.collection_names(): print '1' MongoDatabase.initSlides(self) else: print '2' MongoDatabase.updateSlides(self)
def __init__(self):
self.components = []
self.database = MongoDatabase("localhost")
self.load_components()
# Specify the port number.
PORT = 1496
# Directory to save videos to.
ROOT_DIR = os.getcwd()
SAVE_DIR = f'{os.getcwd()}/videos'
# Create the server.
app = Flask(__name__)
api = Api(app)
valid_headers = ['Content-Type', 'Access-Control-Allow-Origin', '*']
cors = CORS(app, allow_headers=valid_headers)
# Connect to the database.
mongo = MongoDatabase()
# Video processing.
processor = Processor()
# Do a little server-side checking.
ALLOWED_EXTENSIONS = set(['webm', 'mp4', 'mp3', 'wav', 'jpeg', 'gif', 'png'])
# -------------------------------------------------------------
# Global functions.
def allowed_file(filename):
'''Ensure we want to keep this file.'''
return True
from datetime import datetime
from database import MongoDatabase
db = MongoDatabase()
experiments = db.experiments
format = '%Y-%m-%d'
available_times = []
for a in range(1):
dish = list(db.experiments.find({'dishNumber': str(a + 1)}))
for b in range(1):
start_date = list(dish)[b]['startDate']
start_experiment = start_date
end_date = list(dish)[b]['endDate']
end_experiment = end_date
start_experiment = datetime.strptime(start_experiment, format)
end_experiment = datetime.strptime(end_experiment, format)
start_experiment = int(datetime.timestamp(start_experiment))
end_experiment = int(datetime.timestamp(end_experiment))
print(start_experiment, type(start_experiment))
print(end_experiment, type(end_experiment))
return tweet
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# instantiate twitter object
twitter = Twitter(key=API_KEY, secret=API_SECRET)
# create connection
twitter.connect_to_API()
if get_sanders_tweets:
"""
The Sanders dataset consists out of 5,513 hand classified tweets related to the topics Apple (@Apple), Google (#Google), Microsoft (#Microsoft), and Twitter (#Twitter). Tweets were
classified as positive, neutral, negative, or irrelevant; the latter referring to non-English tweets which we discarded. The Sanders dataset has been used for boosting Twitter
sentiment classification using different sentiment dimensions, combining automatically and hand-labeled twitter sentiment labels, and combining community detection and sentiment
analysis}. The dataset is available from http://www.sananalytics.com/lab/.
Tweets are saved into the collection 'sanders_tweets_raw'
class Frec(QMainWindow):
def __init__(self):
super().__init__()
self.ui = mainwindow.Ui_Frec()
self.ui.setupUi(self)
self.db = MongoDatabase()
self.init_frec()
def init_frec(self):
self.ui.btn_register.clicked.connect(self.save_new_member)
self.ui.btn_delete.clicked.connect(self.delete_member)
self.ui.btn_clear.clicked.connect(self.clear_form)
self.ui.btn_export.clicked.connect(self.export_cvs)
self.ui.btn_import.clicked.connect(self.import_cvs)
self.ui.btn_connectDb.clicked.connect(self.connect_db)
self.connect_db()
self.show_member_at_tableWidget()
def show_message(self, msg, timeout=3000):
self.ui.statusBar.showMessage(msg, timeout)
def connect_db(self):
dbName = self.ui.lne_dbName.text()
if self.db.connect(dbName):
self.show_message("Database connection with '" + dbName +
"' is successful.")
self.ui.lbl_dbConnection.setText("Database : Connected '" +
dbName + "'")
self.ui.lbl_dbConnection.setStyleSheet('color: green')
else:
self.ui.lbl_dbConnection.setText("Database : Disconnected")
self.ui.lbl_dbConnection.setStyleSheet('color: red')
self.show_message("Database connection couldn't established!")
return False
return True
# Save new member to database
def save_new_member(self):
new = {
"firstname": self.ui.lne_firstName.text(),
"surname": self.ui.lne_lastName.text(),
"department": self.ui.comboBox_department.currentText(),
"email": self.ui.lne_email.text(),
"mobilencc": self.ui.lne_mobileCyp.text(),
"mobileother": self.ui.lne_mobileOther.text(),
}
self.db.add_new_member(new['firstname'], new['surname'], new['email'],
new['department'], new['mobilencc'],
new['mobileother'])
self.add_member_to_tableWidget(new['firstname'], new['surname'],
new['email'], new['department'],
new['mobilencc'], new['mobileother'])
# show member on tableview
def add_member_to_tableWidget(self, firstname, surname, email, department,
mobilencc, mobileother):
rowPoint = self.ui.tableWidget.rowCount()
self.table.insertRow(rowPoint)
self.table.setItem(rowPoint, 0, QTableWidgetItem(firstname))
self.table.setItem(rowPoint, 1, QTableWidgetItem(surname))
self.table.setItem(rowPoint, 2, QTableWidgetItem(email))
self.table.setItem(rowPoint, 3, QTableWidgetItem(department))
self.table.setItem(rowPoint, 4, QTableWidgetItem(mobilencc))
self.table.setItem(rowPoint, 5, QTableWidgetItem(mobileother))
def show_member_at_tableWidget(self):
self.table = self.ui.tableWidget
self.table.setColumnCount(6)
self.table.setHorizontalHeaderLabels([
"firstname", "surname", "department", "email", "mobilencc",
"mobileother"
])
quary = {}
result = self.db.query_result_multi("Member", quary)
for member in result:
self.add_member_to_tableWidget(member["name"]["first"],
member["name"]["last"],
member["department"],
member["email"],
member["mobileNo"]["ncc"],
member["mobileNo"]["other"])
# Delete chosen member on tableView
def delete_member(self):
pass # TODO
# Clear inside the line edits in registration form
def clear_form(self):
self.ui.lne_email.setText("")
self.ui.lne_firstName.setText("")
self.ui.lne_lastName.setText("")
self.ui.lne_mobileCyp.setText("")
self.ui.lne_mobileOther.setText("")
self.ui.comboBox_department.setCurrentIndex(0)
# Export members in the database as csv
def export_cvs(self):
pass # TODO
# Import members from a csv file
def import_cvs(self):
pass # TODO
def create_desktop_entry(self):
# TODO: Replace [-HOMEDIR-] inside ./data/freg.desktop and copy it to ~/.local/share/applications/
pass
class Preprocessing():
def __init__(self):
logging.info('Initialized {}'.format(self.__class__.__name__))
# instantiate database
self.db = MongoDatabase()
# set utf8 encoding
reload(sys)
sys.setdefaultencoding('utf8')
def full_text_preprocessing(self, pdf_folder=os.path.join('files', 'pdf')):
"""
preprocess full-text publications
- convert pdf to plain text
- correct for carriage returns
- correct for end-of-line hyphenation
- remove boilerplate
- remove bibliography
- remove acknowledgements
Parameters
----------
pdf_folder : os.path
location where PDF documents are stored
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# read pdf files that need to be converted
F = [x for x in read_directory(pdf_folder) if x[-4:] == '.pdf']
# read documents from DB that have already been processed so we can skip them
processed_documents = [
'{}-{}-{}'.format(x['journal'], x['year'], x['title'])
for x in self.db.read_collection(collection='publications_raw')
]
# loop over each file and convert pdf to plain and save meta data to DB
for i, f in enumerate(F):
# extract meta data from folder structure and file name
journal = f.split('/')[2]
year = f.split('/')[3]
title = f.split('/')[4].replace('-', ' ')[4:-4].strip()
# console output
print_doc_verbose(i, len(F), journal, year, title)
# check if PDF has already been processed
if '{}-{}-{}'.format(journal, year, title) in processed_documents:
logging.info('PDF document already processed, skipping ...')
continue
# convert content of PDF to plain text
content = pdf_to_plain(f)
# check if content could be extracted
if content is not None:
# fix soft hyphen
content = content.replace(u'\xad', "-")
# fix em-dash
content = content.replace(u'\u2014', "-")
# fix en-dash
content = content.replace(u'\u2013', "-")
# minus sign
content = content.replace(u'\u2212', "-")
# fix hyphenation that occur just before a new line
content = content.replace('-\n', '')
# remove new lines/carriage returns
content = content.replace('\n', ' ')
# correct for ligatures
content = content.replace(u'\ufb02', "fl") # fl ligature
content = content.replace(u'\ufb01', "fi") # fi ligature
content = content.replace(u'\ufb00', "ff") # ff ligature
content = content.replace(u'\ufb03', "ffi") # ffi ligature
content = content.replace(u'\ufb04', "ffl") # ffl ligature
"""
Remove boilerplate content:
Especially journal publications have lots of boilerplate content on the titlepage. Removing of this is specific for each
journal and you can use some regular expressions to identify and remove it.
"""
"""
Remove acknowledgemends and/or references
This is a somewhat crude example
"""
if content.rfind("References") > 0:
content = content[:content.rfind("References")]
"""
Remove acknowledgements
"""
if content.rfind("Acknowledgment") > 0:
content = content[:content.rfind("Acknowledgment")]
# prepare dictionary to save into MongoDB
doc = {
'journal': journal,
'title': title,
'year': year,
'content': content
}
# save to database
self.db.insert_one_to_collection(doc=doc,
collection='publications_raw')
def general_preprocessing(self, min_bigram_count=5):
"""
General preprocessing of publications (used for abstracts and full-text)
Parameters
----------
min_bigram_count : int (optional)
frequency of bigram to occur to include into list of bigrams. Thus lower frequency than min_bigram_count will not be included.
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# read document collection
D = self.db.read_collection(collection='publications_raw')
# setup spacy natural language processing object
nlp = setup_spacy()
# loop through the documents and correct content
for i, d in enumerate(D):
# check if tokens are already present, if so, skip
if d.get('tokens') is None:
# print to console
print_doc_verbose(i, D.count(), d['journal'], d['year'],
d['title'])
# get content from document and convert to spacy object
content = nlp(d['content'])
# tokenize, lemmatization, remove punctuation, remove single character words
unigrams = word_tokenizer(content)
# get entities
entities = named_entity_recognition(content)
# get bigrams
bigrams = get_bigrams(" ".join(unigrams))
bigrams = [['{} {}'.format(x[0], x[1])] * y
for x, y in Counter(bigrams).most_common()
if y >= min_bigram_count]
bigrams = list(itertools.chain(*bigrams))
d['tokens'] = unigrams + bigrams + entities
# save dictionary to datbase
self.db.update_collection(collection='publications_raw', doc=d)
else:
logging.debug('Document already tokenized, skipping ...')
def __init__(self):
self.processing_queue = deque()
self.actively_processing = {}
self.mongo = MongoDatabase()
from helper_functions import *
from database import MongoDatabase
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# location of target tweets
location_tweets = os.path.join('files', 'target_tweets')
# modes of research
modes_of_research = ['interdisciplinary', 'multidisciplinary','transdisciplinary']
# process tweets for each mode of research
for mode in modes_of_research:
logging.info('Processing mode of research: {}'.format(mode))
# read tweets files
F = read_directory(os.path.join(location_tweets, mode))
return tweet
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# instantiate twitter object
twitter = Twitter(key = API_KEY, secret = API_SECRET)
# create connection
twitter.connect_to_API()
if get_sanders_tweets:
"""
The Sanders dataset consists out of 5,513 hand classified tweets related to the topics Apple (@Apple), Google (#Google), Microsoft (#Microsoft), and Twitter (#Twitter). Tweets were
classified as positive, neutral, negative, or irrelevant; the latter referring to non-English tweets which we discarded. The Sanders dataset has been used for boosting Twitter
sentiment classification using different sentiment dimensions, combining automatically and hand-labeled twitter sentiment labels, and combining community detection and sentiment
analysis}. The dataset is available from http://www.sananalytics.com/lab/.
class Interpretation():
def __init__(self):
logging.info('Initialized {}'.format(self.__class__.__name__))
# instantiate database
self.db = MongoDatabase()
# location to store plots
self.plot_save_folder = os.path.join('files', 'plots')
# location to store tables to
self.table_save_folder = os.path.join('files', 'tables')
def infer_document_topic_distribution(
self,
K=10,
dir_prior='auto',
random_state=42,
num_pass=15,
iteration=200,
top_n_words=10,
models_folder=os.path.join('files', 'models'),
lda_files_folder=os.path.join('files', 'lda')):
"""
Infer the document topic distribition per publication. The LDA model shows us the word probabilies per topic, but we also want to know what
topics we find within each document. Here we infer such document-topic distribution and save it to the databse so we can use it later
to plot some interesting views of the corpus
Values for K, dir_prior, random_state, num_pass and iteratrion will become visible when plotting the coherence score. Use the model that
achieved the highest coherence score.
Parameters
-----------
k: int
number of topics that resulted in the best decomposition of the underlying corpora
dir_prior: string
dirichlet priors 'auto', 'symmetric', 'asymmetric'
random_state: int
seed value for random initialization
num_pass: int
number of passes over the full corpus
iteration: int
max iterations for convergence
top_n_words: int
only print out the top N high probability words
models_folder: os.path
location of created LDA models
lda_files_folder: os.path
location of LDA corpus and dictionary
save_folder: os.path
location to store the tables
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# read dictionary and corpus
dictionary, corpus = get_dic_corpus(lda_files_folder)
# load LDA model according to parameters
model = load_lda_model(
os.path.join(models_folder, str(K), dir_prior, str(random_state),
str(num_pass), str(iteration)))
# load docs
D = self.db.read_collection(collection='publications_raw')
# loop through all the documents to infer document-topics distribition
for i, d in enumerate(D):
# check if tokens are present; in case some documents couldn't properly be tokenized during pre-processing phase
if d.get('tokens') is not None:
# print to console
print_doc_verbose(i, D.count(), d['journal'], d['year'],
d['title'])
# create bag of words from tokens
bow = model.id2word.doc2bow(d['tokens'])
# infer document-topic distribution
topics = model.get_document_topics(bow, per_word_topics=False)
# convert to dictionary: here we convert the topic number to string because mongodb will complain otherwise
# you will get a message that documents can only have string keys
dic_topics = {}
for t in topics:
dic_topics[str(t[0])] = float(t[1])
# create a new document to add to the database, this time in a different collection
insert_doc = {
'journal': d['journal'],
'year': d['year'],
'title': d['title'],
'topics': dic_topics
}
# save insert_doc to database within publications collection
self.db.insert_one_to_collection('publications', insert_doc)
def get_document_title_per_topic(self):
"""
Get document title per topic
Here we obtain the publication title of the most dominant topic within that publication
Most dominant topic is the topic proportion that is the largest
So if document has topic A = 10%, B = 30%, and C = 60%, then C is the dominant topic
We can use the titles for the dominant topics to get insights into the label of that topic
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# load docs
D = self.db.read_collection(collection='publications')
# empty list where we can append publication titles to
titles = []
# loop trough all the docs
for i, d in enumerate(D):
# print to console
print_doc_verbose(i, D.count(), d['journal'], d['year'],
d['title'])
# get the dominant topic
dominant_topic = max(d['topics'].iteritems(), key=itemgetter(1))
# get the topic ID and percentage
dominant_topic_id, dominant_topic_percentage = dominant_topic[
0], dominant_topic[1]
# append to list
titles.append([
d['year'], d['title'], d['journal'], dominant_topic_id,
dominant_topic_percentage
])
# save to CSV
save_csv(titles, 'titles-to-topics', folder=self.table_save_folder)
def plot_topics_over_time(self, plot_save_name='topics-over-time.pdf'):
"""
Plot cumulative topic distribution over time
Parameters
----------
plot_save_name: string
name of the plot
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# load docs
D = self.db.read_collection(collection='publications')
# create dictionary where we can obtain the topic distribution per year
year_to_topics = get_year_to_topics(D)
# calculate the cumulative topic distribution: basically the average distribution per year
year_to_cum_topics = get_year_to_cum_topics(year_to_topics)
# convert dictionary to pandas dataframe
df = pd.DataFrame.from_dict(year_to_cum_topics)
# create the plot
fig, axs = plt.subplots(2, 5, figsize=(15, 10))
axs = axs.ravel()
# loop over each row of the dataframe
for index, row in df.iterrows():
# get year values
x = df.columns.values.tolist()
# get topic proportions
y = row.tolist()
# add to plot
axs[index].plot(x,
y,
'o--',
color='black',
linewidth=1,
label="Topic prevalence")
axs[index].set_title(get_topic_label(index), fontsize=14)
axs[index].set_ylim([0, 0.4])
# save plot
plt.savefig(os.path.join(self.plot_save_folder, plot_save_name),
bbox_inches='tight')
plt.close()
def plot_topics_over_time_stacked(
self, plot_save_name='topics-over-time-stacked.pdf'):
"""
Plot topics over time stacked
Parameters
----------
plot_save_name: string
name of the plot
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# load docs
D = self.db.read_collection(collection='publications')
# create dictionary where we can obtain the topic distribution per year
year_to_topics = get_year_to_topics(D)
# calculate the cumulative topic distribution: basically the average distribution per year
year_to_cum_topics = get_year_to_cum_topics(year_to_topics)
# convert dictionary to pandas dataframe
df = pd.DataFrame.from_dict(year_to_cum_topics)
# transpose dataframe
df = df.transpose()
# change column headers into topic labels
df.columns = [get_topic_label(x) for x in df.columns.values]
# plot the dataframe
ax = df.plot(figsize=(15, 8),
kind='area',
colormap='Spectral_r',
rot=45,
grid=False)
# set values for x-axis
plt.xticks(df.index)
# limit the x-axis
plt.xlim(min(df.index), max(df.index))
# limit the y-axis
plt.ylim(0, 1)
# get the legend
handles, labels = ax.get_legend_handles_labels()
# position it right of the figure
plt.legend(reversed(handles),
reversed(labels),
loc='right',
bbox_to_anchor=(1.35, 0.50),
ncol=1,
fancybox=False,
shadow=False,
fontsize=16)
# save plot
plt.savefig(os.path.join(self.plot_save_folder, plot_save_name),
bbox_inches='tight')
plt.close()
def plot_topic_co_occurrence(self,
plot_save_name='topic-co-occurrence.pdf'):
"""
Plot topic co-occurrence
Parameters
----------
plot_save_name: string
name of the plot
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# load docs
D = self.db.read_collection(collection='publications')
# create empty dictionary where we can store the dominant topic id and remaining other proportions
dominant_id_to_topics = {}
for d in D:
# sort topics and create list
topics = [
value for key, value in sorted(d['topics'].iteritems(),
key=lambda x: int(x[0]))
]
# get max topix id
max_topic_id = topics.index(max(topics))
# check if topic ID key already created
if max_topic_id not in dominant_id_to_topics:
dominant_id_to_topics[max_topic_id] = []
dominant_id_to_topics[max_topic_id].append(topics)
# create empty dictionary where we can have the cumulative topic distribution per dominant topic ID
dominant_id_to_cum_topics = {}
for k, v in dominant_id_to_topics.iteritems():
# calculate mean and add to dictionary
dominant_id_to_cum_topics[k] = np.mean(np.array(v), axis=0) * 100.
# convert dictionary to pandas dataframe
df = pd.DataFrame.from_dict(dominant_id_to_cum_topics)
# change column headers into topic labels
df.columns = [get_topic_label(x) for x in df.columns.values]
df.index = [get_topic_label(x) for x in df.index.values]
# create max column
df['max'] = 0.
# keep track of new index
new_index = []
# add max column so we can sort on it later
for index, row in df.iterrows():
# add value to max column
df['max'][index] = max(row)
# make self co-occurrence zero
df[index][index] = 0.0
# add new index names to tracker so we can rename it later
new_index.append('{} ({}%)'.format(index, round(max(row), 2)))
# update index name
df.index = new_index
# sort by max column
df = df.sort_values(by=['max'], ascending=False)
# delete max column
df = df.drop(['max'], axis=1)
# sort based on column totals
df = df.reindex(sorted(df.columns), axis=1)
# plot the heatmap
ax = sns.heatmap(df,
cmap="Blues",
annot=True,
vmin=0.,
vmax=10.,
square=True,
annot_kws={"size": 11},
fmt='.1f',
mask=df <= 0.0,
linewidths=.5,
cbar=False,
yticklabels=True)
# adjust the figure somewhat
ax.xaxis.tick_top()
plt.yticks(rotation=0)
plt.xticks(rotation=90, ha='left')
fig = ax.get_figure()
fig.set_size_inches(19, 6)
# save figure
fig.savefig(os.path.join(self.plot_save_folder, plot_save_name),
bbox_inches='tight')
def plot_topics_in_journals(self, plot_save_name='topics-in-journals.pdf'):
"""
Plot the distribution of topics within each of the journals in our dataset.
This plot provides an overview of the topical content published by a journal given the time frame of our dataset
Parameters
----------
plot_save_name: string
name of the plot
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# create dictionary where we have key = journal, and value = [topic_distributions]
journal_to_topics = {}
# load documents from database
D = self.db.read_collection(collection='publications')
# loop over the documents, read in the topic distribution, and add to the correct journal key
for i, d in enumerate(D):
# verbose process every 1000th document
if i % 1000 == 0:
logging.debug('Processing document {}/{}'.format(i, D.count()))
# get the name of the journal
journal = d['journal']
# check if topics are created
if d.get('topics') is not None:
# add journal as key to the dictionary if not already exists
if journal not in journal_to_topics:
# add journal as key with empty list
journal_to_topics[journal] = []
# sort topics and create as list
topics = [
value for key, value in sorted(d['topics'].iteritems(),
key=lambda x: int(x[0]))
]
# append topic distribution to dictionary
journal_to_topics[journal].append(topics)
# get cumulative topic distributions for each journa
journal_to_cum_topics = get_journal_to_cum_topics(journal_to_topics)
# convert to Pandas DataFrame
df = pd.DataFrame.from_dict(journal_to_cum_topics).T
# change column labels to topic labels
df.columns = [get_topic_label(x) for x in df.columns.values]
# plot the heatmap
ax = sns.heatmap(df,
cmap="Blues",
annot=True,
vmin=0.,
vmax=.3,
square=True,
annot_kws={"size": 11},
fmt='.2f',
mask=df <= 0.0,
linewidths=.5,
cbar=False,
yticklabels=True)
# adjust the figure somewhat
ax.xaxis.tick_top()
plt.yticks(rotation=0)
plt.xticks(rotation=90, ha='left')
fig = ax.get_figure()
fig.set_size_inches(10, 10)
# save figure
fig.savefig(os.path.join(self.plot_save_folder, plot_save_name),
bbox_inches='tight')
# close thee plot
plt.close()
import pathlib
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# location of target tweets
location_tweets = os.path.join('files', 'target_tweets_UNCW')
# modes of research
modes_of_research = ['UNCW']
# process tweets for each mode of research
for mode in modes_of_research:
logging.info('Processing mode of research: {}'.format(mode))
# read tweets files
F = read_directory(os.path.join(location_tweets, mode))
print(F)
tweet_id_to_label = {l[0]: l[1] for l in labels}
# define the label types
label_types = ['positive', 'neutral', 'negative']
# define the colors for each sentiment (postive, neutral and negative)
colors = ['#52bf80', '#088bdc', '#fe6300']
# plot location
plot_location = os.path.join('files', 'plots')
# create location if not exists
create_directory(plot_location)
# create database connection
db = MongoDatabase()
if create_donot_plot:
"""
Create a donutplot for interdisciplinary, transdisciplinary and multidisciplinary target tweets where each
donutplot shows the percentage of positive, negative and neutral tweets
file will be saved to files/plots/donutplot.pdf
"""
# create the subplts
fig, axs = plt.subplots(1, 3, figsize=(21, 6))
# make axes available like ax[0] instead of ax[0,0]
axs = axs.ravel()
import json
from helper_functions import *
from database import MongoDatabase
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# name of collection to store all the training tweets to
db_collection = 'training_tweets'
# setup spacy object, so we can do some NLP things
nlp = setup_spacy()
# sources to process and the collections they are stored in
process_sources = {
'sanders': 'sanders_tweets_raw',
'semeval': 'semeval_tweets_raw',
'clarin13': 'clarin13_tweets_raw',
'hcr': 'hcr_tweets_raw',
'omd': 'omd_tweets_raw',
'stanford': 'stanford_tweets_raw',
from database import MongoDatabase
from sklearn.externals import joblib
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
# verbose
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# load classifier
clf = joblib.load(os.path.join('files', 'ml_models', 'LinearSVC.pkl'))
# read labels for target tweets that have been manually labeled and convert to dictionary with key = tweet ID and value = label
true_labels = {
d['tweet_id']: d['label']
for d in db.read_collection(collection='manual_tweets_raw')
}
# load tweets for which we want to infer the sentiment label
D = db.read_collection(collection='target_tweets')
# create empty numpy array so we can retrieve labels later on somewhat faster
labels = np.zeros((D.count(), 3), dtype=np.int)
class Evaluation():
def __init__(self):
logging.info('Initialized {}'.format(self.__class__.__name__))
# instantiate database
self.db = MongoDatabase()
def calculate_coherence(self,
file_folder=os.path.join('files', 'lda'),
models_folder=os.path.join('files', 'models')):
"""
Calculate the CV coherence score for each of the created LDA models
Parameters
----------
file_folder: os.path
location of the dictionary and corpus for gensim
models_folder: os.path
location where the lda model is saved
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# read dictionary and corpus
dictionary, corpus = get_dic_corpus(file_folder)
# load bag of words features of each document from the database
texts = [
x['tokens'] for x in self.db.read_collection('publications_raw')
]
# get path location for models
M = [
x for x in read_directory(models_folder) if x.endswith('lda.model')
]
# read processed models from database
processed_models = [
'{}-{}-{}-{}-{}'.format(x['k'], x['dir_prior'], x['random_state'],
x['num_pass'], x['iteration'])
for x in self.db.read_collection('coherence')
]
# calculate coherence score for each model
for i, m in enumerate(M):
logging.info('Calculating coherence score: {}/{}'.format(
i + 1, len(M)))
print m
# number of topics
k = m.split(os.sep)[2]
# different dirichlet priors
dir_prior = m.split(os.sep)[3]
# random initiatilizations
random_state = m.split(os.sep)[4]
# passes over the corpus
num_pass = m.split(os.sep)[5]
# max iteration for convergence
iteration = m.split(os.sep)[6]
logging.info(
'k: {}, dir_prior: {}, random_state: {}, num_pass: {}, iteration: {}'
.format(k, dir_prior, random_state, num_pass, iteration))
# check if coherence score already obtained
if '{}-{}-{}-{}-{}'.format(k, dir_prior, random_state, num_pass,
iteration) not in processed_models:
# load LDA model
model = models.LdaModel.load(m)
# get coherence c_v score
coherence_c_v = CoherenceModel(model=model,
texts=texts,
dictionary=dictionary,
coherence='c_v')
# get coherence score
score = coherence_c_v.get_coherence()
# logging output
logging.info('coherence score: {}'.format(score))
# save score to database
doc = {
'k': k,
'dir_prior': dir_prior,
'random_state': random_state,
'num_pass': num_pass,
'iteration': iteration,
'coherence_score': score
}
self.db.insert_one_to_collection('coherence', doc)
else:
logging.info(
'coherence score already calculated, skipping ...')
continue
def plot_coherence(self,
min_k=2,
max_k=20,
save_location=os.path.join('files', 'plots'),
plot_save_name='coherence_scores_heatmap.pdf'):
"""
Read coherence scores from database and create heatmap to plot scores
Parameters
-----------
min_k: int
owest number of topics created when creating LDA models. Here 2
max_k: int
highest number of topics created when creating LDA models. Here 20
save_location: os.path
location where to save the plot
plot_save_name: string
name for the plot
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# make sure plot save location exists
create_directory(save_location)
# read documents from database that contain coherence scores
D = list(self.db.read_collection(collection='coherence'))
# convert data from document into a list
data = [[
int(x['k']), x['dir_prior'], x['random_state'], x['num_pass'],
x['iteration'], x['coherence_score']
] for x in D]
# create empty dataframe where we can store our scores
df = pd.DataFrame()
# loop trough values of k parameter and find relevant scores for each grid search combination
for k in range(min_k, max_k + 1):
# create dataframe to temporarily store values
df_temp = pd.DataFrame(index=[k])
# loop trough the data to obtain only the scores for a specific k value
for row in sorted(data):
if row[0] == k:
df_temp['{}-{}-{}-{}'.format(
row[1], row[2], row[3], row[4])] = pd.Series(row[5],
index=[k])
# append temporarary dataframe of only 1 k value to the full dataframe
df = df.append(df_temp)
# transpose the dataframe
df = df.transpose()
# plot the heatmap
ax = sns.heatmap(df,
cmap="Blues",
annot=True,
vmin=0.500,
vmax=0.530,
square=True,
annot_kws={"size": 11},
fmt='.3f',
linewidths=.5,
cbar_kws={'label': 'coherence score'})
# adjust the figure somewhat
ax.xaxis.tick_top()
plt.yticks(rotation=0)
plt.xticks(rotation=0, ha='left')
fig = ax.get_figure()
fig.set_size_inches(19, 6)
# save figure
fig.savefig(os.path.join(save_location, plot_save_name),
bbox_inches='tight')
def output_lda_topics(self,
K=9,
dir_prior='auto',
random_state=42,
num_pass=15,
iteration=200,
top_n_words=10,
models_folder=os.path.join('files', 'models'),
save_folder=os.path.join('files', 'tables')):
"""
Create table with LDA topic words and probabilities
Creates a table of topic words and probabilties + topics in a list format
Values for K, dir_prior, random_state, num_pass and iteratrion will become visible when plotting the coherence score. Use the model that
achieved the highest coherence score and plug in the correct values. The values will create the correct file location of the LDA model
for example : files/models/2/auto/42/5/200/lda.model
Parameters
-----------
k: int
number of topics that resulted in the best decomposition of the underlying corpora
dir_prior: string
dirichlet priors 'auto', 'symmetric', 'asymmetric'
random_state: int
seed value for random initialization
num_pass: int
number of passes over the full corpus
iteration: int
max iterations for convergence
top_n_words: int
only print out the top N high probability words
models_folder: os.path
location of created LDA models
save_folder: os.path
location to store the tables
"""
logging.info('Start {}'.format(sys._getframe().f_code.co_name))
# load LDA model according to parameters
model = load_lda_model(
os.path.join(models_folder, str(K), dir_prior, str(random_state),
str(num_pass), str(iteration)))
# define empty lists so we can fill them with words
topic_table, topic_list = [], []
# loop trough all the topics found within K
for k in range(K):
# create topic header, e.g. (1) TOPIC X
topic_table.append([
'{}'.format(
get_topic_label(k, labels_available=False).upper())
])
# add column for word and probability
topic_table.append(["word", "prob."])
list_string = ""
topic_string = ""
topic_string_list = []
# get topic distribution for topic k and return only top-N words
scores = model.print_topic(k, top_n_words).split("+")
# loop trough each word and probability
for score in scores:
# extract score and trimm spaces
score = score.strip()
# split on *
split_scores = score.split('*')
# get percentage
percentage = split_scores[0]
# get word
word = split_scores[1].strip('"')
# add word and percentage to table
topic_table.append(
[word.upper(), "" + percentage.replace("0.", ".")])
# add word to list table
list_string += word + ", "
# add empty line for the table
topic_table.append([""])
# add topic words to list
topic_list.append([str(k + 1), list_string.rstrip(", ")])
# save to CSV
save_csv(topic_list, 'topic-list', folder=save_folder)
save_csv(topic_table, 'topic-table', folder=save_folder)
def exitProgram(): print 'Ending Process!' MongoDatabase.terminate() exit()
def __init__(self):
logging.info('Initialized {}'.format(self.__class__.__name__))
# instantiate database
self.db = MongoDatabase()
# switches, set to True what needs to be executed
filter_tweets = True
clean_tweets = True
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# execute if set to True
if filter_tweets:
"""
Filter raw target tweet
- remove non-English tweets
- remove retweet
- remove tweets that do not originate from an academic or scientist (by using bio text)
raw tweets are stored in the collection 'raw_tweets'
filtered tweets will be stored in the collectin 'filtered_tweets'
"""
# read tweets documents from database
D = db.read_collection(collection='raw_tweets')
class Frec(QMainWindow):
def __init__(self):
super().__init__()
self.ui = mainwindow.Ui_Frec()
self.ui.setupUi(self)
self.db = MongoDatabase()
self.init_frec()
def init_frec(self):
self.ui.btn_register.clicked.connect(self.save_new_member)
self.ui.btn_delete.clicked.connect(self.delete_member)
self.ui.btn_createDesktopEntry.clicked.connect(self.create_desktop_entry)
self.ui.btn_clear.clicked.connect(self.clear_form)
self.ui.btn_export.clicked.connect(self.export_cvs)
self.ui.btn_import.clicked.connect(self.import_cvs)
self.ui.btn_exportAsCVS.clicked.connect(self.export_cvs)
self.ui.btn_connectDb.clicked.connect(self.connect_db)
self.connect_db()
self.show_member_at_tableWidget()
def show_message(self, msg, timeout=3000):
self.ui.statusBar.showMessage(msg, timeout)
def connect_db(self):
dbName = self.ui.lne_dbName.text()
if self.db.connect(dbName):
self.show_message("Database connection with '" + dbName + "' is successful.")
self.ui.lbl_dbConnection.setText("Database : Connected '" + dbName + "'")
self.ui.lbl_dbConnection.setStyleSheet('color: green')
else:
self.ui.lbl_dbConnection.setText("Database : Disconnected")
self.ui.lbl_dbConnection.setStyleSheet('color: red')
self.show_message("Database connection couldn't established!")
return False
return True
# Save new member to database
def save_new_member(self):
new = {
"firstname": self.ui.lne_firstName.text(),
"surname": self.ui.lne_lastName.text(),
"department": self.ui.comboBox_department.currentText(),
"email": self.ui.lne_email.text(),
"mobilecyp": self.ui.lne_mobileCyp.text(),
"mobileother": self.ui.lne_mobileOther.text(),
}
if self.db.add_new_member(new['firstname'], new['surname'], new['email'], new['department'], new['mobilecyp'],
new['mobileother']) is False:
return False
self.add_member_to_tableWidget(new['firstname'], new['surname'], new['email'], new['department'],
new['mobilecyp'], new['mobileother'])
# show member on tableview
def add_member_to_tableWidget(self, firstname, surname, email, department, mobilecyp, mobileother):
rowPoint = self.ui.tableWidget.rowCount()
self.table.insertRow(rowPoint)
self.table.setItem(rowPoint, 0, QTableWidgetItem(firstname))
self.table.setItem(rowPoint, 1, QTableWidgetItem(surname))
self.table.setItem(rowPoint, 2, QTableWidgetItem(email))
self.table.setItem(rowPoint, 3, QTableWidgetItem(department))
self.table.setItem(rowPoint, 4, QTableWidgetItem(mobilecyp))
self.table.setItem(rowPoint, 5, QTableWidgetItem(mobileother))
def show_member_at_tableWidget(self):
self.table = self.ui.tableWidget
self.table.setRowCount(0)
self.table.setColumnCount(6)
self.table.setHorizontalHeaderLabels(
["Name", "Surname", "E-mail", "Department", "Mobile No Cyp", "Mobile No Other"])
query = {} # means no condition. so it will get everyone.
result = self.db.query_result_multi("Member", query)
if not result:
return False
for member in result:
self.add_member_to_tableWidget(member["name"]["first"], member["name"]["last"], member["email"],
member["department"], member["mobileNo"]["cyp"], member["mobileNo"]["other"])
# Delete chosen member on tableView
def delete_member(self):
indexes = self.ui.tableWidget.selectedIndexes()
for i in indexes:
# Following line get the selected row's email address
email_to_delete = str(self.ui.tableWidget.item(i.row(), 2).text())
# Delete member who has the email
self.db.delete_member_by_email(email_to_delete)
self.show_member_at_tableWidget()
# Clear inside the line edits in registration form
def clear_form(self):
self.ui.lne_email.setText("")
self.ui.lne_firstName.setText("")
self.ui.lne_lastName.setText("")
self.ui.lne_mobileCyp.setText("")
self.ui.lne_mobileOther.setText("")
self.ui.comboBox_department.setCurrentIndex(0)
# Export members in the database as csv
def export_cvs(self):
try:
file = open("freg_export.cvs", "w+")
csv_file = csv.writer(file)
data = [self.ui.lne_firstName.text(), self.ui.lne_lastName.text()]
csv_file.writerow(data)
finally:
file.close()
# Import members from a csv file
def import_cvs(self):
pass # TODO
def create_desktop_entry(self):
fd = open("~/local/share/applications/freg.desktop", "w")
fd.write("[Desktop Entry]\n")
fd.write("Version=1.0 \n")
fd.write("Type=Application\n")
fd.write("Name=Freg\n")
fd.write("Exec=python3 ~/.faunus/frec/frec.py\n")
fd.write("Icon=~/.faunus/frec/data/icon/appicon.png\n")
fd.write("Comment=Small member registration system\n")
fd.write("Terminal=false\n")
fd.close()
def arrange_for_cvs(self):
dbb = self.db.query_result_multi("Member", {})
arranged_str = ""
for member in dbb:
one = member["name"]["first"] + "," + member["name"]["last"] + "," + member["email"]
arranged_str += one + "\n"
return arranged_str
from database import MongoDatabase
from config import MoR
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# location of target tweets
location_tweets = os.path.join('files', 'target_tweets')
# process tweets for each mode of research
for mode in MoR:
logging.info('Processing mode of research: {}'.format(mode))
# read tweets files
F = read_directory(os.path.join(location_tweets, mode))
# tracker to keep track of processed tweet ids
tweet_tracker = set(['{}{}'.format(x['tweet_type'], x['id'])
for x in db.read_collection(collection='raw_tweets')])
import os
import database
from database import MongoDatabase
from config import PitConfig
from routes import PitRoutes
reload(sys)
sys.setdefaultencoding('utf8')
# ---------------------------------------------------------------------------- #
# Create Flask web application object.
pit_app = flask.Flask(__name__, static_url_path="")
pit_app.secret_key = "roflmao"
# Set up and verify database connection.
database = MongoDatabase(PitConfig)
database.connect()
# Set up API endpoints.
routes = PitRoutes(pit_app, database, PitConfig, flask.make_response,
flask.render_template)
if __name__ == "__main__":
# Start application.
debug = PitConfig['web']['debug'] == 'True'
port = int(PitConfig['web']['port'])
if debug:
pit_app.run(debug=debug, host="0.0.0.0", port=port)
else:
context = ('fullchain.pem', 'privkey.pem')
pit_app.run(debug=debug,
class Core:
def __init__(self):
self.components = []
self.database = MongoDatabase("localhost")
self.load_components()
def load_components(self):
log.info("Loading components...")
for record in self.database.get_all_records("components"):
try:
self.create_component(record, save=False)
except ValueError:
pass
def open_command_tunnel(self, content):
matches = commands.recognise_command(content)
if len(matches) == 0:
return None
tunnel = commands.MessageTunnel()
context = commands.CommandContext(matches[0], tunnel)
context.initial_message = content
context.core = self
context.run_command()
return tunnel
def create_component(self, data, save=True):
component = components.create_component(data, core=self)
if save:
component.save(self.database)
self.components.append(component)
def find_component_by_filter(self, record_filter):
return self.find_components_by_filter(record_filter)[0]
def find_components_by_filter(self, record_filter):
out = []
for record in self.database.find_all("components", record_filter):
for component in self.components:
if component.uuid == record["uuid"]:
out.append(component)
return out
def save_all(self):
log.info("Saving everything...")
for component in self.components:
component.save(self.database)
def cleanup_all(self):
log.info("Cleaning everything...")
for component in self.components:
component.cleanup()
###### Install Pillow, not PIL ###### from constants import Constants import math import glob, os import Tkinter from Tkinter import * import ttk from PIL import Image, ImageTk import zipfile import shutil from database import MongoDatabase from inner_folder.ui_class import UI if __name__=='__main__': ###### Begin Script ###### ###### Connection to Mongo DB ###### MongoDatabase.initlialize() # process = MongoDatabase.get_mongo_process() ui = UI(MongoDatabase.db) ui.checkPath() ui.root.mainloop()
return
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# name of collection to store all the training tweets to
db_collection = 'training_tweets'
# location to save machine learning classification models to
model_save_location = os.path.join('files', 'ml_models2')
# get all the training tweet documents
D = db.read_collection(collection=db_collection)
# get values from list and assign to X and Y
X, Y = zip(*[(x['text'], str(x['label'])) for x in D])
# define pipeline options
pipeline_setup = get_pipeline_setup()
from database import MongoDatabase
import joblib
"""
Script starts here
"""
if __name__ == "__main__":
# create logging to console
set_logger()
# verbose
logging.info('Start: {} '.format(__file__))
# create database connection
db = MongoDatabase()
# load classifier
clf = joblib.load(os.path.join('files', 'ml_models3', 'LinearSVC.pkl'))
# read labels for target tweets that have been manually labeled and convert to dictionary with key = tweet ID and value = label
true_labels = {
d['tweet_id']: d['label']
for d in db.read_collection(collection='sanders_tweets_raw')
}
# load tweets for which we want to infer the sentiment label
D = db.read_collection(collection='target_tweets_UNCW')
# create empty numpy array so we can retrieve labels later on somewhat faster
labels = np.zeros((D.count(), 3), dtype=np.int64)
