class OWTweetProfiler(OWWidget):
name = "Tweet Profiler"
description = "Detect Ekman's, Plutchik's or Profile of Mood States's " \
"emotions in tweets."
icon = "icons/TweetProfiler.svg"
priority = 46
inputs = [(IO.CORPUS, Corpus, 'set_corpus')]
outputs = [(IO.CORPUS, Corpus)]
want_main_area = False
resizing_enabled = False
token = Setting('')
model_name = Setting('')
output_mode = Setting('')
tweet_attr = Setting(0)
auto_commit = Setting(True)
class Error(OWWidget.Error):
server_down = Msg('Our servers are not responding. '
'Please try again later.')
invalid_token = Msg('This token is invalid')
no_credit = Msg('Too little credits for this data set')
def __init__(self):
super().__init__()
self.corpus = None
self.strings_attrs = []
self.profiler = TweetProfiler(
token=self.token,
on_server_down=self.Error.server_down,
on_invalid_token=self.Error.invalid_token,
on_too_little_credit=self.Error.no_credit,
)
# Info box
self.n_documents = ''
self.credit = 0
box = gui.widgetBox(self.controlArea, "Info")
gui.label(box, self, 'Documents: %(n_documents)s')
gui.label(box, self, 'Credits: %(credit)s')
# Settings
self.controlArea.layout().addWidget(self.generate_grid_layout())
# Server token
box = gui.vBox(self.controlArea, 'Server Token')
gui.lineEdit(box, self, 'token', callback=self.token_changed,
controlWidth=300)
gui.button(box, self, 'Get Token', callback=self.get_new_token)
# Auto commit
buttons_layout = QtGui.QHBoxLayout()
buttons_layout.addWidget(self.report_button)
buttons_layout.addSpacing(15)
buttons_layout.addWidget(
gui.auto_commit(None, self, 'auto_commit', 'Commit', box=False)
)
self.controlArea.layout().addLayout(buttons_layout)
self.refresh_token_info()
def generate_grid_layout(self):
box = QtGui.QGroupBox(title='Options')
layout = QGridLayout()
layout.setSpacing(10)
row = 0
self.tweet_attr_combo = gui.comboBox(None, self, 'tweet_attr',
callback=self.apply)
layout.addWidget(QLabel('Attribute:'))
layout.addWidget(self.tweet_attr_combo, row, 1)
row += 1
self.model_name_combo = gui.comboBox(None, self, 'model_name',
items=self.profiler.model_names,
sendSelectedValue=True,
callback=self.apply)
if self.profiler.model_names:
self.model_name = self.profiler.model_names[0] # select 0th
layout.addWidget(QLabel('Emotions:'))
layout.addWidget(self.model_name_combo, row, 1)
row += 1
self.output_mode_combo = gui.comboBox(None, self, 'output_mode',
items=self.profiler.output_modes,
sendSelectedValue=True,
callback=self.apply)
if self.profiler.output_modes:
self.output_mode = self.profiler.output_modes[0] # select 0th
layout.addWidget(QLabel('Output:'))
layout.addWidget(self.output_mode_combo, row, 1)
box.setLayout(layout)
return box
def set_corpus(self, corpus):
self.corpus = corpus
if corpus is not None:
self.strings_attrs = [a for a in self.corpus.domain.metas
if isinstance(a, StringVariable)]
self.tweet_attr_combo.setModel(VariableListModel(self.strings_attrs))
self.tweet_attr_combo.currentIndexChanged.emit(self.tweet_attr)
# select the first feature from 'text_features' if present
ind = [self.strings_attrs.index(tf)
for tf in corpus.text_features
if tf in self.strings_attrs]
if ind:
self.tweet_attr = ind[0]
self.n_documents = len(corpus)
self.commit()
def apply(self):
self.commit()
def commit(self):
self.Error.clear()
if self.corpus is not None:
with self.progressBar(iterations=len(self.corpus)) as pb:
out = self.profiler.transform(
self.corpus, self.strings_attrs[self.tweet_attr],
self.model_name, self.output_mode,
on_advance=pb.advance)
self.send(IO.CORPUS, out)
else:
self.send(IO.CORPUS, None)
self.refresh_token_info()
def get_new_token(self):
self.Warning.clear()
self.profiler.new_token()
self.token = self.profiler.token
self.refresh_token_info()
self.commit()
def token_changed(self):
self.profiler.token = self.token
self.refresh_token_info()
self.commit()
def refresh_token_info(self):
self.credit = str(self.profiler.get_credit())
def send_report(self):
self.report_items([
('Documents', self.n_documents),
('Attribute', self.strings_attrs[self.tweet_attr]
if len(self.strings_attrs) > self.tweet_attr else ''),
('Emotions', self.model_name),
('Output', self.output_mode),
])
class TestTweetProfiler(unittest.TestCase):
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def setUp(self):
self.data = Corpus.from_file('Election-2016-Tweets.tab')[:100]
self.profiler = TweetProfiler()
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_get_server_address(self):
server = self.profiler.get_server_address()
self.assertTrue(server.startswith('http'))
@patch(SERVER_CALL, MockServerCall())
def test_get_configuration(self):
self.assertEqual(self.profiler.model_names, MODELS)
self.assertEqual(self.profiler.output_modes, MODES)
@patch(SERVER_CALL, MockServerCall())
def test_get_token(self):
self.assertIsNone(self.profiler.token)
self.profiler.new_token()
self.assertEqual(self.profiler.token, TOKEN)
@patch(SERVER_CALL, MockServerCall())
def test_is_token_valid(self):
self.assertEqual(self.profiler.is_token_valid(), VALID)
@patch(SERVER_CALL, MockServerCall())
def test_get_credit(self):
self.assertEqual(self.profiler.get_credit(), COINS)
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_transform_embeddings(self):
advance_call_mock = Mock()
text_var = self.data.domain.metas[1]
corp = self.profiler.transform(self.data,
text_var,
'model-mc',
'Embeddings',
on_advance=advance_call_mock)
self.assertIsInstance(corp, Corpus)
self.assertEqual(advance_call_mock.call_count, 2)
self.assertEqual(
len(corp.domain.attributes) - len(self.data.domain.attributes),
EMBEDDINGS_NUM)
self.assertEqual(corp.X.shape[1] - self.data.X.shape[1],
EMBEDDINGS_NUM)
self.assertEqual(np.sum(corp.X[:, -EMBEDDINGS_NUM]), 0)
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_transform_probabilities_and_ml_classes(self):
text_var = self.data.domain.metas[1]
for mode in ['Probabilities', 'Classes']:
corp = self.profiler.transform(self.data, text_var, 'model-ml',
mode)
self.assertIsInstance(corp, Corpus)
self.assertEqual(
len(corp.domain.attributes) - len(self.data.domain.attributes),
len(PROFILE_CLASSES))
self.assertEqual(corp.X.shape[1] - self.data.X.shape[1],
len(PROFILE_CLASSES))
self.assertEqual(np.sum(corp.X[:, -len(PROFILE_CLASSES)]), 0)
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_transform_mc_classes(self):
text_var = self.data.domain.metas[1]
corp = self.profiler.transform(self.data, text_var, 'model-mc',
'Classes')
self.assertIsInstance(corp, Corpus)
self.assertEqual(
len(corp.domain.attributes) - len(self.data.domain.attributes), 1)
self.assertEqual(corp.X.shape[1] - self.data.X.shape[1], 1)
self.assertEqual(np.sum(corp.X[:, -1]), 0)
@patch(CHECK_ALIVE, Mock(return_value=False))
def test_transform_probabilities(self):
text_var = self.data.domain.metas[1]
corp = self.profiler.transform(self.data, text_var, MODELS[0],
'Classes')
self.assertIsNone(corp)
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_transform_empty_corpus(self):
text_var = self.data.domain.metas[1]
corp = self.profiler.transform(self.data[:0], text_var, 'model-mc',
'Classes')
self.assertIsNone(corp)
@patch(SERVER_CALL, MockServerCallSuddenlyUnavailable())
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_transform_server_call_error(self):
text_var = self.data.domain.metas[1]
corp = self.profiler.transform(self.data, text_var, 'model-mc',
'Classes')
self.assertIsNone(corp)
class TestTweetProfiler(unittest.TestCase):
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def setUp(self):
self.data = Corpus.from_file('Election-2016-Tweets.tab')[:100]
self.profiler = TweetProfiler()
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_get_server_address(self):
server = self.profiler.get_server_address()
self.assertTrue(server.startswith('http'))
@patch(SERVER_CALL, MockServerCall())
def test_get_configuration(self):
self.assertEqual(self.profiler.model_names, MODELS)
self.assertEqual(self.profiler.output_modes, MODES)
@patch(SERVER_CALL, MockServerCall())
def test_get_token(self):
self.assertIsNone(self.profiler.token)
self.profiler.new_token()
self.assertEqual(self.profiler.token, TOKEN)
@patch(SERVER_CALL, MockServerCall())
def test_is_token_valid(self):
self.assertEqual(self.profiler.is_token_valid(), VALID)
@patch(SERVER_CALL, MockServerCall())
def test_get_credit(self):
self.assertEqual(self.profiler.get_credit(), COINS)
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_transform_embeddings(self):
advance_call_mock = Mock()
text_var = self.data.domain.metas[1]
corp = self.profiler.transform(self.data, text_var,
'model-mc', 'Embeddings',
on_advance=advance_call_mock)
self.assertIsInstance(corp, Corpus)
self.assertEqual(advance_call_mock.call_count, 2)
self.assertEqual(len(corp.domain.attributes) -
len(self.data.domain.attributes),
EMBEDDINGS_NUM)
self.assertEqual(corp.X.shape[1] - self.data.X.shape[1],
EMBEDDINGS_NUM)
self.assertEqual(np.sum(corp.X[:, -EMBEDDINGS_NUM]), 0)
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_transform_probabilities_and_ml_classes(self):
text_var = self.data.domain.metas[1]
for mode in ['Probabilities', 'Classes']:
corp = self.profiler.transform(self.data, text_var,
'model-ml', mode)
self.assertIsInstance(corp, Corpus)
self.assertEqual(len(corp.domain.attributes) -
len(self.data.domain.attributes),
len(PROFILE_CLASSES))
self.assertEqual(corp.X.shape[1] - self.data.X.shape[1],
len(PROFILE_CLASSES))
self.assertEqual(np.sum(corp.X[:, -len(PROFILE_CLASSES)]), 0)
@patch(SERVER_CALL, MockServerCall())
@patch(CHECK_ALIVE, Mock(return_value=True))
def test_transform_mc_classes(self):
text_var = self.data.domain.metas[1]
corp = self.profiler.transform(self.data, text_var,
'model-mc', 'Classes')
self.assertIsInstance(corp, Corpus)
self.assertEqual(len(corp.domain.attributes) -
len(self.data.domain.attributes),
1)
self.assertEqual(corp.X.shape[1] - self.data.X.shape[1], 1)
self.assertEqual(np.sum(corp.X[:, -1]), 0)
@patch(CHECK_ALIVE, Mock(return_value=False))
def test_transform_probabilities(self):
text_var = self.data.domain.metas[1]
corp = self.profiler.transform(self.data, text_var,
MODELS[0], 'Classes')
self.assertIs(corp, self.data)
class OWTweetProfiler(OWWidget):
name = "Tweet Profiler"
description = "Detect Ekman's, Plutchik's or Profile of Mood States's " \
"emotions in tweets."
icon = "icons/TweetProfiler.svg"
priority = 46
inputs = [(IO.CORPUS, Corpus, 'set_corpus')]
outputs = [(IO.CORPUS, Corpus)]
want_main_area = False
resizing_enabled = False
token = Setting('')
model_name = Setting('')
output_mode = Setting('')
tweet_attr = Setting(0)
auto_commit = Setting(True)
class Error(OWWidget.Error):
server_down = Msg('Our servers are not responding. '
'Please try again later.')
invalid_token = Msg('This token is invalid')
no_credit = Msg('Too little credits for this data set')
def __init__(self):
super().__init__()
self.corpus = None
self.strings_attrs = []
self.profiler = TweetProfiler(
token=self.token,
on_server_down=self.Error.server_down,
on_invalid_token=self.Error.invalid_token,
on_too_little_credit=self.Error.no_credit,
)
# Info box
self.n_documents = ''
self.credit = 0
box = gui.widgetBox(self.controlArea, "Info")
gui.label(box, self, 'Documents: %(n_documents)s')
gui.label(box, self, 'Credits: %(credit)s')
# Settings
self.controlArea.layout().addWidget(self.generate_grid_layout())
# Server token
box = gui.vBox(self.controlArea, 'Server Token')
gui.lineEdit(box,
self,
'token',
callback=self.token_changed,
controlWidth=300)
gui.button(box, self, 'Get Token', callback=self.get_new_token)
# Auto commit
buttons_layout = QtGui.QHBoxLayout()
buttons_layout.addWidget(self.report_button)
buttons_layout.addSpacing(15)
buttons_layout.addWidget(
gui.auto_commit(None, self, 'auto_commit', 'Commit', box=False))
self.controlArea.layout().addLayout(buttons_layout)
self.refresh_token_info()
def generate_grid_layout(self):
box = QtGui.QGroupBox(title='Options')
layout = QGridLayout()
layout.setSpacing(10)
row = 0
self.tweet_attr_combo = gui.comboBox(None,
self,
'tweet_attr',
callback=self.apply)
layout.addWidget(QLabel('Attribute:'))
layout.addWidget(self.tweet_attr_combo, row, 1)
row += 1
self.model_name_combo = gui.comboBox(None,
self,
'model_name',
items=self.profiler.model_names,
sendSelectedValue=True,
callback=self.apply)
if self.profiler.model_names:
self.model_name = self.profiler.model_names[0] # select 0th
layout.addWidget(QLabel('Emotions:'))
layout.addWidget(self.model_name_combo, row, 1)
row += 1
self.output_mode_combo = gui.comboBox(None,
self,
'output_mode',
items=self.profiler.output_modes,
sendSelectedValue=True,
callback=self.apply)
if self.profiler.output_modes:
self.output_mode = self.profiler.output_modes[0] # select 0th
layout.addWidget(QLabel('Output:'))
layout.addWidget(self.output_mode_combo, row, 1)
box.setLayout(layout)
return box
def set_corpus(self, corpus):
self.corpus = corpus
if corpus is not None:
self.strings_attrs = [
a for a in self.corpus.domain.metas
if isinstance(a, StringVariable)
]
self.tweet_attr_combo.setModel(
VariableListModel(self.strings_attrs))
self.tweet_attr_combo.currentIndexChanged.emit(self.tweet_attr)
# select the first feature from 'text_features' if present
ind = [
self.strings_attrs.index(tf) for tf in corpus.text_features
if tf in self.strings_attrs
]
if ind:
self.tweet_attr = ind[0]
self.n_documents = len(corpus)
self.commit()
def apply(self):
self.commit()
def commit(self):
self.Error.clear()
if self.corpus is not None:
with self.progressBar(iterations=len(self.corpus)) as pb:
out = self.profiler.transform(
self.corpus,
self.strings_attrs[self.tweet_attr],
self.model_name,
self.output_mode,
on_advance=pb.advance)
self.send(IO.CORPUS, out)
else:
self.send(IO.CORPUS, None)
self.refresh_token_info()
def get_new_token(self):
self.Warning.clear()
self.profiler.new_token()
self.token = self.profiler.token
self.refresh_token_info()
self.commit()
def token_changed(self):
self.profiler.token = self.token
self.refresh_token_info()
self.commit()
def refresh_token_info(self):
self.credit = str(self.profiler.get_credit())
def send_report(self):
self.report_items([
('Documents', self.n_documents),
('Attribute', self.strings_attrs[self.tweet_attr]
if len(self.strings_attrs) > self.tweet_attr else ''),
('Emotions', self.model_name),
('Output', self.output_mode),
])
