def get_topics_table_by_id(self, topic_id):
"""
Transform topics from gensim LDA model to table.
:param lda: gensim LDA model.
:return: `Orange.data.table.Table`.
"""
topics = self.lda.show_topics(num_topics=-1,
num_words=MAX_WORDS,
formatted=False)
if topic_id >= len(topics):
raise ValueError("Too large topic ID.")
num_words = max([len(it) for it in topics])
data = np.zeros((num_words, 2), dtype=object)
data[:, 0] = [item[1] for item in topics[topic_id]]
data[:, 1] = [item[0] for item in topics[topic_id]]
metas = [
StringVariable(self.topic_names[topic_id]),
ContinuousVariable("Topic{}_weights".format(topic_id + 1))
]
metas[-1]._out_format = '%.2e'
domain = Domain([], metas=metas)
t = Topics.from_numpy(domain, X=np.zeros((num_words, 0)), metas=data)
t.W = data[:, 1]
return t
def __init__(self, parent=None, signalManager=None, settings=None):
super().__init__(parent, signalManager, settings)
self.data = None
self.undoStack = QtGui.QUndoStack(self)
self.plot = PaintDataPlot(self.mainArea, "Painted Plot", widget=self)
self.classValuesModel = ColoredListModel(
["Class-1", "Class-2"],
self,
self.plot,
flags=QtCore.Qt.ItemIsSelectable | QtCore.Qt.ItemIsEnabled
| QtCore.Qt.ItemIsEditable)
self.classValuesModel.dataChanged.connect(self.classNameChange)
self.data = Table(
Domain([
ContinuousVariable(self.attr1),
ContinuousVariable(self.attr2)
], DiscreteVariable("Class", values=self.classValuesModel)))
self.toolsStackCache = {}
self.initUI()
self.initPlot()
self.updatePlot()
def test_corpus_not_eq(self):
c = Corpus.from_file('bookexcerpts')
n_doc = c.X.shape[0]
c2 = Corpus(c.X, c.Y, c.metas, c.domain, [])
self.assertNotEqual(c, c2)
c2 = Corpus(np.ones((n_doc, 1)), c.Y, c.metas, c.domain,
c.text_features)
self.assertNotEqual(c, c2)
c2 = Corpus(c.X, np.ones((n_doc, 1)), c.metas, c.domain,
c.text_features)
self.assertNotEqual(c, c2)
broken_metas = np.copy(c.metas)
broken_metas[0, 0] = ''
c2 = Corpus(c.X, c.Y, broken_metas, c.domain, c.text_features)
self.assertNotEqual(c, c2)
new_meta = [StringVariable('text2')]
broken_domain = Domain(c.domain.attributes, c.domain.class_var,
new_meta)
c2 = Corpus(c.X, c.Y, c.metas, broken_domain, new_meta)
self.assertNotEqual(c, c2)
def get_topics_table_by_id(self, topic_id):
"""
Transform topics from gensim LDA model to table.
:param lda: gensim LDA model.
:return: `Orange.data.table.Table`.
"""
words = self._topics_words(MAX_WORDS)
weights = self._topics_weights(MAX_WORDS)
if topic_id >= len(words):
raise ValueError("Too large topic ID.")
num_words = len(words[topic_id])
data = np.zeros((num_words, 2), dtype=object)
data[:, 0] = words[topic_id]
data[:, 1] = weights[topic_id]
metas = [
StringVariable(self.topic_names[topic_id]),
ContinuousVariable("Topic{}_weights".format(topic_id + 1))
]
metas[-1]._out_format = '%.2e'
domain = Domain([], metas=metas)
t = Topics.from_numpy(domain, X=np.zeros((num_words, 0)), metas=data)
t.W = data[:, 1]
return t
def sendData(self):
data = self.data
values = set([str(ex.get_class()) for ex in data])
if len(values) == 1:
# Remove the useless class variable.
domain = Domain(data.domain.attributes)
data = Table(domain, data)
self.send("Data", data)
def k_net_table(self, v_list):
values = [
ContinuousVariable(name='Distance', number_of_decimals=0),
ContinuousVariable(name='Knet score'),
]
domain = Domain(values)
return Table(domain, [[n, v] for n, v in enumerate(v_list)])
def k_node_table(self, v_list):
dist_values = {l[0] for l in v_list}
values = [
DiscreteVariable(name='Protein', values=dist_values),
ContinuousVariable(name='Knode score'),
ContinuousVariable(name='Distance', number_of_decimals=0),
]
domain = Domain(values)
return Table(domain, v_list)
def _create_table(words, scores: Mapping[str, np.array]) -> Table:
"""
Create an Orange table from the word scores
:param words: list of words
:param scores: mapping of {label: score_array}. Use ordereddict to preserve column order
:return: a Table object
"""
values = list(scores.values())
order = (-values[0]).argsort()
data = np.column_stack(values)[order]
words = np.array(words).reshape(len(words), 1)[order]
domain = Domain([ContinuousVariable(label) for label in scores],
metas=[StringVariable("term")])
return Table(domain, data, metas=words)
def insert_topics_into_corpus(self, corp_in):
"""
Insert topical representation into corpus.
:param corp_in: Corpus into whic we want to insert topical representations
:return: `Orange.data.table.Table`
"""
matrix = matutils.corpus2dense(self.corpus,
num_terms=self.num_topics).T
# Generate the new table.
attr = [ContinuousVariable(n) for n in self.topic_names]
domain = Domain(attr,
corp_in.domain.class_vars,
metas=corp_in.domain.metas)
return Table.from_numpy(domain,
matrix,
Y=corp_in._Y,
metas=corp_in.metas)
def redo(self):
self.label = self.classValuesModel.pop(self.index)
examples = [
ex for ex in self.data if str(ex.get_class()) != self.label
]
newdomain = Domain([
ContinuousVariable(self.widget.attr1),
ContinuousVariable(self.widget.attr2)
], DiscreteVariable("Class", values=self.classValuesModel))
newdata = Table(newdomain)
for ex in examples:
if str(ex.get_class()) != self.label and \
str(ex.get_class()) in self.classValuesModel:
newdata.append(
Instance(newdomain,
[float(ex[a]) for a in ex.domain.attributes] +
[str(ex.get_class())]))
self.widget.data = newdata
self.widget.updatePlot()
self.widget.updateCursor()
def redo(self):
newdomain = Domain([
ContinuousVariable(self.widget.attr1),
ContinuousVariable(self.widget.attr2)
], DiscreteVariable("Class", values=self.classValuesModel))
newdata = Table(newdomain)
for ex in self.data:
print(ex.get_class())
if str(ex.get_class()) not in self.classValuesModel:
self.oldLabelName = str(ex.get_class())
instance = Instance(
newdomain,
[float(ex[a])
for a in ex.domain.attributes] + [self.changedLabel])
newdata.append(instance)
else:
newdata.append(
Instance(newdomain,
[float(ex[a]) for a in ex.domain.attributes] +
[str(ex.get_class())]))
self.widget.data = newdata
self.widget.updatePlot()
def redo(self):
self.classValuesModel.append(self.newClassLabel)
newdomain = Domain([
ContinuousVariable(self.widget.attr1),
ContinuousVariable(self.widget.attr2)
], DiscreteVariable("Class", values=self.classValuesModel))
newdata = Table(newdomain)
instances = [
Instance(newdomain,
[float(ex[a])
for a in ex.domain.attributes] + [str(ex.get_class())])
for ex in self.data
]
newdata.extend(instances)
self.widget.data = newdata
self.widget.removeClassLabel.setEnabled(len(self.classValuesModel) > 1)
newindex = self.classValuesModel.index(len(self.classValuesModel) - 1)
self.widget.classValuesView.selectionModel().select(
newindex, QtGui.QItemSelectionModel.ClearAndSelect)
self.widget.updatePlot()
self.widget.updateCursor()