def classify_files():
"""
The POST method allows API clients to use their model(s)
The GET method allows API clients to get a list of available models.
:return:
"""
if request.method == "POST":
# Log that we got a request
app.logger.error("Got a file POST request")
# Extract and validate credentials.
app_id, app_key = check_credentials()
try:
submitted_file = request.files.get('file')
ctype = submitted_file.content_type
if ctype not in ACCEPTED_MIMETYPES:
app.logger.error("Unsupported Media Type: %s", ctype)
raise UnsupportedMediaType
data = json.loads(request.data)
if 'textcol' in data:
textcol = data['textcol']
validate_input_file(submitted_file, ctype)
header, rows = sh.get_spreadsheet_rows(submitted_file,
textcol,
dedupe=True)
model = data['model']
ideanet = IDEANETS[model]
predictions = ideanet.classify(rows)
data = {"status": "success", "data": predictions}
except:
data = {"status": "failure", "data": {}}
def classify_files():
"""
The POST method allows API clients to use their model(s)
The GET method allows API clients to get a list of available models.
:return:
"""
if request.method == "POST":
# Log that we got a request
app.logger.error("Got a file POST request")
# Extract and validate credentials.
app_id, app_key = check_credentials()
try:
submitted_file = request.files.get('file')
ctype = submitted_file.content_type
if ctype not in ACCEPTED_MIMETYPES:
app.logger.error("Unsupported Media Type: %s", ctype)
raise UnsupportedMediaType
data = json.loads(request.data)
if 'textcol' in data:
textcol = data['textcol']
validate_input_file(submitted_file, ctype)
header, rows = sh.get_spreadsheet_rows(submitted_file,textcol,dedupe=True)
model = data['model']
ideanet = IDEANETS[model]
predictions = ideanet.classify(rows)
data = {"status":"success",
"data":predictions}
except:
data = {"status":"failure",
"data":{}}
def preprocess(self):
# For Synapsify Core output, the comments are in the first column
# and the sentiment is in the 6th column
if self._raw_rows==None:
header, rows = sh.get_spreadsheet_rows(os.path.join(self._data_directory, self._data_file), self._text_col, dedupe=self._dedupe)
else:
header = []
rows = self._raw_rows
sentences = [str(S[self._text_col]) for S in rows]
classes = [str(S[self._label_col]) for S in rows]
self._DICTIONARY = self._build_dict(sentences)
if self._class_type=="Sentiment":
# Grab the indices for the Core sentiment
self._train_xx = self._get_rand_sentiment_indices(classes, self._train_size,[])
self._test_xx = self._get_rand_sentiment_indices(classes, self._test_size, self._train_xx)
self._trXX = self._get_sentiment_indices([rows[r] for r in self._train_xx], self._label_col, [])
self._teXX = self._get_sentiment_indices([rows[r] for r in self._test_xx], self._label_col, self._train_xx)
# Munge training and test sets for the classes provided
train = self._munge_class_freqs(sentences,[self._trXX['neg'],self._trXX['pos']])
test = self._munge_class_freqs(sentences,[self._teXX['neg'],self._teXX['pos']])
else:
self._get_rand_indices(classes)
# max_sentence_length(self.train_x_sets)
# Munge training and test sets for the classes provided
train_classes = []
train_unique_classes = np.unique([int(key) for key in self._trXX.keys()]).tolist()
for trc in train_unique_classes:
train_classes.append(self._trXX[str(trc)])
test_classes = []
test_unique_classes = np.unique([int(key) for key in self._teXX.keys()]).tolist()
for trc in test_unique_classes:
test_classes.append(self._teXX[str(trc)])
train = self._munge_class_freqs(sentences, train_classes)
test = self._munge_class_freqs(sentences, test_classes)
# Split training into a validation set per the model parameter
valid_set_x, valid_set_y, train_set_x, train_set_y = self._split_train_w_valid_set( train)
# Remove unknown words
train_set_x = self._remove_unk(train_set_x)
valid_set_x = self._remove_unk(valid_set_x)
test_set_x = self._remove_unk(test[0])
self.train_set = (train_set_x, train_set_y)
self.valid_set = (valid_set_x, valid_set_y)
self.test_set = (test_set_x, test[1])
# TVT = {
# 'train': (train_set_x, train_set_y),
# 'valid': (valid_set_x, valid_set_y),
# 'test': (test_set_x,test[1])
# }
return self
# if __name__ == '__main__':
# directory = sys.argv[1]
# filename = sys.argv[2]
# textcol = 0
# if len(sys.argv)>2: textcol = int(sys.argv[3])
# sentcol = 5
# if len(sys.argv)>3: sentcol = int(sys.argv[4])
#
# ### I added the following part of codes here
# train_size = int(sys.argv[5])
# test_size = int(sys.argv[6])
# ### The main function should take six parameters instead of four
# ### The original code is:
# ### main(directory, filename, textcol, sentcol)
# load(directory, filename, textcol, sentcol, train_size, test_size)
