def getTrainingData():
"""
Gets all the available training data.
Required params:\n
-action = 'getData'\n
-stoken\n
:return: Returns the training data for that user and 200.
"""
request.get_data()
if request.json is None:
return jsonify({
'status': {
'error':
True,
'message':
"Incorrect header type. Header type should be application/json."
},
'data': None
}), 400
chkMissingData, data = getRequiredParameters(request, stoken='', action='')
if not chkMissingData:
return jsonify({
'status': {
'message': 'Missing required parameter \'' + data + '\'',
'error': True
},
'data': None
}), 400
con = Conn()
userID = con.checkToken(data["stoken"])
if userID == -1:
return jsonify(INVALID_SESSION_TOKEN), 403
if data['action'] != 'getData':
return jsonify({
'status': {
'message':
'Wrong action. This endpoint is only for retrieving training data.',
'error': True
},
'data': None
}), 403
modelID, modelName = con.getModelInfo(str(userID))
availableTraining = con.getAvailableTraining(modelID)
formattedTraining = list()
for item in availableTraining:
trainingDict = dict()
trainingDict["id"] = item[0]
trainingDict["rawText"] = item[1]
trainingDict["sentiment"] = item[3]
trainingDict["tags"] = item[4]
trainingDict["dateTrained"] = item[5]
formattedTraining.append(trainingDict)
return jsonify({
'status': RETURN_SUCCESS_STATUS,
'data': {
'availableTraining': formattedTraining
}
}), 200
def processAsk():
"""
Gets a prediction from the model.
Required params:\n
-action = 'ask'\n
-stoken\n
-datatype => text or blob\n
-data => raw text or base64 blob\n
-tags => list of tags\n
:return: The predicted sentiment, related tags and the original submitted text and 201
"""
request.get_data()
if request.json is None:
return jsonify({
'status': {
'error':
True,
'message':
"Incorrect header type. Header type should be application/json."
},
'data': None
}), 400
data = request.json
noMissingData, data = getRequiredParameters(request,
stoken='',
action='',
datatype='',
data='')
if not noMissingData:
return jsonify({
'status': {
'message': 'Missing required parameter \'' + data + '\'',
'error': True
},
'data': None
}), 400
con = Conn()
userID = con.checkToken(data['stoken'])
if userID == -1:
return jsonify(INVALID_SESSION_TOKEN), 403
if data['action'] != "ask":
return jsonify({
'status': {
'message': 'Wrong action. This endpoint is only for teach.',
'error': True
},
'data': None
}), 403
if data['datatype'] != 'text' and data['datatype'] != 'blob':
return jsonify({
'status': {
'message':
'Wrong datatype. This endpoint only accepts raw text as \'text\' and pdf or txt files as \'blob\'.',
'error': True
},
'data': None
}), 403
if data['datatype'] == 'text':
txtdata = str(request.json['data'])
# remove escape characters, if any
txtdata = txtdata.replace("\\", "")
if con.getModelInfo(str(userID)) == [None, None]:
return jsonify({
'status': {
'message': "No model created yet.",
'error': True
},
'data': None
}), 400
modelID, modelName = con.getModelInfo(str(userID))
availTraining = con.getAvailableTraining(modelID)
ifPrevKnowSimilar = False
for eachTraining in availTraining:
if getCosineSimilarity(
txtdata,
eachTraining[1]) >= PREVIOUS_KNOWLEDGE_SIMILARITY_RATE:
ifPrevKnowSimilar = True
break
if not ifPrevKnowSimilar:
return jsonify({
'status': RETURN_SUCCESS_STATUS,
'data': {
'predictedSentiment': "Not available",
'suggested': "Not available",
'text': txtdata
}
}), 200
model = Classifier(makeNewModel=False, modelName=modelName)
predSentiment, retTags = model.predict(txtdata)
return jsonify({
'status': RETURN_SUCCESS_STATUS,
'data': {
'predictedSentiment': predSentiment,
'suggested': retTags,
'text': txtdata
}
}), 200
elif data['datatype'] == 'blob':
if type(data['data']) != list and type(data['data']) != str:
return jsonify({
'status': {
'error':
True,
'message':
'Data in the \'data\' field should be sent as a string or a list of items.'
},
'data': None
}), 400
if type(data['data']) == list: # if multiple files
modelID, modelName = con.getModelInfo(str(userID))
model = Classifier(makeNewModel=False, modelName=modelName)
predSentiment = list()
retTags = list()
for item in data['data']:
filename = str(uuid.uuid4()) # generate temp file name
status, filepath = getFileBase64(item, filename)
# in case file received is not a text-based (.pdf/.txt) file
if not filepath.endswith('pdf') and not filepath.endswith(
'txt'):
os.remove(filepath)
return jsonify({
'status': {
'error': True,
'message': "Wrong format for data."
},
'data': None
}), 400
if not status:
return jsonify({
'status': {
'error': False,
'message': ERRORFILEEXTEN
}
}), 400
textData = readTextFileContents(filepath)
sent, tag = model.predict(textData)
predSentiment.append(sent)
retTags.append(tag)
os.remove(filepath)
return jsonify({
'status': RETURN_SUCCESS_STATUS,
'data': {
'predictedSentiment': predSentiment,
'suggested': retTags,
'text': data['data']
}
}), 200
elif type(data['data']) == str: # if single file
modelID, modelName = con.getModelInfo(str(userID))
model = Classifier(makeNewModel=False, modelName=modelName)
filename = str(uuid.uuid4()) # generate temp file name
status, filepath = getFileBase64(data['data'], filename)
# in case file received is not a text-based (.pdf/.txt) file
if not filepath.endswith('pdf') and not filepath.endswith('txt'):
os.remove(filepath)
return jsonify({
'status': {
'error': True,
'message': "Wrong format for data."
},
'data': None
}), 400
textData = readTextFileContents(filepath)
os.remove(filepath)
predSentiment, retTags = model.predict(textData)
return jsonify({
'status': RETURN_SUCCESS_STATUS,
'data': {
'predictedSentiment': predSentiment,
'suggested': retTags,
'text': data['text']
}
}), 200
def retrain():
"""
Used to retrain the model.
Required params:\n
-action = 'retrain'\n
-stoken => the session token \n
-data => a list of trainingID that will be deleted\n
:return: A success message and 200.
"""
request.get_data()
if request.json is None:
return jsonify({
'status': {
'error':
True,
'message':
"Incorrect header type. Header type should be application/json."
},
'data': None
}), 400
chkMissingData, data = getRequiredParameters(request,
stoken='',
action='',
data='')
if not chkMissingData:
return jsonify({
'status': {
'message': 'Missing required parameter \'' + data + '\'',
'error': True
},
'data': None
}), 400
con = Conn()
userID = con.checkToken(data["stoken"])
if userID == -1:
return jsonify(INVALID_SESSION_TOKEN), 403
if data['action'] != 'retrain':
return jsonify({
'status': {
'message':
'Wrong action. This endpoint is only for retraining.',
'error': True
},
'data': None
}), 403
if type(data['data']) != list:
return jsonify({
'status': {
'message':
'Bad data. Data in the \'data\' field must be sent as list.',
'error': True
},
'data': None
}), 400
# get old model info
oldModelID, oldModelName = con.getModelInfo(str(userID))
if not con.removeTraining(oldModelID, data['data']):
return jsonify({
'status': {
'message': 'Internal server encountered. Please try again.',
'error': True
},
'data': None
}), 500
availableTraining = con.getAvailableTraining(oldModelID)
# make a new model
model = Classifier(makeNewModel=True)
for eachData in availableTraining:
model.train(text=eachData[2],
sentiment=eachData[3],
tags=eachData[4],
fromDB=True)
# update db with new model info
if not con.updateModelInfo(oldModelID, modelName=model.modelName):
return jsonify({
'status': {
'message': 'Internal server error when updating the resource.',
'error': True
},
'data': None
}), 500
if not oldModelName.endswith(PICKLE_FILE_EXTENSION):
oldModelName += PICKLE_FILE_EXTENSION
os.remove(os.path.join(os.path.dirname(__file__),
'Models/' + oldModelName))
return jsonify({
'status': {
'message': 'Model successfully retrained.',
'error': False
},
'data': None
}), 200