def classify():
global stress
ERROR_THRESHOLD = 0.75
context = None
sentence = request.json['query']
# print(sentence)
sentiment = sentiment_analysis.sentiment_analyzer(sentence)
if "context" in request.json:
# print("Context present : ", request.json['context'])
context = request.json['context']
else:
print("Context not present!")
context = None
# entities = entity_extraction.named_entity_extraction(sentence)
# generate probabilities from the model
input_data = pd.DataFrame([bow(sentence, words)],
dtype=float,
index=['input'])
results = model.predict([input_data])[0]
# filter out predictions below a threshold
# print("Before Filter : ", results)
results = [[i, r] for i, r in enumerate(results) if r > ERROR_THRESHOLD]
# print("After Filter : ", results)
# sort by strength of probability
results.sort(key=lambda x: x[1], reverse=True)
return_list = []
output_context = None
# print("Results Length : ", len(results))
if len(results) == 0:
return jsonify({
"result": {
"fulfillment": {
"messages": [{
"type": 0,
"platform": "facebook",
"speech": random.choice(fallback_dict)
}]
}
}
})
# return_list.append({"query": sentence, "intent": "fallback", "response": random.choice(fallback_dict), "context": None, "probability": "0.00", "sentiment":sentiment})
else:
# print("Inference Exists")
for r in results:
if context != None:
classes[r[0]] = context
# print("Class Value : ", classes[r[0]])
for x_tend in intents['intents']:
if classes[r[0]] == x_tend['tag']:
# print("Entities Length : ", len(entities))
if x_tend['context'] == "":
output_context = None
# if entities is None:
# entities = None
# elif len(entities) == 0:
# entities = None
# print(random.choice(x_tend['responses']))
return jsonify({
"result": {
"fulfillment": {
"messages": [{
"type":
0,
"platform":
"facebook",
"speech":
random.choice(x_tend['responses'])
}]
}
}
})
def classify():
global stress
# ERROR_THRESHOLD = THRESHOLD
context = None
sentence = request.json['sentence']
nlu = json.loads(luis.luis_api(sentence))
return_list = []
# output_context = None
entities = None
# print("LUIS NLU Inference : ", nlu['topScoringIntent']['intent'])
sentiment = sentiment_analysis.sentiment_analyzer(
sentence, nlu['sentimentAnalysis']['score'],
nlu['sentimentAnalysis']['label'])
if nlu['topScoringIntent']['intent'] == "None":
# print("Fallback detected")
stress_payload = stress_analysis.stress_analyzer(
sentiment['polarity'], 'fallback', stress)
stress = stress_payload['stress']
trigger = stress_payload['trigger']
responsive = stress_payload['responsive']
reaction = stress_payload['reaction']
completion = stress_payload['completion']
return_list.append({
"query":
sentence,
"intent":
"fallback",
"response":
random.choice(fallback_dict),
"context":
context,
"probability":
"{0:.2f}".format(nlu['topScoringIntent']['score']),
"entities":
entities,
"sentiment":
sentiment,
"stress":
stress,
"trigger":
trigger,
"responsive":
responsive,
"reaction":
reaction,
'completion':
False
})
else:
# print("Inference Exists")
stress_payload = stress_analysis.stress_analyzer(
sentiment['polarity'], nlu['topScoringIntent']['intent'], stress)
stress = stress_payload['stress']
trigger = stress_payload['trigger']
responsive = stress_payload['responsive']
reaction = stress_payload['reaction']
completion = stress_payload['completion']
if completion is True:
# print("Extraction completion event triggered!")
return_list.append({
"query":
sentence,
"intent":
nlu['topScoringIntent']['intent'],
"response":
random.choice(extraction_dict),
"context":
context,
"probability":
"{0:.2f}".format(nlu['topScoringIntent']['score']),
"entities":
entities,
"sentiment":
sentiment,
"stress":
stress,
"trigger":
trigger,
"responsive":
responsive,
"reaction":
reaction,
'completion':
completion
})
else:
if stress_payload['repeat'] is not None:
return_list.append({
"query":
sentence,
"intent":
nlu['topScoringIntent']['intent'],
"response":
random.choice(repeat_dict) + " " +
"You are only talking about " +
nlu['topScoringIntent']['intent'].replace("_", " "),
"context":
context,
"probability":
"{0:.2f}".format(nlu['topScoringIntent']['score']),
"entities":
entities,
"sentiment":
sentiment,
"stress":
stress,
"trigger":
trigger,
"responsive":
responsive,
"reaction":
reaction,
'completion':
completion
})
else:
for x_tend in intents['intents']:
if nlu['topScoringIntent']['intent'] == x_tend['tag']:
normal_response = random.choice(x_tend['responses'])
return_list.append({
"query":
sentence,
"intent":
nlu['topScoringIntent']['intent'],
"response":
normal_response,
"context":
context,
"probability":
"{0:.2f}".format(nlu['topScoringIntent']['score']),
"entities":
entities,
"sentiment":
sentiment,
"stress":
stress,
"trigger":
trigger,
"responsive":
responsive,
"reaction":
reaction,
'completion':
completion
})
response = jsonify({"result": return_list, "error": None})
# print("Completion Status : {}".format(completion))
if completion:
stress = DEFAULT_STRESS
# print("Level complete. Resetting Stress to default {}".format(DEFAULT_STRESS))
print("GLOBAL STRESS LEVEL : {}".format(stress))
return response
def classify():
global stress
ERROR_THRESHOLD = 0.75
context = None
sentence = request.json['sentence']
# print(sentence)
sentiment = sentiment_analysis.sentiment_analyzer(sentence)
if "context" in request.json:
# print("Context present : ", request.json['context'])
context = request.json['context']
else:
# print("Context not present!")
context = None
entities = entity_extraction.named_entity_extraction(sentence)
# generate probabilities from the model
input_data = pd.DataFrame([bow(sentence, words)],
dtype=float,
index=['input'])
results = model.predict([input_data])[0]
# filter out predictions below a threshold
# print("Before Filter : ", results)
results = [[i, r] for i, r in enumerate(results) if r > ERROR_THRESHOLD]
# print("After Filter : ", results)
# sort by strength of probability
results.sort(key=lambda x: x[1], reverse=True)
return_list = []
output_context = None
# print("Results Length : ", len(results))
if len(results) == 0:
return_list.append({
"query": sentence,
"intent": "fallback",
"response": random.choice(fallback_dict),
"context": None,
"probability": "0.00",
"entities": None,
"sentiment": sentiment,
"stress": stress,
"trigger": 'confused',
"responsive": True,
"reaction": None,
'completion': False
})
else:
# print("Inference Exists")
for r in results:
if context != None:
classes[r[0]] = context
# print("Class Value : ", classes[r[0]])
for x_tend in intents['intents']:
if classes[r[0]] == x_tend['tag']:
# print("Entities Length : ", len(entities))
if x_tend['context'] == "":
output_context = None
if entities is None:
entities = None
elif len(entities) == 0:
entities = None
stress_payload = stress_analysis.stress_analyzer(
sentiment['polarity'], classes[r[0]], stress)
stress = stress_payload['stress']
trigger = stress_payload['trigger']
responsive = stress_payload['responsive']
reaction = stress_payload['reaction']
completion = stress_payload['completion']
return_list.append({
"query":
sentence,
"intent":
classes[r[0]],
"response":
random.choice(x_tend['responses']),
"context":
output_context,
"probability":
str(round(r[1], 2)),
"entities":
entities,
"sentiment":
sentiment,
"stress":
stress,
"trigger":
trigger,
"responsive":
responsive,
"reaction":
reaction,
'completion':
completion
})
# return tuple of intent and probability
response = jsonify({"result": return_list, "error": None})
print("GLOBAL STRESS LEVEL : {}".format(stress))
return response
def classify():
global stress
ERROR_THRESHOLD = THRESHOLD
context = None
sentence = request.json['sentence']
# print(sentence)
sentiment = sentiment_analysis.sentiment_analyzer(sentence)
if "context" in request.json:
# print("Context present : ", request.json['context'])
context = request.json['context']
else:
# print("Context not present!")
context = None
entities = entity_extraction.named_entity_extraction(sentence)
# generate probabilities from the model
input_data = pd.DataFrame([bow(sentence, words)],
dtype=float,
index=['input'])
results = model.predict([input_data])[0]
# filter out predictions below a threshold
# print("Before Filter : ", results)
results = [[i, r] for i, r in enumerate(results) if r > ERROR_THRESHOLD]
# print("After Filter : ", results)
# sort by strength of probability
results.sort(key=lambda x: x[1], reverse=True)
return_list = []
output_context = None
# print("Results Length : ", len(results))
if len(results) == 0:
# print("Fallback detected")
stress_payload = stress_analysis.stress_analyzer(
sentiment['polarity'], 'fallback', stress)
stress = stress_payload['stress']
trigger = stress_payload['trigger']
responsive = stress_payload['responsive']
reaction = stress_payload['reaction']
completion = stress_payload['completion']
fallback_logger(sentence)
return_list.append({
"query": sentence,
"intent": "fallback",
"response": random.choice(fallback_dict),
"context": None,
"probability": "0.00",
"entities": None,
"sentiment": sentiment,
"stress": stress,
"trigger": trigger,
"responsive": responsive,
"reaction": reaction,
'completion': False
})
else:
# print("Inference Exists")
for r in results:
if context != None:
classes[r[0]] = context
# print("Class Value for context: ", classes[r[0]])
for x_tend in intents['intents']:
if classes[r[0]] == x_tend['tag']:
# print("Entities Length : ", len(entities))
if x_tend['context'] == "":
output_context = None
if entities is None:
entities = None
elif len(entities) == 0:
entities = None
stress_payload = stress_analysis.stress_analyzer(
sentiment['polarity'], classes[r[0]], stress)
stress = stress_payload['stress']
trigger = stress_payload['trigger']
responsive = stress_payload['responsive']
reaction = stress_payload['reaction']
completion = stress_payload['completion']
if completion is True:
print("Extraction completion event triggered!")
return_list.append({
"query":
sentence,
"intent":
classes[r[0]],
"response":
random.choice(extraction_dict),
"context":
output_context,
"probability":
"{0:.5f}".format(r[1]),
"entities":
entities,
"sentiment":
sentiment,
"stress":
stress,
"trigger":
trigger,
"responsive":
responsive,
"reaction":
reaction,
'completion':
completion
})
# if reaction == 'extreme':
if stress_payload['repeat'] is not None:
return_list.append({
"query":
sentence,
"intent":
classes[r[0]],
"response":
random.choice(repeat_dict) + " " +
"You are only talking about " +
classes[r[0]].replace("_", " "),
"context":
output_context,
"probability":
"{0:.5f}".format(r[1]),
"entities":
entities,
"sentiment":
sentiment,
"stress":
stress,
"trigger":
trigger,
"responsive":
responsive,
"reaction":
reaction,
'completion':
completion
})
elif reaction == 'shock':
return_list.append({
"query":
sentence,
"intent":
classes[r[0]],
"response":
"",
"context":
output_context,
"probability":
"{0:.5f}".format(r[1]),
"entities":
entities,
"sentiment":
sentiment,
"stress":
stress,
"trigger":
trigger,
"responsive":
responsive,
"reaction":
reaction,
'completion':
completion
})
normal_response = random.choice(x_tend['responses'])
normal_intent = classes[r[0]]
conversation_logger(sentence, normal_intent,
normal_response)
return_list.append({
"query": sentence,
"intent": normal_intent,
"response": normal_response,
"context": output_context,
"probability": "{0:.5f}".format(r[1]),
"entities": entities,
"sentiment": sentiment,
"stress": stress,
"trigger": trigger,
"responsive": responsive,
"reaction": reaction,
'completion': completion
})
# return tuple of intent and probability
response = jsonify({"result": return_list, "error": None})
# print("Completion Status : {}".format(completion))
if completion:
stress = DEFAULT_STRESS
# print("Level complete. Resetting Stress to default {}".format(DEFAULT_STRESS))
# print("GLOBAL STRESS LEVEL : {}".format(stress))
return response