Module faq_forum.forum
Expand source code
from cluster import connector as cluster
from faq_forum.question_match import match
from faq_forum.auto_moderator import offensiveness, is_nonsense
import logging, sys
import traceback
#logging.basicConfig(stream=sys.stderr, level=logging.DEBUG)
def __get_match(question, question_set):
"""
Find the best semantic match to question from the given question_set.
Args:
question: The question to match
question_set: The questions to match to (given as a list of dicts {"question_id":XXX,"question":YYY})
Returns: A tuple (p,best) consisting of the 'best' match from the given question_set and the probability,
'p', that that 'best' match is semantically equal to the given question. (1 = equal, 0 = unequal,
0 <= p <= 1). If the question_set was empty or None, (0.0, None) is returned
"""
prob = 0.0
best = None
# Only loop over the questions in question_set if it is not None
if question_set is not None:
# Find the best match to the given question
best = max(question_set, key=lambda x: match(question, x["question"]))
prob = match(question, best["question"])
# Right now we only store the best match, but we could respond with the X best
# matches as well
return \
[
{
"question_id": best["question_id"],
"prob": prob
}
]
def __get_offensiveness(sentence):
"""
Estimate the offensiveness of a sentence.
Args:
sentence: The sentence to estimate the offensiveness of
Returns: the probability,'p', that that the given question is offensive. (1 = yes, 0 = no,
0 <= p <= 1). 0.0 is returned if the given sentence was None.
"""
if sentence is None:
return 0.0
return offensiveness(sentence)
def __unwrap_match_request(request):
"""
Unwrap the given "match questions" request into a tuple of Python objects.
Args:
request: A JSON object describing a "match questions" request
Returns: A tuple (question,question_set) where question is a string and question_set is an iterable
collection of questions and their ids (given as dicts {"question_id":XXX,"question":YYY}).
"""
# request = json.loads(request)
question = request["question"]
question_set = request["compare_questions"]
return question, question_set
def __wrap_match_request(request, best_matches):
"""
Wrap the given result of a "match questions" request in a JSON object
Args:
request: The request that was processed
best_matches: A list of best matches and their probabilities, given as a list of
{"question_id":XXX,"prob":YYY} dicts
Returns: A JSON-like dict containing all the given information
(as described on https://clusterdocs.azurewebsites.net/)
"""
ans = \
{
"question_id": request["question_id"],
"possible_matches": best_matches,
"msg_id": request["msg_id"],
"question": request["question"]
}
return ans
def __unwrap_offensive_request(request):
"""
Unwrap the given "estimate offensiveness" request into a string.
Args:
request: A JSON-like dict describing an "estimate offensiveness" request
(as described on https://clusterdocs.azurewebsites.net/)
Returns: A string that represents the sentence of which to estimate the offensiveness
"""
# request = json.loads(request)
sentence = request["sentence"]
return sentence
def __wrap_offensive_request(request, prob):
"""
Wrap the given result of an "estimate offensiveness" request in a JSON-like dict
Args:
request: The request that was processed
prob: The probability that the question is offensive, a float
Returns: A JSON-like dict containing all the given information
(as described on https://clusterdocs.azurewebsites.net/)
"""
ans = \
{
"sentence_id": request["sentence_id"],
"prob": prob,
"msg_id": request["msg_id"],
"sentence": request["sentence"]
}
return ans
def __unwrap_nonsense_request(request):
"""
Unwrap the given "estimate nonsense" request into a string.
Args:
request: A JSON-like dict describing an "estimate nonsense" request
(as described on https://clusterdocs.azurewebsites.net/)
Returns: A string that represents the sentence of which to estimate the offensiveness
"""
# request = json.loads(request)
sentence = request["sentence"]
return sentence
def __wrap_nonsense_request(request, is_nonsense):
"""
Wrap the given result of an "estimate nonsense" request in a JSON-like dict
Args:
request: The request that was processed
is_nonsense: True if the question is nonsense
False if the question is not nonsense
Returns: A JSON-like dict containing all the given information
(as described on https://clusterdocs.azurewebsites.net/)
"""
ans = \
{
"sentence_id": request["sentence_id"],
"nonsense": is_nonsense,
"msg_id": request["msg_id"],
"sentence": request["sentence"]
}
return ans
def process(request):
"""
Process the given request and store the reply in a JSON-like dict.
The given request is not None, it is a JSON-like dict
Args:
request: A JSON-like dict describing the request
(as described on https://clusterdocs.azurewebsites.net/)
Returns: The reply to the given request
"""
error = \
{
"error": "request was invalid"
}
# Process the request
if request is None:
ans = error
else:
# req_dict = json.loads(request)
if "action" not in request:
ans = error
elif request["action"] == cluster.Actions.MATCH_QUESTIONS.value:
inp = __unwrap_match_request(request)
out = __get_match(inp[0], inp[1])
ans = __wrap_match_request(request,
out)
elif request["action"] == cluster.Actions.ESTIMATE_OFFENSIVENESS.value:
inp = __unwrap_offensive_request(request)
out = __get_offensiveness(inp)
ans = __wrap_offensive_request(request,
out)
elif request["action"] == cluster.Actions.IS_NONSENSE.value:
inp = __unwrap_nonsense_request(request)
# Should we try/catch here ?
out = is_nonsense(inp)
ans = __wrap_nonsense_request(request, out)
else:
ans = \
{
"error": "the given request is not supported"
}
# return json.dumps(ans)
return ans
def main():
"""
Go into a while loop and wait for requests from Cluster.
Returns: None
"""
while True:
try:
# Connect to the server
faq = cluster.Connector()
break
except Exception:
# Retry connecting until it succeeds
pass
# Go to an infinite loop of processing requests of the FAQ forum
while True:
try:
# Get the request
print("::: waiting for request")
request = faq.get_next_task(timeout=None)
print("::: received request")
# HOTFIX, needs to be fixed in the connector
if "question" in request and "question_id" in request:
request["sentence"] = request["question"]
request["sentence_id"] = request["question_id"]
# Process the request
ans = process(request)
while True:
try:
# Answer to the request
print("::: sending answer")
print(request)
print(ans)
faq.reply(ans)
print("::: answer sent")
break
except Exception:
traceback.print_exc()
print("::: could not send the answer")
# Retry sending the reply until it succeeds
pass
except Exception:
# Retry getting a request until it succeeds
pass
def _test():
req_1 = \
{
"action": cluster.Actions.MATCH_QUESTIONS.value,
"question": "Where is the coffee machine?",
"question_id": 123,
"compare_questions": [
{
"question_id": 111,
"question": "Where can I find the coffee machine?"
},
{
"question_id": 222,
"question": "When did the Titanic sink?"
},
{
"question_id": 333,
"question": "How can I use the coffee machine?"
}
],
"msg_id": 214
}
ans_1 = process(req_1)
print(ans_1)
req_2 = \
{
"action": cluster.Actions.ESTIMATE_OFFENSIVENESS.value,
"question_id": 100,
"question": "Charlie is a little bitch, haha, what a little shit :p",
"msg_id": 345
}
ans_2 = process(req_2)
print(ans_2)
req_3 = \
{
"action": cluster.Actions.IS_NONSENSE.value,
"question_id": 200,
"question": "xmkjnezoinmkqzm. apeozfimkln. azefpqj wdsoimkalez.",
"msg_id": 654
}
ans_3 = process(req_3)
print(ans_3)
if __name__ == "__main__":
# Remove the test() method in deployed scripts, it is purely used for debugging
main()Functions
def main()-
Go into a while loop and wait for requests from Cluster.
Returns: None
Expand source code
def main(): """ Go into a while loop and wait for requests from Cluster. Returns: None """ while True: try: # Connect to the server faq = cluster.Connector() break except Exception: # Retry connecting until it succeeds pass # Go to an infinite loop of processing requests of the FAQ forum while True: try: # Get the request print("::: waiting for request") request = faq.get_next_task(timeout=None) print("::: received request") # HOTFIX, needs to be fixed in the connector if "question" in request and "question_id" in request: request["sentence"] = request["question"] request["sentence_id"] = request["question_id"] # Process the request ans = process(request) while True: try: # Answer to the request print("::: sending answer") print(request) print(ans) faq.reply(ans) print("::: answer sent") break except Exception: traceback.print_exc() print("::: could not send the answer") # Retry sending the reply until it succeeds pass except Exception: # Retry getting a request until it succeeds pass def process(request)-
Process the given request and store the reply in a JSON-like dict. The given request is not None, it is a JSON-like dict
Args
request- A JSON-like dict describing the request
- (as described on https://clusterdocs.azurewebsites.net/)
Returns:Thereplytothegivenrequest
Expand source code
def process(request): """ Process the given request and store the reply in a JSON-like dict. The given request is not None, it is a JSON-like dict Args: request: A JSON-like dict describing the request (as described on https://clusterdocs.azurewebsites.net/) Returns: The reply to the given request """ error = \ { "error": "request was invalid" } # Process the request if request is None: ans = error else: # req_dict = json.loads(request) if "action" not in request: ans = error elif request["action"] == cluster.Actions.MATCH_QUESTIONS.value: inp = __unwrap_match_request(request) out = __get_match(inp[0], inp[1]) ans = __wrap_match_request(request, out) elif request["action"] == cluster.Actions.ESTIMATE_OFFENSIVENESS.value: inp = __unwrap_offensive_request(request) out = __get_offensiveness(inp) ans = __wrap_offensive_request(request, out) elif request["action"] == cluster.Actions.IS_NONSENSE.value: inp = __unwrap_nonsense_request(request) # Should we try/catch here ? out = is_nonsense(inp) ans = __wrap_nonsense_request(request, out) else: ans = \ { "error": "the given request is not supported" } # return json.dumps(ans) return ans