def find_intent_match(self, responses, user_message): bow_user_message=Counter(preprocess(user_message)) processed_responses=[Counter(preprocess(response)) for response in responses] similarity_list=[compare_overlap(response, bow_user_message) for response in processed_responses] print(similarity_list) response_index=similarity_list.index(max(similarity_list)) return responses[response_index]
def find_intent_match(self, questions, user_message):
bow_user_message = Counter(preprocess(user_message))
processed_questions = [
Counter(preprocess(question)) for question in questions
]
# define similarity_list here:
similarity_list = [
compare_overlap(bow_user_message, question)
for question in processed_questions
]
# define response_index here:
response_index = similarity_list.index(max(similarity_list))
return faq[questions[response_index]]
def find_intent_match(self, responses, user_message):
# create a Bag-of-Words model
bow_user_message = Counter(preprocess(user_message))
processed_responses = [
Counter(preprocess(response)) for response in responses
]
# select the response that best matches the intent of the user message
similarity_list = [
compare_overlap(item, bow_user_message)
for item in processed_responses
]
# select the index of the highest similarity score
response_index = similarity_list.index(max(similarity_list))
return responses[response_index]
def find_entities(self, user_message):
tagged_user_message = pos_tag(preprocess(user_message))
message_nouns = extract_nouns(tagged_user_message)
tokens = word2vec(" ".join(message_nouns))
category = word2vec(blank_spot)
word2vec_result = compute_similarity(tokens, category)
word2vec_result.sort(key=lambda x: x[2])
return word2vec_result[-1][0]
def find_entities(self, user_message):
tagged_user_message = pos_tag(preprocess(user_message))
message_nouns = extract_nouns(tagged_user_message)
# fit a word2vec model on our candidate entities
tokens = word2vec(" ".join(message_nouns))
category = word2vec(blank_spot)
word2vec_result = compute_similarity(tokens, category)
# select the entity with the highest similarity score
word2vec_result.sort(key=lambda x: x[2])
if len(word2vec_result) > 0:
return word2vec_result[-1][0]
else:
return blank_spot
from user_functions import preprocess, compare_overlap, pos_tag, extract_nouns, word2vec, compute_similarity
from collections import Counter
user_message = "Good morning... will it rain in Chicago later this week?"
blank_spot = "illinois city"
response_a = "The average temperature this weekend in {} with be 88 degrees. Bring your sunglasses!"
response_b = "Forget about your umbrella; there is no rain forecasted in {} this weekend."
response_c = "This weekend, a warm front from the southeast will keep skies near {} clear."
responses = [response_a, response_b, response_c]
#preprocess documents
bow_user_message = Counter(preprocess(user_message))
processed_responses = [Counter(preprocess(response)) for response in responses]
#determine intent with BoW
similarity_list = [
compare_overlap(doc, bow_user_message) for doc in processed_responses
]
#extract entities with word2vec
tagged_user_message = pos_tag(preprocess(user_message))
message_nouns = extract_nouns(tagged_user_message)
#execute word2vec below
tokens = word2vec(" ".join(message_nouns))
category = word2vec(blank_spot)
word2vec_result = compute_similarity(tokens, category)
#select final response below
