def test_query(u, v, t, input_list, dimen, no_similar, a, **pca):
stop_words = set(stopwords.words('english'))
word2vec_value = np.load('word2vec_value.npy')
model = Word2Vec.load('model_word2vec.bin')
district = input_list[0]
state = input_list[1]
sent = input_list[2]
sent = data_cleaner.sentence_cleaner(sent)
# sent_words = [district,state] + sent.split(" ")
sent_words = sent.split(" ")
sent_value = np.array([0.0 for k in range(dimen)])
sent_new = []
count = 0
for i in sent_words:
if i not in stop_words:
try:
# if count in [0]:
# factor = a/(a + u[i]/u.N())
# if count in [1]:
# factor = a/(a + v[i]/v.N())
# if count in [2,3,4]:
# factor = a/(a + t[i]/t.N())
factor = a / (a + t[i] / t.N()) #1
# factor = 1000
sent_value += model[i] * factor
count += 1
except:
count += 1
continue
sent_value = sent_value / count
if pca == 'Yes':
sent_value = sent_value - u_ut * np.array(sent_value)
all_dist = []
for i in word2vec_value:
dist = scipy.spatial.distance.cosine(i, sent_value)
all_dist.append(dist)
k = no_similar
ind = np.argpartition(all_dist, k)[:k]
return ind
print '\nSo %s, How can I help you?'%(name)
query = raw_input('')
print '\nSure, which district are you from?'
district = raw_input('')
print '\nAnd which state?'
state = raw_input('')
print '\nGive me a second\n\n'
#### Answer #####
query = query.lower()
input_list = [district,state,query]
query = data_cleaner.sentence_cleaner(query)
# weather_api.daily(district+','+state)
if 'weather' in query:
weat = weather_api.daily(district+','+state)
print '%s\n'%(weat)
print 'Would you like to the weather forecast for the coming week ?\n'
week = raw_input('').lower()
if week == 'yes':
weat = weather_api.weekly(district+','+state)
print weat
print '\nThank You for chatting'
print '\nFor further information, contact KCC'
print '\nAnd which state?'
state = 'maharashtra' #raw_input('')
print '\nGive me a second\n\n'
#### test #####
predicted_ans_list = []
predicted_query_list = []
if 'weather' in query:
print 'weather query not to be tested'
else:
query = query.lower()
query_t = data_cleaner.sentence_cleaner(query)
input_list = [district, state, query_t]
print input_list
u, v, t, new_maharashtra, maharashtra = wv.pre('all_files.csv')
ind = wv.test_query(u, v, t, input_list, dimen, k, a, pca=pca_text)
# wv.print_ans(ind, maharashtra, k)
maharashtra = maharashtra.reset_index()
for i in ind:
predicted_query_list.append(maharashtra['Query'][i])
# print 'Answer: %s\n\n'%(maharashtra['Ans'][i])
exec('mah_list=%s' % (maharashtra['Ans'][i]))
for j in mah_list:
predicted_ans_list.append(j)
query = query.lower()
input_list = [district, state, query]
ind = wv.test_metric(u,
v,
t,
query,
dimen,
k,
a,
model,
word2vec_value,
pca=pca_text)
pdf = maharashtra.reset_index()
query_list = data_cleaner.sentence_cleaner(query)
fin_index = wv.entity(ind, query_list, pdf)
lesk_score = similarity.compute_lesk_score(query,
pdf['Query'][ind[fin_index]])
jaccard_score = similarity.compute_jaccard_sim(
query, pdf['Query'][ind[fin_index]])
# print pdf['Query'][ind[fin_index]]
if lesk_score > threshold_lesk:
count_lesk_threshold += 1
if lesk_score > 0:
count_lesk += 1