1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
|
import re
import funcy
import csv
import numpy as np
import datetime
import json
from collections import defaultdict
import glob
from scipy.spatial.distance import cosine
with open('data/styles.json') as f:
styles = json.load(f)
with open('data/beer_info.json', 'r') as f:
beer_names = json.load(f)
attr_to_styles = defaultdict(list)
style_to_attrs = defaultdict(list)
with open('data/sparser_beer_data.csv') as f:
reader = csv.DictReader(f)
for line in reader:
style = line.pop('Style')
for k, v in line.items():
if v == '1':
attr_to_styles[k].append(style)
style_to_attrs[style].append(k)
# In[7]:
with open('data/final_data_small.json') as f:
final_data = json.load(f)
failed = 0
beer_by_style = defaultdict(list)
for beer, data in final_data.items():
try:
a, b = beer.split('-')
real_style = beer_names[b + '-' + a][1][0]
if beer_names[b + '-' + a][3] > 7:
beer_by_style[real_style].append(data['embed'])
except:
failed += 1
style_centers = {}
for style, datas in beer_by_style.items():
style_centers[style.strip()] = np.mean(datas, axis=0)
style_name_to_num = {y[0]: x for x,y in styles.items()}
attr_centers = {}
for attr, rel_styles in attr_to_styles.items():
centers = [style_centers[x] for x in rel_styles]
attr_mean = np.mean(centers, axis=0)
attr_centers[attr] = attr_mean
embeddings = [(x[0], x[1]['embed']) for x in final_data.items()]
small_embeddings = []
small_styles = []
bad = 0
for e in embeddings:
b, a = e[0].split('-')
newkey = a + '-' + b
if newkey in beer_names:
info= beer_names[newkey]
if info[3] > 25:
small_embeddings.append(e)
small_styles.append(info[1][0])
else:
bad += 1
def get_closest(beer_id):
one_embed = final_data[beer_id]['embed']
for thing in sorted(small_embeddings, key = lambda x: cosine(one_embed, x[1]), reverse=False)[:25]:
b, a = thing[0].split('-')
# print(b, a)
if a + '-' + b in beer_names:
print(beer_names[a + '-' + b])
print('=' * 50)
def get_closest_to_point(one_embed, style_limit=[]):
# one_embed = final_data[beer_id]['embed']
if style_limit:
possible_beers = []
for style, e in zip(small_styles, small_embeddings):
if style == style_limit:
possible_beers.append(e)
else:
possible_beers = small_embeddings
print(len(possible_beers))
for thing in sorted(possible_beers, key = lambda x: cosine(one_embed, x[1]), reverse=False)[:5]:
b, a = thing[0].split('-')
# print(b, a)
if a + '-' + b in beer_names:
bn = beer_names[a + '-' + b]
print(bn[0], '---', bn[1][0])
print('=' * 50)
# In[16]:
def translate_to_attr(embedding, to_attr):
if to_attr.startswith('-'):
back = True
to_attr = to_attr[1:]
else:
back = False
relevant_styles = attr_to_styles[to_attr]
small_rel_centers = {x: y for x, y in style_centers.items() if x in relevant_styles}
sorted_centers = sorted(small_rel_centers, key=lambda x: cosine(small_rel_centers[x], embedding))
closest_center = sorted_centers[0]
closest_center_vector = style_centers[closest_center]
vector_between = closest_center_vector - embedding
print('Moving Towards/From: {}'.format(closest_center))
print('=' *10)
print('\n')
dist_between = np.linalg.norm(closest_center_vector-embedding)
for x in [4,2, 1]:
print('Moving {}%'.format(1/float(x) * 100))
if back:
new_point = embedding - (vector_between/x/5)
else:
new_point = embedding + (vector_between/x)
get_closest_to_point(new_point)
print('-' * 25)
def translate_to_style(embedding, style):
closest_center_vector = style_centers[style]
vector_between = closest_center_vector - embedding
# print(vector_between)
print('Moving Towards/From: {}'.format(style))
print('=' *10)
print('\n')
dist_between = np.linalg.norm(closest_center_vector-embedding)
for x in [4,2, 1]:
print('Moving {}%'.format(1/float(x) * 100))
new_point = embedding + (vector_between/x)
get_closest_to_point(new_point, style_limit=style)
print('-' * 25)
# In[17]:
# beer_id = '388-1703' # cantillio gueueze
# beer_id = '140-276' # SN pale ale
beer_id = '4-59' # allagash white
# beer_id = '64-33832' # palo salto marron
info = beer_names[beer_id.split('-')[1] + '-' + beer_id.split('-')[0]]
print('Real ABV: {}'.format(info[2]))
one_embed = final_data[beer_id]['embed']
translate_to_style(one_embed, 'Belgian Quadrupel (Quad)')
# In[22]:
# beer_id = '388-1703' # cantillio gueueze
# beer_id = '140-276' # SN pale ale
beer_id = '4-59' # allagash white
# beer_id = '64-33832' # palo salto marron
info = beer_names[beer_id.split('-')[1] + '-' + beer_id.split('-')[0]]
print('Real ABV: {}'.format(info[2]))
one_embed = final_data[beer_id]['embed']
translate_to_attr(one_embed, 'Hoppy')
# In[ ]:
# In[ ]:
|
