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
|
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)
beer_by_style = defaultdict(list)
for beer, data in final_data.items():
try:
real_style = beer_names[beer][1][0]
if beer_names[beer][3] > 7:
beer_by_style[real_style].append(data['embed'])
except:
pass
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:
newkey = e[0]
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']
beer_ids = []
for thing in sorted(small_embeddings, key = lambda x: cosine(one_embed, x[1]), reverse=False)[:10]:
if thing[0] in beer_names:
beer_ids.append(thing[0])
print(thing[0])
print('=' * 50)
return beer_ids
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))
to_return = []
for thing in sorted(possible_beers, key = lambda x: cosine(one_embed, x[1]), reverse=False)[:5]:
# print(b, a)
if thing[0] in beer_names:
bn = beer_names[thing[0]]
to_return.append((thing[0], bn[0], bn[1][0]))
return to_return
def translate_to_attr(beer_id, to_attr, amt):
embedding = final_data[beer_id]['embed']
if amt < 0:
back = True
amt = abs(amt)
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)
amt_dict = {
1: 4,
2: 2,
3: 1
}
x = amt_dict[amt]
print('Moving {}%'.format(1/float(x) * 100))
if back:
new_point = embedding - (vector_between/x)
else:
new_point = embedding + (vector_between/x)
return get_closest_to_point(new_point)
def translate_to_style(beer_id, style):
embedding = final_data[beer_id]['embed']
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)
TRANSLATION_CONSTANT = 4 # 4 is slight
new_point = embedding + (vector_between/TRANSLATION_CONSTANT)
return get_closest_to_point(new_point, style_limit=style)
def get_drinks_like(beer_id):
data = final_data[beer_id]
return data['alc']
|
