-
Notifications
You must be signed in to change notification settings - Fork 0
/
make_scatterplots.py
270 lines (231 loc) · 8.38 KB
/
make_scatterplots.py
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
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
import argparse
import pickle as pkl
import warnings
from collections import defaultdict
from pathlib import Path
import matplotlib.pyplot as plt
import numpy as np
from adjustText import adjust_text
from utils import calc_c_I
warnings.filterwarnings("ignore")
def plot_scatterplot(axistour_embed, topk, words, left_axis_index, length, output_path):
n, dim = axistour_embed.shape
normed_axistour_embed = axistour_embed / np.linalg.norm(
axistour_embed, axis=1, keepdims=True
)
axis_idxs = np.array([left_axis_index + i for i in range(length)])
axis_idx2axis_idx_idx = {axis_idx: idx for idx, axis_idx in enumerate(axis_idxs)}
k = 5
axis2top_word_ids = {}
for axis_idx in axis_idxs:
top_word_ids = np.argsort(normed_axistour_embed[:, axis_idx])[-k:]
axis2top_word_ids[axis_idx] = top_word_ids
print(f"axis indexes: {axis_idxs}")
proj_matrix = []
for idx in range(length - 1):
theta = np.pi * idx / (length - 1)
proj_matrix.append((np.cos(theta), np.sin(theta)))
proj_matrix.append((-1, 0))
proj_matrix = np.array(proj_matrix)
picked_emb = normed_axistour_embed[:, axis_idxs] # (n, length)
word_idx2axis_idx_idx = {}
axis_idx2word_idx_list = defaultdict(list)
for i in range(n):
max_idx = np.argmax(picked_emb[i])
word_idx2axis_idx_idx[i] = max_idx
axis_idx2word_idx_list[left_axis_index + max_idx].append(i)
for axis_idx in axis_idxs:
assert len(axis_idx2word_idx_list[axis_idx]) > 0
axis_idx2word_idx_list[axis_idx] = np.array(axis_idx2word_idx_list[axis_idx])
# make color map for each axis, total length colors
color_map = {}
for i in range(length):
color_map[i] = plt.cm.get_cmap("rainbow")(i / length)
proj_emb = np.dot(picked_emb, proj_matrix) # (n, 2)
max_x = np.max(np.abs(proj_emb[:, 0]))
max_y = np.max(np.abs(proj_emb[:, 1]))
max_x = max(max_x, max_y)
max_y = max_x
fig, axes = plt.subplots(2, figsize=(11, 10))
for ax_idx in range(2):
if ax_idx == 0:
print("\nAxis Tour")
else:
print("\nSkew Sort")
ax = axes[ax_idx]
if ax_idx == 1:
picked_emb = normed_axistour_embed[:, axis_idxs]
picked_skews = np.sum(axistour_embed[:, axis_idxs] ** 3, axis=0)
picked_skew_sort_idex = np.argsort(-picked_skews)
picked_emb = picked_emb[:, picked_skew_sort_idex]
proj_matrix = []
for idx in range(length - 1):
theta = 2 * np.pi - np.pi * idx / (length - 1)
proj_matrix.append((np.cos(theta), np.sin(theta)))
proj_matrix.append((-1, 0))
proj_matrix = np.array(proj_matrix)
proj_emb = np.dot(picked_emb, proj_matrix) # (n, 2)
axis_idxs = axis_idxs[picked_skew_sort_idex]
for idx, axis_idx in enumerate(axis_idxs):
if ax_idx == 0:
if idx < len(axis_idxs) - 1:
theta = np.pi * idx / (length - 1)
else:
theta = np.pi
else:
if idx < len(axis_idxs) - 1:
theta = 2 * np.pi - np.pi * idx / (length - 1)
else:
theta = np.pi
x, y = np.cos(theta), np.sin(theta)
x *= max_x
y *= max_y
point = {"start": (0, 0), "end": (x, y)}
ax.annotate(
"",
xytext=point["start"],
xy=point["end"],
arrowprops=dict(
shrink=0,
width=2,
headwidth=7,
headlength=7,
connectionstyle="arc3",
facecolor=color_map[axis_idx2axis_idx_idx[axis_idx]],
edgecolor="black",
linewidth=0.5,
),
)
ax.text(
1.05 * x,
1.05 * y,
f"{axis_idx}",
fontsize=18,
ha="center",
va="center",
color=color_map[axis_idx2axis_idx_idx[axis_idx]],
)
word_idxs = np.argsort(np.linalg.norm(picked_emb, axis=1))
xs = proj_emb[word_idxs, 0]
ys = proj_emb[word_idxs, 1]
colors = [color_map[word_idx2axis_idx_idx[word_idx]] for word_idx in word_idxs]
ax.scatter(xs, ys, s=10, marker="o", color=colors, alpha=0.5)
texts = []
lens = []
for axis_idx in axis_idxs:
idx = axis_idx2axis_idx_idx[axis_idx]
top_word_ids = axis2top_word_ids[axis_idx]
for id_ in top_word_ids:
x, y = proj_emb[id_]
if ax_idx == 0:
if y < 0:
continue
else:
if y > 0:
continue
lens.append(np.linalg.norm((x, y)))
ax.scatter(
x,
y,
s=30,
marker="o",
color=color_map[idx],
edgecolors="black",
linewidths=0.5,
zorder=11,
)
texts.append(
ax.text(
x,
y,
words[id_],
fontsize=13,
color="black",
bbox=dict(
facecolor=color_map[idx],
boxstyle="round,pad=0.1",
edgecolor="gray",
linewidth=1,
),
)
)
print(f"d_I: {np.mean(lens):.2f}")
print(f"c_I: {calc_c_I(picked_emb, normed_axistour_embed, topk): .2f}")
ax.set_xlim(-max_x * 1.1, max_x * 1.1)
if ax_idx == 0:
ax.set_ylim(0, max_y * 1.1)
else:
ax.set_ylim(-max_y * 1.1, 0.0)
ax.axis("off")
ax.set_aspect("equal")
# adjust text
adjust_text(
texts,
ax=ax,
arrowprops=dict(arrowstyle="-", color="k", lw=0.5, zorder=10),
force_pull=(0.05, 0.05),
expand=(1.1, 1.1),
)
for text in texts:
text.set_zorder(10)
if length == 9:
scale_y = 0.9
elif length == 10:
scale_y = 0.98
else:
raise NotImplementedError
if ax_idx == 0:
ax.text(
-1.15 * max_x,
scale_y * max_y,
"Axis Tour",
fontsize=25,
ha="left",
va="center",
fontweight="bold",
)
else:
ax.text(
-1.15 * max_x,
-scale_y * max_y,
"Skewness Sort",
fontsize=25,
ha="left",
va="center",
fontweight="bold",
)
fig.tight_layout()
fig.subplots_adjust(
left=0.01, right=0.99, bottom=0.01, top=0.99, wspace=0.01, hspace=0.05
)
fig.savefig(output_path, dpi=150)
plt.close()
def parse_args():
parser = argparse.ArgumentParser(description="Scatterplot for Axis Tour.")
parser.add_argument("--emb_type", type=str, default="glove")
parser.add_argument("--topk", type=int, default=100)
parser.add_argument("--left_axis_index", type=int, default=86)
parser.add_argument("--length", type=int, default=9)
return parser.parse_args()
def main():
args = parse_args()
emb_type = args.emb_type
topk = args.topk
left_axis_index = args.left_axis_index
length = args.length
axistour_embed_path = (
f"output/axistour_embeddings/axistour_top{topk}_{emb_type}.pkl"
)
if not Path(axistour_embed_path).exists():
raise FileNotFoundError(f"{axistour_embed_path} does not exist")
with open(axistour_embed_path, "rb") as f:
axistour_embed, words = pkl.load(f)
output_dir = Path("output/images/scatterplots")
output_dir.mkdir(exist_ok=True, parents=True)
output_path = (
output_dir / f"scatterplot_{emb_type}_top{topk}_"
f"left{left_axis_index}_length{length}.png"
)
plot_scatterplot(axistour_embed, topk, words, left_axis_index, length, output_path)
if __name__ == "__main__":
main()