RaphaelThesmar commited on
Commit
97291c9
·
verified ·
1 Parent(s): 64a6ac3

File used to make the kaggle competition as well as hugging dace dataset

Browse files
Files changed (1) hide show
  1. clean_data_for_hugging_face.ipynb +380 -0
clean_data_for_hugging_face.ipynb ADDED
@@ -0,0 +1,380 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "cells": [
3
+ {
4
+ "cell_type": "code",
5
+ "execution_count": 3,
6
+ "id": "bfa8dc49-e3e5-4c07-9ce8-576617bba299",
7
+ "metadata": {},
8
+ "outputs": [],
9
+ "source": [
10
+ "# Importing relevant libraries\n",
11
+ "import pandas as pd\n",
12
+ "import numpy as np\n",
13
+ "import json"
14
+ ]
15
+ },
16
+ {
17
+ "cell_type": "code",
18
+ "execution_count": 20,
19
+ "id": "faeb31dd-97d3-46e8-b956-2fac4ee6d463",
20
+ "metadata": {},
21
+ "outputs": [],
22
+ "source": [
23
+ "# Fixing and cleaning up the new categories mapping\n",
24
+ "with open(r\"./dataset/categories.txt\")as f:\n",
25
+ " CATEGORIES = json.load(f)\n",
26
+ "\n",
27
+ "CATEGORIES = set(CATEGORIES.keys())\n",
28
+ "\n",
29
+ "IGNORED_CATEGOREIES = set([\n",
30
+ " \"atom-ph\", \"bayes-an\", \"chao-dyn\", \"chem-ph\", \"cmp-lg\", \"comp-gas\", \"cond-mat\", \"dg-ga\", \"funct-an\",\n",
31
+ " \"mtrl-th\", \"patt-sol\", \"plasm-ph\", \"q-alg\", \"q-bio\", \"solv-int\", \"supr-con\", \"test\", \"test.dis-nn\", \n",
32
+ " \"test.mes-hall\", \"test.mtrl-sci\", \"test.soft\", \"test.stat-mech\", \"test.str-el\", \"test.supr-con\"\n",
33
+ "])\n",
34
+ "\n",
35
+ "CATEGORY_ALIASES = {\n",
36
+ " 'math.MP': 'math-ph',\n",
37
+ " 'stat.TH': 'math.ST',\n",
38
+ " 'math.IT': 'cs.IT',\n",
39
+ " 'q-fin.EC': 'econ.GN',\n",
40
+ " 'cs.SY': 'eess.SY',\n",
41
+ " 'cs.NA': 'math.NA'\n",
42
+ "}"
43
+ ]
44
+ },
45
+ {
46
+ "cell_type": "code",
47
+ "execution_count": null,
48
+ "id": "cd452c2f-6b3b-4ea0-812b-2886b33bf09b",
49
+ "metadata": {},
50
+ "outputs": [],
51
+ "source": [
52
+ "for I_CAT in IGNORED_CATEGOREIES:\n",
53
+ " if I_CAT in CATEGORIES:\n",
54
+ " CATEGORIES.remove(I_CAT)\n",
55
+ "\n",
56
+ "for C_ALIAS in CATEGORY_ALIASES:\n",
57
+ " if C_ALIAS in CATEGORIES:\n",
58
+ " CATEGORIES.remove(C_ALIAS)\n",
59
+ " CATEGORIES.add(CATEGORY_ALIASES[C_ALIAS])\n",
60
+ "\n",
61
+ "CATEGORIES = sorted(list(CATEGORIES))\n",
62
+ "CATEGORIES = {CATEGORIES[i]:i for i in range(len(CATEGORIES))}\n",
63
+ "with open(r\"./dataset/categories.txt\", \"w\")as f:\n",
64
+ " json.dump(CATEGORIES, f)"
65
+ ]
66
+ },
67
+ {
68
+ "cell_type": "markdown",
69
+ "id": "eff91d4f-a44c-4d2b-a96f-0849354c8273",
70
+ "metadata": {},
71
+ "source": [
72
+ "#### For now only keep version 1"
73
+ ]
74
+ },
75
+ {
76
+ "cell_type": "code",
77
+ "execution_count": 5,
78
+ "id": "db881114-a624-456b-a4ab-d4ec8a53ad60",
79
+ "metadata": {
80
+ "scrolled": true
81
+ },
82
+ "outputs": [],
83
+ "source": [
84
+ "# Cleaning the dataset, ie. removing duplicates, verifying none of the ignored tags are there and replace category aliases\n",
85
+ "test_minor_path = r\"dataset/full_text_sample/test_minor_cats_full.json\"\n",
86
+ "train_minor_path = r\"dataset/full_text_sample/train_minor_cats_full.json\""
87
+ ]
88
+ },
89
+ {
90
+ "cell_type": "code",
91
+ "execution_count": 41,
92
+ "id": "e3bbe301-4e28-4a03-bd3c-1bbc99b43a0e",
93
+ "metadata": {
94
+ "scrolled": true
95
+ },
96
+ "outputs": [],
97
+ "source": [
98
+ "# Fixing test\n",
99
+ "pd_test_minor = pd.read_json(test_minor_path, lines=True)\n",
100
+ "\n",
101
+ "pd_test_minor.rename(columns={'major_category':'field','prime_category':'primary_subfield','abs_categories':'secondary_subfield'}, inplace=True)\n",
102
+ "pd_test_minor = pd_test_minor[~pd_test_minor['field'].isin(IGNORED_CATEGOREIES)]\n",
103
+ "pd_test_minor = pd_test_minor[(pd_test_minor['version'] == 1)]\n",
104
+ "if pd_test_minor['primary_subfield'].isin(CATEGORY_ALIASES.keys()).sum():\n",
105
+ " ValueError(\"IMPLEMENT CODE TO CHANGE CATEGORY_ALIASES\")\n",
106
+ " \n",
107
+ "pd_test_minor[['primary_subfield', \"secondary_subfield\"]].to_csv(\"dataset/baseline_datasets/test_field.csv\")\n",
108
+ "test_fields[\"Usage\"] = \"Public\"\n",
109
+ "pd_test_minor.drop(columns=['field', 'primary_subfield', \"secondary_subfield\"]).to_csv(r\"dataset/baseline_datasets/test_papers.csv\")"
110
+ ]
111
+ },
112
+ {
113
+ "cell_type": "code",
114
+ "execution_count": 43,
115
+ "id": "5e0c00ef-29c5-44e3-a73e-e3417e07d53c",
116
+ "metadata": {
117
+ "scrolled": true
118
+ },
119
+ "outputs": [
120
+ {
121
+ "name": "stdout",
122
+ "output_type": "stream",
123
+ "text": [
124
+ "Chunk 1\n",
125
+ "Chunk 2\n",
126
+ "Chunk 3\n",
127
+ "Chunk 4\n",
128
+ "Chunk 5\n",
129
+ "Chunk 6\n",
130
+ "Chunk 7\n",
131
+ "Chunk 8\n",
132
+ "Chunk 9\n",
133
+ "Chunk 10\n",
134
+ "Chunk 11\n",
135
+ "Chunk 12\n"
136
+ ]
137
+ }
138
+ ],
139
+ "source": [
140
+ "# Fixing train\n",
141
+ "chunksize = 10**4\n",
142
+ "for i, pd_train_minor in enumerate(pd.read_json(train_minor_path, lines=True, chunksize=chunksize)):\n",
143
+ " print(f\"Chunk {i+1}\")\n",
144
+ " pd_train_minor.rename(columns={'major_category':'field','prime_category':'primary_subfield','abs_categories':'secondary_subfield'}, inplace=True)\n",
145
+ " pd_train_minor = pd_train_minor[~pd_train_minor['field'].isin(IGNORED_CATEGOREIES)]\n",
146
+ " pd_train_minor = pd_train_minor[(pd_train_minor['version'] == 1)]\n",
147
+ " if pd_train_minor['primary_subfield'].isin(CATEGORY_ALIASES.keys()).sum():\n",
148
+ " ValueError(\"IMPLEMENT CODE TO CHANGE CATEGORY_ALIASES \")\n",
149
+ "\n",
150
+ " if i==0:\n",
151
+ " pd_train_minor.to_csv(r\"dataset/baseline_datasets/train.csv\")\n",
152
+ " else:\n",
153
+ " pd_train_minor.to_csv(r\"dataset/baseline_datasets/train.csv\", header=False, mode='a')"
154
+ ]
155
+ },
156
+ {
157
+ "cell_type": "markdown",
158
+ "id": "e7e5c112-eb78-4fb9-ad12-c3c4a90339a5",
159
+ "metadata": {},
160
+ "source": [
161
+ "#### Removing using titles"
162
+ ]
163
+ },
164
+ {
165
+ "cell_type": "code",
166
+ "execution_count": 123,
167
+ "id": "e968294d-a933-4dba-8142-cc593f9cfbf9",
168
+ "metadata": {},
169
+ "outputs": [
170
+ {
171
+ "data": {
172
+ "text/html": [
173
+ "<div>\n",
174
+ "<style scoped>\n",
175
+ " .dataframe tbody tr th:only-of-type {\n",
176
+ " vertical-align: middle;\n",
177
+ " }\n",
178
+ "\n",
179
+ " .dataframe tbody tr th {\n",
180
+ " vertical-align: top;\n",
181
+ " }\n",
182
+ "\n",
183
+ " .dataframe thead th {\n",
184
+ " text-align: right;\n",
185
+ " }\n",
186
+ "</style>\n",
187
+ "<table border=\"1\" class=\"dataframe\">\n",
188
+ " <thead>\n",
189
+ " <tr style=\"text-align: right;\">\n",
190
+ " <th></th>\n",
191
+ " <th>paper_id</th>\n",
192
+ " <th>version</th>\n",
193
+ " <th>yymm</th>\n",
194
+ " <th>created</th>\n",
195
+ " <th>title</th>\n",
196
+ " <th>secondary_subfield</th>\n",
197
+ " <th>abstract</th>\n",
198
+ " <th>primary_subfield</th>\n",
199
+ " <th>field</th>\n",
200
+ " <th>fulltext</th>\n",
201
+ " </tr>\n",
202
+ " </thead>\n",
203
+ " <tbody>\n",
204
+ " <tr>\n",
205
+ " <th>3888</th>\n",
206
+ " <td>1006.1164</td>\n",
207
+ " <td>2</td>\n",
208
+ " <td>1006</td>\n",
209
+ " <td>2010-06-10T05:56:49</td>\n",
210
+ " <td>Non-perturbative renormalization of quark mass...</td>\n",
211
+ " <td>hep-lat</td>\n",
212
+ " <td>We present an evaluation of the quark mass ren...</td>\n",
213
+ " <td>hep-lat</td>\n",
214
+ " <td>hep-lat</td>\n",
215
+ " <td>UTHEP-609\\nUTCCS-P-59\\n\\nNon-perturbative reno...</td>\n",
216
+ " </tr>\n",
217
+ " <tr>\n",
218
+ " <th>3957</th>\n",
219
+ " <td>1011.0157</td>\n",
220
+ " <td>1</td>\n",
221
+ " <td>1011</td>\n",
222
+ " <td>2010-10-31T14:09:45</td>\n",
223
+ " <td>Non-perturbative renormalization of quark mass...</td>\n",
224
+ " <td>hep-lat</td>\n",
225
+ " <td>We present an evaluation of the quark mass ren...</td>\n",
226
+ " <td>hep-lat</td>\n",
227
+ " <td>hep-lat</td>\n",
228
+ " <td>Non-perturbative renormalization of quark mass...</td>\n",
229
+ " </tr>\n",
230
+ " </tbody>\n",
231
+ "</table>\n",
232
+ "</div>"
233
+ ],
234
+ "text/plain": [
235
+ " paper_id version yymm created \\\n",
236
+ "3888 1006.1164 2 1006 2010-06-10T05:56:49 \n",
237
+ "3957 1011.0157 1 1011 2010-10-31T14:09:45 \n",
238
+ "\n",
239
+ " title secondary_subfield \\\n",
240
+ "3888 Non-perturbative renormalization of quark mass... hep-lat \n",
241
+ "3957 Non-perturbative renormalization of quark mass... hep-lat \n",
242
+ "\n",
243
+ " abstract primary_subfield \\\n",
244
+ "3888 We present an evaluation of the quark mass ren... hep-lat \n",
245
+ "3957 We present an evaluation of the quark mass ren... hep-lat \n",
246
+ "\n",
247
+ " field fulltext \n",
248
+ "3888 hep-lat UTHEP-609\\nUTCCS-P-59\\n\\nNon-perturbative reno... \n",
249
+ "3957 hep-lat Non-perturbative renormalization of quark mass... "
250
+ ]
251
+ },
252
+ "execution_count": 123,
253
+ "metadata": {},
254
+ "output_type": "execute_result"
255
+ }
256
+ ],
257
+ "source": [
258
+ "# TBD whether we do this or not\n",
259
+ "duplicates = pd_test_minor[pd_test_minor.duplicated(subset=['title'], keep=False)]\n",
260
+ "duplicates"
261
+ ]
262
+ },
263
+ {
264
+ "cell_type": "code",
265
+ "execution_count": null,
266
+ "id": "2a5ea0fe-cec7-4f76-891c-fb80852c75a3",
267
+ "metadata": {},
268
+ "outputs": [],
269
+ "source": []
270
+ },
271
+ {
272
+ "cell_type": "markdown",
273
+ "id": "cbc215aa-3b3d-4fbe-ae85-30ca8b9568e2",
274
+ "metadata": {},
275
+ "source": [
276
+ "#### Score metrics used on Kaggle"
277
+ ]
278
+ },
279
+ {
280
+ "cell_type": "code",
281
+ "execution_count": 66,
282
+ "id": "88a5a364-f8f6-4817-aaf7-4e6ce43f5090",
283
+ "metadata": {},
284
+ "outputs": [],
285
+ "source": [
286
+ "def score(solution: pd.DataFrame, submission: pd.DataFrame, row_id_column_name: str) -> float:\n",
287
+ "\n",
288
+ " try:\n",
289
+ " # First calculate the primary subfield accuracy which is relatively simple\n",
290
+ " res = solution[\"primary_subfield\"]==submission[\"primary_subfield\"]\n",
291
+ " primary_field_acc = res.sum()/len(res)\n",
292
+ "\n",
293
+ " # We then calculate the secondary field accuracy\n",
294
+ " sol_sec_list = solution[\"secondary_subfield\"].tolist()\n",
295
+ " sub_sec_list = solution[\"secondary_subfield\"].tolist()\n",
296
+ "\n",
297
+ " secondary_field_acc = 0\n",
298
+ " for i in range(len(sol_sec_list)):\n",
299
+ " paper_secondary_field_pred = sub_sec_list[i].split(\" \")\n",
300
+ " paper_secondary_field_sol = sol_sec_list[i].split(\" \")\n",
301
+ "\n",
302
+ " diff = len(paper_secondary_field_pred)-len(paper_secondary_field_sol)\n",
303
+ "\n",
304
+ " if diff<0:\n",
305
+ " tot = 0\n",
306
+ " for sf_sol in paper_secondary_field_sol:\n",
307
+ " if sf_sol in paper_secondary_field_pred:\n",
308
+ " tot+=1\n",
309
+ " tot /= len(paper_secondary_field_sol)\n",
310
+ " else:\n",
311
+ " tot = 0\n",
312
+ " for sf_pred in paper_secondary_field_pred:\n",
313
+ " if sf_pred in paper_secondary_field_sol:\n",
314
+ " tot+=1\n",
315
+ "\n",
316
+ " tot /= len(paper_secondary_field_pred)\n",
317
+ "\n",
318
+ " secondary_field_acc += tot\n",
319
+ "\n",
320
+ " loss = secondary_field_acc/len(sol_sec_list) + primary_field_acc\n",
321
+ " \n",
322
+ " return loss\n",
323
+ "\n",
324
+ " except Exception as e:\n",
325
+ " \n",
326
+ " raise ParticipantVisibleError(f\"Error occurred: {str(e)}\")"
327
+ ]
328
+ },
329
+ {
330
+ "cell_type": "code",
331
+ "execution_count": 67,
332
+ "id": "0497a438-13b7-47d9-a493-bd7c9fc1892d",
333
+ "metadata": {},
334
+ "outputs": [
335
+ {
336
+ "data": {
337
+ "text/plain": [
338
+ "2.0"
339
+ ]
340
+ },
341
+ "execution_count": 67,
342
+ "metadata": {},
343
+ "output_type": "execute_result"
344
+ }
345
+ ],
346
+ "source": [
347
+ "score(pd_sol, pd_sol, \"id\")"
348
+ ]
349
+ },
350
+ {
351
+ "cell_type": "code",
352
+ "execution_count": null,
353
+ "id": "38a4812b-26ce-4063-b6a7-245544e8e02a",
354
+ "metadata": {},
355
+ "outputs": [],
356
+ "source": []
357
+ }
358
+ ],
359
+ "metadata": {
360
+ "kernelspec": {
361
+ "display_name": "Python 3 (ipykernel)",
362
+ "language": "python",
363
+ "name": "python3"
364
+ },
365
+ "language_info": {
366
+ "codemirror_mode": {
367
+ "name": "ipython",
368
+ "version": 3
369
+ },
370
+ "file_extension": ".py",
371
+ "mimetype": "text/x-python",
372
+ "name": "python",
373
+ "nbconvert_exporter": "python",
374
+ "pygments_lexer": "ipython3",
375
+ "version": "3.9.19"
376
+ }
377
+ },
378
+ "nbformat": 4,
379
+ "nbformat_minor": 5
380
+ }