removed restriction on paper versions for major, minor, all2023
Browse files- README.md +10 -0
- all2023/val.json +2 -2
- major/test.json +2 -2
- major/train.json +2 -2
- minor/test.json +2 -2
- minor/train.json +2 -2
- preprocess_all2023.py +6 -21
- preprocess_major_minor.py +7 -42
- test_paper_id.py +37 -0
README.md
CHANGED
@@ -35,6 +35,16 @@ Original data: https://www.dropbox.com/scl/fo/wwu0ifghw4sco09g67frb/h?rlkey=6ddg
|
|
35 |
|
36 |
<!-- Working doc: https://docs.google.com/document/d/1ZTiR5M7NBpsfYbJdlP6qcIldBJVzJI1Hm8srgI8hfnI/edit?usp=sharing -->
|
37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
## Preprocessing
|
39 |
|
40 |
Transform raw data into JSON format:
|
|
|
35 |
|
36 |
<!-- Working doc: https://docs.google.com/document/d/1ZTiR5M7NBpsfYbJdlP6qcIldBJVzJI1Hm8srgI8hfnI/edit?usp=sharing -->
|
37 |
|
38 |
+
## Set up dependencies
|
39 |
+
|
40 |
+
Get subfields to ignore and subfield aliases:
|
41 |
+
```
|
42 |
+
# in your conda env
|
43 |
+
git clone https://github.com/ag2435/arxiv-classifier-next
|
44 |
+
cd arxiv-classifier-next
|
45 |
+
conda develop .
|
46 |
+
```
|
47 |
+
|
48 |
## Preprocessing
|
49 |
|
50 |
Transform raw data into JSON format:
|
all2023/val.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:728f5ae7ff4f3907c9a5e6f3b35d1a2968a02cd825f50dc575efe94c7d4e4ffa
|
3 |
+
size 409838242
|
major/test.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c7849149e5f504fd8f99efa3b5f1c2d65b541b2b76640d563bababa4b4f6d395
|
3 |
+
size 406464012
|
major/train.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:37aa1843751273b4e27633b1ec1a0fad6256a9143ed48d81bbf34ffae03116c4
|
3 |
+
size 1639035113
|
minor/test.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:667308dc5ddfa3cf17869673427e5ce85abf054fee301e8a1cf58ee04e6d9d3a
|
3 |
+
size 1453222079
|
minor/train.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8834e6cfa89d589cffcc5335a3bc2371ce91f6627c4a653ec689b3e370582b8e
|
3 |
+
size 5823161950
|
preprocess_all2023.py
CHANGED
@@ -37,23 +37,9 @@ import pandas as pd
|
|
37 |
import numpy as np
|
38 |
import os
|
39 |
from tqdm import tqdm
|
40 |
-
|
41 |
-
|
42 |
-
IGNORED_CATEGOREIES
|
43 |
-
"atom-ph", "bayes-an", "chao-dyn", "chem-ph", "cmp-lg", "comp-gas", "cond-mat", "dg-ga", "funct-an",
|
44 |
-
"mtrl-th", "patt-sol", "plasm-ph", "q-alg", "q-bio", "solv-int", "supr-con", "test", "test.dis-nn",
|
45 |
-
"test.mes-hall", "test.mtrl-sci", "test.soft", "test.stat-mech", "test.str-el", "test.supr-con"
|
46 |
-
])
|
47 |
-
print("number of ignored categories: ", len(IGNORED_CATEGOREIES))
|
48 |
-
CATEGORY_ALIASES = {
|
49 |
-
'math.MP': 'math-ph',
|
50 |
-
'stat.TH': 'math.ST',
|
51 |
-
'math.IT': 'cs.IT',
|
52 |
-
'q-fin.EC': 'econ.GN',
|
53 |
-
'cs.SY': 'eess.SY',
|
54 |
-
'cs.NA': 'math.NA'
|
55 |
-
}
|
56 |
-
print("number of category aliases: ", len(CATEGORY_ALIASES))
|
57 |
|
58 |
# %% [markdown]
|
59 |
# Rename columns to reflect standarized terminology:
|
@@ -104,16 +90,15 @@ for i, original_df in tqdm(enumerate(pd.read_csv(file_path, sep='\t', chunksize=
|
|
104 |
original_df[feature] = ""
|
105 |
|
106 |
# ignore entries from ignored categories (i.e., primary subfield)
|
107 |
-
df = original_df[~original_df['primary_subfield'].isin(IGNORED_CATEGOREIES)]
|
108 |
-
df = df[(df['version'] == 1)]
|
109 |
if df['primary_subfield'].isin(CATEGORY_ALIASES.keys()).sum():
|
110 |
ValueError("IMPLEMENT CODE TO CHANGE CATEGORY_ALIASES ")
|
111 |
|
112 |
# convert all columns to string per huggingface requirements
|
113 |
if i == 0:
|
114 |
-
df.
|
115 |
else:
|
116 |
-
df.
|
117 |
|
118 |
print("Saved to: ", save_path)
|
119 |
# %%
|
|
|
37 |
import numpy as np
|
38 |
import os
|
39 |
from tqdm import tqdm
|
40 |
+
from arxiv_classifier.utils import (CATEGORY_ALIASES,
|
41 |
+
GENERAL_CATEGORIES,
|
42 |
+
IGNORED_CATEGOREIES)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
|
44 |
# %% [markdown]
|
45 |
# Rename columns to reflect standarized terminology:
|
|
|
90 |
original_df[feature] = ""
|
91 |
|
92 |
# ignore entries from ignored categories (i.e., primary subfield)
|
93 |
+
df = original_df[~original_df['primary_subfield'].isin(GENERAL_CATEGORIES | IGNORED_CATEGOREIES)]
|
|
|
94 |
if df['primary_subfield'].isin(CATEGORY_ALIASES.keys()).sum():
|
95 |
ValueError("IMPLEMENT CODE TO CHANGE CATEGORY_ALIASES ")
|
96 |
|
97 |
# convert all columns to string per huggingface requirements
|
98 |
if i == 0:
|
99 |
+
df.to_json(save_path, lines=True, orient="records")
|
100 |
else:
|
101 |
+
df.to_json(save_path, lines=True, orient="records", mode='a')
|
102 |
|
103 |
print("Saved to: ", save_path)
|
104 |
# %%
|
preprocess_major_minor.py
CHANGED
@@ -37,23 +37,9 @@ import pandas as pd
|
|
37 |
import numpy as np
|
38 |
import os
|
39 |
from tqdm import tqdm
|
40 |
-
|
41 |
-
|
42 |
-
IGNORED_CATEGOREIES
|
43 |
-
"atom-ph", "bayes-an", "chao-dyn", "chem-ph", "cmp-lg", "comp-gas", "cond-mat", "dg-ga", "funct-an",
|
44 |
-
"mtrl-th", "patt-sol", "plasm-ph", "q-alg", "q-bio", "solv-int", "supr-con", "test", "test.dis-nn",
|
45 |
-
"test.mes-hall", "test.mtrl-sci", "test.soft", "test.stat-mech", "test.str-el", "test.supr-con"
|
46 |
-
])
|
47 |
-
print("number of ignored categories: ", len(IGNORED_CATEGOREIES))
|
48 |
-
CATEGORY_ALIASES = {
|
49 |
-
'math.MP': 'math-ph',
|
50 |
-
'stat.TH': 'math.ST',
|
51 |
-
'math.IT': 'cs.IT',
|
52 |
-
'q-fin.EC': 'econ.GN',
|
53 |
-
'cs.SY': 'eess.SY',
|
54 |
-
'cs.NA': 'math.NA'
|
55 |
-
}
|
56 |
-
print("number of category aliases: ", len(CATEGORY_ALIASES))
|
57 |
|
58 |
# %% [markdown]
|
59 |
# Rename columns to reflect standarized terminology:
|
@@ -74,16 +60,7 @@ os.makedirs(save_dir, exist_ok=True)
|
|
74 |
save_path = os.path.join(save_dir, f"{split}.json")
|
75 |
print("Reading from file: ", file_path)
|
76 |
|
77 |
-
|
78 |
-
# if os.path.exists(save_path):
|
79 |
-
# print("Overwriting existing file: ", save_path)
|
80 |
-
# os.remove(save_path)
|
81 |
-
|
82 |
-
# with open(save_path, 'a') as f:
|
83 |
-
for i, original_df in tqdm(enumerate(pd.read_json(file_path, lines=True, chunksize=10**4))):
|
84 |
-
# print(f"Chunk {i+1}")
|
85 |
-
# print("original columns: ", original_df.columns)
|
86 |
-
|
87 |
# Rename columns
|
88 |
original_df.rename(columns={
|
89 |
'major_category': 'field',
|
@@ -91,25 +68,13 @@ for i, original_df in tqdm(enumerate(pd.read_json(file_path, lines=True, chunksi
|
|
91 |
'abs_categories': 'secondary_subfield'
|
92 |
}, inplace=True)
|
93 |
# ignore entries from ignored categories (i.e., primary subfield)
|
94 |
-
df = original_df[~original_df['field'].isin(IGNORED_CATEGOREIES)]
|
95 |
-
df = df[(df['version'] == 1)]
|
96 |
if df['primary_subfield'].isin(CATEGORY_ALIASES.keys()).sum():
|
97 |
ValueError("IMPLEMENT CODE TO CHANGE CATEGORY_ALIASES ")
|
98 |
|
99 |
-
# # preprocess the chunk here
|
100 |
-
# df_chunk = df.astype(str)
|
101 |
-
# # convert the DataFrame to a JSON string
|
102 |
-
# json_str = df_chunk.to_json(orient='records', lines=True)
|
103 |
-
|
104 |
-
# # write the JSON string to file
|
105 |
-
# f.write(json_str)
|
106 |
-
|
107 |
-
# # add a newline character after each chunk except the last one
|
108 |
-
# # if i != num_chunks - 1:
|
109 |
-
# f.write('\n')
|
110 |
if i == 0:
|
111 |
-
df.
|
112 |
else:
|
113 |
-
df.
|
114 |
|
115 |
# %%
|
|
|
37 |
import numpy as np
|
38 |
import os
|
39 |
from tqdm import tqdm
|
40 |
+
from arxiv_classifier.utils import (CATEGORY_ALIASES,
|
41 |
+
GENERAL_CATEGORIES,
|
42 |
+
IGNORED_CATEGOREIES)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
|
44 |
# %% [markdown]
|
45 |
# Rename columns to reflect standarized terminology:
|
|
|
60 |
save_path = os.path.join(save_dir, f"{split}.json")
|
61 |
print("Reading from file: ", file_path)
|
62 |
|
63 |
+
for i, original_df in tqdm(enumerate(pd.read_json(file_path, lines=True, chunksize=10**4, dtype=str))):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
64 |
# Rename columns
|
65 |
original_df.rename(columns={
|
66 |
'major_category': 'field',
|
|
|
68 |
'abs_categories': 'secondary_subfield'
|
69 |
}, inplace=True)
|
70 |
# ignore entries from ignored categories (i.e., primary subfield)
|
71 |
+
df = original_df[~original_df['field'].isin(GENERAL_CATEGORIES | IGNORED_CATEGOREIES)]
|
|
|
72 |
if df['primary_subfield'].isin(CATEGORY_ALIASES.keys()).sum():
|
73 |
ValueError("IMPLEMENT CODE TO CHANGE CATEGORY_ALIASES ")
|
74 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
if i == 0:
|
76 |
+
df.to_json(save_path, lines=True, orient="records")
|
77 |
else:
|
78 |
+
df.to_json(save_path, lines=True, orient="records", mode='a')
|
79 |
|
80 |
# %%
|
test_paper_id.py
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# script to test for any paper id overlap between the training and test set
|
2 |
+
# to run:
|
3 |
+
# python test_paper_id.py
|
4 |
+
|
5 |
+
import pandas as pd
|
6 |
+
import re
|
7 |
+
from tqdm import tqdm
|
8 |
+
|
9 |
+
def test(dataset):
|
10 |
+
print(f"Checking dataset: {dataset}")
|
11 |
+
# load the training and test data
|
12 |
+
train_file = f"data/raw/{dataset}/train_{dataset}_cats_full.json"
|
13 |
+
# read json in chunks of 1000 rows, get the paper_id column and convert to list
|
14 |
+
train_ids = []
|
15 |
+
for df in tqdm(pd.read_json(train_file, lines=True, chunksize=1000, dtype=str)):
|
16 |
+
train_ids.extend(df['paper_id'].tolist())
|
17 |
+
test_file = f"data/raw/{dataset}/test_{dataset}_cats_full.json"
|
18 |
+
test_ids = []
|
19 |
+
for df in tqdm(pd.read_json(test_file, lines=True, chunksize=1000, dtype=str)):
|
20 |
+
test_ids.extend(df['paper_id'].tolist())
|
21 |
+
|
22 |
+
# check paper id formats
|
23 |
+
ARXIV_IDENTIFIER_PATTERN = r"^\d{2}(0[1-9]|1[0-2])\.\d{4,5}$"
|
24 |
+
OLD_ARXIV_IDENTIFIER_PATTERN = r"^[\w\-]+/\d{7}$"
|
25 |
+
assert not any([re.match(ARXIV_IDENTIFIER_PATTERN, pid) is None for pid in train_ids if '.' in pid])
|
26 |
+
assert not any([re.match(ARXIV_IDENTIFIER_PATTERN, pid) is None for pid in test_ids if '.' in pid])
|
27 |
+
assert not any([re.match(OLD_ARXIV_IDENTIFIER_PATTERN, pid) is None for pid in train_ids if '/' in pid])
|
28 |
+
assert not any([re.match(OLD_ARXIV_IDENTIFIER_PATTERN, pid) is None for pid in test_ids if '/' in pid])
|
29 |
+
|
30 |
+
# check for overlap
|
31 |
+
overlap = set(train_ids).intersection(set(test_ids))
|
32 |
+
assert len(overlap) == 0, f"Overlap found: {overlap}"
|
33 |
+
|
34 |
+
if __name__ == "__main__":
|
35 |
+
test("major")
|
36 |
+
test("minor")
|
37 |
+
print("All tests passed!")
|