Add script for raw metadata
Browse files- Amazon-Review-2023.py +66 -0
Amazon-Review-2023.py
ADDED
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
|
3 |
+
import datasets
|
4 |
+
|
5 |
+
|
6 |
+
_AMAZON_REVIEW_2023_DESCRIPTION = """\
|
7 |
+
Amazon Review 2023 is an updated version of the Amazon Review 2018 dataset.
|
8 |
+
This dataset mainly includes reviews (ratings, text) and item metadata (desc-
|
9 |
+
riptions, category information, price, brand, and images). Compared to the pre-
|
10 |
+
vious versions, the 2023 version features larger size, newer reviews (up to Sep
|
11 |
+
2023), richer and cleaner meta data, and finer-grained timestamps (from day to
|
12 |
+
milli-second).
|
13 |
+
|
14 |
+
"""
|
15 |
+
|
16 |
+
|
17 |
+
class RawMetaAmazonReview2023Config(datasets.BuilderConfig):
|
18 |
+
def __init__(self, **kwargs):
|
19 |
+
super(RawMetaAmazonReview2023Config, self).__init__(**kwargs)
|
20 |
+
|
21 |
+
self.suffix = 'jsonl'
|
22 |
+
self.domain = self.name[len(f'raw_meta_'):]
|
23 |
+
self.description = f'This is a subset for items in domain: {self.domain}.'
|
24 |
+
self.data_dir = f'raw/meta_categories/meta_{self.domain}.jsonl'
|
25 |
+
|
26 |
+
|
27 |
+
class AmazonReview2023(datasets.GeneratorBasedBuilder):
|
28 |
+
BUILDER_CONFIGS = [
|
29 |
+
RawMetaAmazonReview2023Config(
|
30 |
+
name='raw_meta_All_Beauty'
|
31 |
+
)
|
32 |
+
]
|
33 |
+
|
34 |
+
def _info(self):
|
35 |
+
return datasets.DatasetInfo(
|
36 |
+
description=_AMAZON_REVIEW_2023_DESCRIPTION + self.config.description
|
37 |
+
)
|
38 |
+
|
39 |
+
def _split_generators(self, dl_manager):
|
40 |
+
dl_dir = dl_manager.download_and_extract(self.config.data_dir)
|
41 |
+
return [
|
42 |
+
datasets.SplitGenerator(
|
43 |
+
name='full',
|
44 |
+
gen_kwargs={"filepath": dl_dir}
|
45 |
+
)
|
46 |
+
]
|
47 |
+
|
48 |
+
def _generate_examples(self, filepath):
|
49 |
+
with open(filepath, 'r', encoding='utf-8') as file:
|
50 |
+
for idx, line in enumerate(file):
|
51 |
+
if self.config.suffix == 'jsonl':
|
52 |
+
try:
|
53 |
+
dp = json.loads(line)
|
54 |
+
"""
|
55 |
+
For item metadata, 'details' is free-form structured data
|
56 |
+
Here we dump it to string to make huggingface datasets easy
|
57 |
+
to store.
|
58 |
+
"""
|
59 |
+
if isinstance(self.config, RawMetaAmazonReview2023Config) and \
|
60 |
+
'details' in dp:
|
61 |
+
dp['details'] = json.dumps(dp['details'])
|
62 |
+
except:
|
63 |
+
continue
|
64 |
+
else:
|
65 |
+
raise ValueError(f'Unknown suffix {self.config.suffix}.')
|
66 |
+
yield idx, dp
|