Datasets:
create
Browse files- dataset.py +92 -0
dataset.py
ADDED
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import json
|
3 |
+
import datasets
|
4 |
+
|
5 |
+
logger = datasets.logging.get_logger(__name__)
|
6 |
+
|
7 |
+
_CITATION = """\
|
8 |
+
@dataset{gotthatdata_stargate_2024,
|
9 |
+
title = {STARGATE: CIA Remote Viewing Archive},
|
10 |
+
author = {GotThatData},
|
11 |
+
year = {2024},
|
12 |
+
url = {https://huggingface.co/datasets/GotThatData/STARGATE}
|
13 |
+
}
|
14 |
+
"""
|
15 |
+
|
16 |
+
_DESCRIPTION = """\
|
17 |
+
STARGATE is a dataset of 12,000+ declassified CIA PDFs related to remote viewing (RV), extrasensory perception (ESP), and anomalous cognition.
|
18 |
+
This loader includes structured metadata and binary access to the original scanned PDFs.
|
19 |
+
"""
|
20 |
+
|
21 |
+
_HOMEPAGE = "https://huggingface.co/datasets/GotThatData/STARGATE"
|
22 |
+
|
23 |
+
_LICENSE = "CC-BY-4.0"
|
24 |
+
|
25 |
+
class StargatePDFConfig(datasets.BuilderConfig):
|
26 |
+
def __init__(self, **kwargs):
|
27 |
+
super(StargatePDFConfig, self).__init__(**kwargs)
|
28 |
+
|
29 |
+
class StargatePDFDataset(datasets.GeneratorBasedBuilder):
|
30 |
+
VERSION = datasets.Version("1.0.0")
|
31 |
+
BUILDER_CONFIGS = [
|
32 |
+
StargatePDFConfig(name="default", version=VERSION, description="STARGATE raw PDFs with metadata")
|
33 |
+
]
|
34 |
+
|
35 |
+
def _info(self):
|
36 |
+
return datasets.DatasetInfo(
|
37 |
+
description=_DESCRIPTION,
|
38 |
+
features=datasets.Features({
|
39 |
+
"filename": datasets.Value("string"),
|
40 |
+
"document_id": datasets.Value("string"),
|
41 |
+
"page_count": datasets.Value("int32"),
|
42 |
+
"image_count": datasets.Value("int32"),
|
43 |
+
"processed_at": datasets.Value("string"),
|
44 |
+
"ocr_status": datasets.Value("string"),
|
45 |
+
"text_extracted": datasets.Value("bool"),
|
46 |
+
"source": datasets.Value("string"),
|
47 |
+
"tags": datasets.Sequence(datasets.Value("string")),
|
48 |
+
"pdf": datasets.Value("binary"),
|
49 |
+
}),
|
50 |
+
supervised_keys=None,
|
51 |
+
homepage=_HOMEPAGE,
|
52 |
+
license=_LICENSE,
|
53 |
+
citation=_CITATION,
|
54 |
+
)
|
55 |
+
|
56 |
+
def _split_generators(self, dl_manager):
|
57 |
+
# Path to local or downloaded files
|
58 |
+
archive_path = dl_manager.download_and_extract("./") # Change if remote
|
59 |
+
metadata_path = os.path.join(archive_path, "metadata.json")
|
60 |
+
data_dir = os.path.join(archive_path, "data")
|
61 |
+
|
62 |
+
return [
|
63 |
+
datasets.SplitGenerator(
|
64 |
+
name=datasets.Split.TRAIN,
|
65 |
+
gen_kwargs={"metadata_path": metadata_path, "data_dir": data_dir}
|
66 |
+
)
|
67 |
+
]
|
68 |
+
|
69 |
+
def _generate_examples(self, metadata_path, data_dir):
|
70 |
+
logger.info(f"⏳ Loading metadata from {metadata_path}")
|
71 |
+
with open(metadata_path, "r", encoding="utf-8") as f:
|
72 |
+
records = json.load(f)
|
73 |
+
|
74 |
+
for idx, record in enumerate(records):
|
75 |
+
pdf_path = os.path.join(data_dir, record["filename"])
|
76 |
+
if not os.path.isfile(pdf_path):
|
77 |
+
logger.warning(f"🚫 Missing PDF: {pdf_path}")
|
78 |
+
continue
|
79 |
+
|
80 |
+
with open(pdf_path, "rb") as pdf_file:
|
81 |
+
yield idx, {
|
82 |
+
"filename": record.get("filename"),
|
83 |
+
"document_id": record.get("document_id", record["filename"].replace(".pdf", "")),
|
84 |
+
"page_count": record.get("page_count", 0),
|
85 |
+
"image_count": record.get("image_count", 0),
|
86 |
+
"processed_at": record.get("processed_at", ""),
|
87 |
+
"ocr_status": record.get("ocr_status", "pending"),
|
88 |
+
"text_extracted": record.get("text_extracted", False),
|
89 |
+
"source": record.get("source", "CIA Stargate Archive"),
|
90 |
+
"tags": record.get("tags", []),
|
91 |
+
"pdf": pdf_file.read(),
|
92 |
+
}
|