Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code:   DatasetGenerationError
Exception:    CastError
Message:      Couldn't cast
id: string
seed: int64
region: string
crossBorder: bool
seller: struct<name: string, taxId: string, address: list<item: string>, email: string, contact: string, ban (... 41 chars omitted)
  child 0, name: string
  child 1, taxId: string
  child 2, address: list<item: string>
      child 0, item: string
  child 3, email: string
  child 4, contact: string
  child 5, bank: struct<bank: string, account: string>
      child 0, bank: string
      child 1, account: string
buyer: struct<name: string, taxId: string, address: list<item: string>, email: string, contact: string, ban (... 41 chars omitted)
  child 0, name: string
  child 1, taxId: string
  child 2, address: list<item: string>
      child 0, item: string
  child 3, email: string
  child 4, contact: string
  child 5, bank: struct<bank: string, account: string>
      child 0, bank: string
      child 1, account: string
docs: list<item: struct<id: string, kind: string, number: string, date: timestamp[s], region: string, curr (... 1115 chars omitted)
  child 0, item: struct<id: string, kind: string, number: string, date: timestamp[s], region: string, currency: strin (... 1103 chars omitted)
      child 0, id: string
      child 1, kind: string
      child 2, number: string
      child 3, date: timestamp[s]
      child 4, region: string
      child 5, currency: string
      child 6, lang: string
      child 7, seller: struct<name: string, taxId: string, address: list<item: string>, email: string, contact: string, ban (... 41 ch
...

          child 3, Zero-rated export: string
      child 19, signature: struct<name: string, at: timestamp[s], remarks: string>
          child 0, name: string
          child 1, at: timestamp[s]
          child 2, remarks: string
      child 20, dueDate: timestamp[s]
      child 21, title: string
      child 22, body: list<item: string>
          child 0, item: string
packets: list<item: struct<id: string, type: string, parts: list<item: struct<docId: string, kind: string>>>>
  child 0, item: struct<id: string, type: string, parts: list<item: struct<docId: string, kind: string>>>
      child 0, id: string
      child 1, type: string
      child 2, parts: list<item: struct<docId: string, kind: string>>
          child 0, item: struct<docId: string, kind: string>
              child 0, docId: string
              child 1, kind: string
match: struct<decision: string, reasons: list<item: null>>
  child 0, decision: string
  child 1, reasons: list<item: null>
      child 0, item: null
invoiceApproval: list<item: struct<level: int64, role: string, name: string, decision: string, at: timestamp[s]>>
  child 0, item: struct<level: int64, role: string, name: string, decision: string, at: timestamp[s]>
      child 0, level: int64
      child 1, role: string
      child 2, name: string
      child 3, decision: string
      child 4, at: timestamp[s]
notes: list<item: string>
  child 0, item: string
path: string
chain: string
scenario: string
outcome: string
pages: int64
decision: string
to
{'chain': Value('string'), 'region': Value('string'), 'scenario': Value('string'), 'outcome': Value('string'), 'pages': Value('int64'), 'decision': Value('string'), 'path': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1835, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 621, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              id: string
              seed: int64
              region: string
              crossBorder: bool
              seller: struct<name: string, taxId: string, address: list<item: string>, email: string, contact: string, ban (... 41 chars omitted)
                child 0, name: string
                child 1, taxId: string
                child 2, address: list<item: string>
                    child 0, item: string
                child 3, email: string
                child 4, contact: string
                child 5, bank: struct<bank: string, account: string>
                    child 0, bank: string
                    child 1, account: string
              buyer: struct<name: string, taxId: string, address: list<item: string>, email: string, contact: string, ban (... 41 chars omitted)
                child 0, name: string
                child 1, taxId: string
                child 2, address: list<item: string>
                    child 0, item: string
                child 3, email: string
                child 4, contact: string
                child 5, bank: struct<bank: string, account: string>
                    child 0, bank: string
                    child 1, account: string
              docs: list<item: struct<id: string, kind: string, number: string, date: timestamp[s], region: string, curr (... 1115 chars omitted)
                child 0, item: struct<id: string, kind: string, number: string, date: timestamp[s], region: string, currency: strin (... 1103 chars omitted)
                    child 0, id: string
                    child 1, kind: string
                    child 2, number: string
                    child 3, date: timestamp[s]
                    child 4, region: string
                    child 5, currency: string
                    child 6, lang: string
                    child 7, seller: struct<name: string, taxId: string, address: list<item: string>, email: string, contact: string, ban (... 41 ch
              ...
              
                        child 3, Zero-rated export: string
                    child 19, signature: struct<name: string, at: timestamp[s], remarks: string>
                        child 0, name: string
                        child 1, at: timestamp[s]
                        child 2, remarks: string
                    child 20, dueDate: timestamp[s]
                    child 21, title: string
                    child 22, body: list<item: string>
                        child 0, item: string
              packets: list<item: struct<id: string, type: string, parts: list<item: struct<docId: string, kind: string>>>>
                child 0, item: struct<id: string, type: string, parts: list<item: struct<docId: string, kind: string>>>
                    child 0, id: string
                    child 1, type: string
                    child 2, parts: list<item: struct<docId: string, kind: string>>
                        child 0, item: struct<docId: string, kind: string>
                            child 0, docId: string
                            child 1, kind: string
              match: struct<decision: string, reasons: list<item: null>>
                child 0, decision: string
                child 1, reasons: list<item: null>
                    child 0, item: null
              invoiceApproval: list<item: struct<level: int64, role: string, name: string, decision: string, at: timestamp[s]>>
                child 0, item: struct<level: int64, role: string, name: string, decision: string, at: timestamp[s]>
                    child 0, level: int64
                    child 1, role: string
                    child 2, name: string
                    child 3, decision: string
                    child 4, at: timestamp[s]
              notes: list<item: string>
                child 0, item: string
              path: string
              chain: string
              scenario: string
              outcome: string
              pages: int64
              decision: string
              to
              {'chain': Value('string'), 'region': Value('string'), 'scenario': Value('string'), 'outcome': Value('string'), 'pages': Value('int64'), 'decision': Value('string'), 'path': Value('string')}
              because column names don't match
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1379, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 956, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1702, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1887, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

chain
string
region
string
scenario
string
outcome
string
pages
int64
decision
string
path
string
c000007
ID
missing-page
hold
5
hold
fabricated/ID/c000007/degraded/missing-page
c000007
ID
quantity-ticks
read_same
6
post
fabricated/ID/c000007/degraded/quantity-ticks
c000007
ID
receipt-box
read_same
6
post
fabricated/ID/c000007/degraded/receipt-box
c000007
ID
shuffled-pages
hold
6
hold
fabricated/ID/c000007/degraded/shuffled-pages
c000007
ID
stamped
read_same
6
post
fabricated/ID/c000007/degraded/stamped
c000019
ID
altered-total
hold
5
hold
fabricated/ID/c000019/degraded/altered-total
c000019
ID
folded
read_same
5
post
fabricated/ID/c000019/degraded/folded
c000019
ID
worst-case
hold
5
hold
fabricated/ID/c000019/degraded/worst-case
c000023
ID
quantity-written-short
hold
1
hold
fabricated/ID/c000023/degraded/quantity-written-short
c000023
ID
receipt-box
read_same
1
post
fabricated/ID/c000023/degraded/receipt-box
c000023
ID
thermal-fade
best_effort
1
post
fabricated/ID/c000023/degraded/thermal-fade
c000035
ID
duty-stamp
read_same
6
post
fabricated/ID/c000035/degraded/duty-stamp
c000035
ID
missing-page
hold
5
hold
fabricated/ID/c000035/degraded/missing-page
c000035
ID
shuffled-pages
hold
6
hold
fabricated/ID/c000035/degraded/shuffled-pages
c000035
ID
watermark
read_same
6
post
fabricated/ID/c000035/degraded/watermark
c000039
ID
blank-page
read_same
5
post
fabricated/ID/c000039/degraded/blank-page
c000039
ID
blur-motion
best_effort
4
post
fabricated/ID/c000039/degraded/blur-motion
c000039
ID
duty-stamp
read_same
4
post
fabricated/ID/c000039/degraded/duty-stamp
c000039
ID
missing-page
hold
3
hold
fabricated/ID/c000039/degraded/missing-page
c000039
ID
shuffled-pages
hold
4
hold
fabricated/ID/c000039/degraded/shuffled-pages
c000045
ID
stamp-over-total
hold
2
hold
fabricated/ID/c000045/degraded/stamp-over-total
c000045
ID
stamped
read_same
2
post
fabricated/ID/c000045/degraded/stamped
c000046
ID
checker-total
read_same
5
hold
fabricated/ID/c000046/degraded/checker-total
c000046
ID
paid-stamp
hold
5
hold
fabricated/ID/c000046/degraded/paid-stamp
c000046
ID
receipt-box-mismatch
hold
5
hold
fabricated/ID/c000046/degraded/receipt-box-mismatch
c000053
ID
checker-total
read_same
2
post
fabricated/ID/c000053/degraded/checker-total
c000053
ID
punched-stapled
read_same
2
post
fabricated/ID/c000053/degraded/punched-stapled
c000053
ID
scan
read_same
2
post
fabricated/ID/c000053/degraded/scan
c000065
ID
cropped
hold
4
hold
fabricated/ID/c000065/degraded/cropped
c000065
ID
digital-stamp-draft
hold
4
hold
fabricated/ID/c000065/degraded/digital-stamp-draft
c000065
ID
missing-page
hold
3
hold
fabricated/ID/c000065/degraded/missing-page
c000065
ID
quantity-written-short
hold
4
hold
fabricated/ID/c000065/degraded/quantity-written-short
c000065
ID
shuffled-pages
hold
4
hold
fabricated/ID/c000065/degraded/shuffled-pages
c000079
ID
altered-total
hold
5
hold
fabricated/ID/c000079/degraded/altered-total
c000079
ID
digital-stamp
read_same
5
hold
fabricated/ID/c000079/degraded/digital-stamp
c000079
ID
receipt-box
read_same
5
hold
fabricated/ID/c000079/degraded/receipt-box
c000086
ID
blur-motion
best_effort
11
post
fabricated/ID/c000086/degraded/blur-motion
c000086
ID
missing-page
hold
10
hold
fabricated/ID/c000086/degraded/missing-page
c000086
ID
shuffled-pages
hold
11
hold
fabricated/ID/c000086/degraded/shuffled-pages
c000086
ID
stained
best_effort
11
post
fabricated/ID/c000086/degraded/stained
c000086
ID
two-up
read_same
6
post
fabricated/ID/c000086/degraded/two-up
c000087
ID
blank-page
read_same
6
post
fabricated/ID/c000087/degraded/blank-page
c000087
ID
low-res
best_effort
5
post
fabricated/ID/c000087/degraded/low-res
c000087
ID
multi-signatures
read_same
5
post
fabricated/ID/c000087/degraded/multi-signatures
c000090
ID
approval-level-missing
hold
5
hold
fabricated/ID/c000090/degraded/approval-level-missing
c000090
ID
duplicate-page
read_same
6
post
fabricated/ID/c000090/degraded/duplicate-page
c000090
ID
folded
read_same
5
post
fabricated/ID/c000090/degraded/folded
c000096
ID
blank-page
read_same
6
post
fabricated/ID/c000096/degraded/blank-page
c000096
ID
crossed-out-line
hold
5
hold
fabricated/ID/c000096/degraded/crossed-out-line
c000096
ID
scan
read_same
5
post
fabricated/ID/c000096/degraded/scan
c000012
MY
annotated
read_same
2
post
fabricated/MY/c000012/degraded/annotated
c000012
MY
missing-page
hold
1
hold
fabricated/MY/c000012/degraded/missing-page
c000012
MY
multi-signatures
read_same
2
post
fabricated/MY/c000012/degraded/multi-signatures
c000012
MY
shuffled-pages
hold
2
hold
fabricated/MY/c000012/degraded/shuffled-pages
c000012
MY
watermark
read_same
2
post
fabricated/MY/c000012/degraded/watermark
c000018
MY
received-stamp
read_same
1
post
fabricated/MY/c000018/degraded/received-stamp
c000018
MY
skewed
read_same
1
post
fabricated/MY/c000018/degraded/skewed
c000018
MY
watermark
read_same
1
post
fabricated/MY/c000018/degraded/watermark
c000024
MY
altered-total
hold
2
hold
fabricated/MY/c000024/degraded/altered-total
c000024
MY
office-workflow
read_same
2
post
fabricated/MY/c000024/degraded/office-workflow
c000024
MY
photo-hard
best_effort
2
post
fabricated/MY/c000024/degraded/photo-hard
c000027
MY
paid-stamp
hold
2
hold
fabricated/MY/c000027/degraded/paid-stamp
c000027
MY
receipt-box
read_same
2
post
fabricated/MY/c000027/degraded/receipt-box
c000027
MY
stamp-over-total
hold
2
hold
fabricated/MY/c000027/degraded/stamp-over-total
c000028
MY
crossed-out-line
hold
1
hold
fabricated/MY/c000028/degraded/crossed-out-line
c000028
MY
punched-stapled
read_same
1
post
fabricated/MY/c000028/degraded/punched-stapled
c000028
MY
received-stamp
read_same
1
post
fabricated/MY/c000028/degraded/received-stamp
c000032
MY
blur-motion
best_effort
11
post
fabricated/MY/c000032/degraded/blur-motion
c000032
MY
missing-page
hold
10
hold
fabricated/MY/c000032/degraded/missing-page
c000032
MY
receipt-box
read_same
11
post
fabricated/MY/c000032/degraded/receipt-box
c000032
MY
shuffled-pages
hold
11
hold
fabricated/MY/c000032/degraded/shuffled-pages
c000049
MY
digital-stamp-void
hold
1
hold
fabricated/MY/c000049/degraded/digital-stamp-void
c000049
MY
photo-hard
best_effort
1
post
fabricated/MY/c000049/degraded/photo-hard
c000049
MY
quantity-written-short
hold
1
hold
fabricated/MY/c000049/degraded/quantity-written-short
c000059
MY
missing-page
hold
4
hold
fabricated/MY/c000059/degraded/missing-page
c000059
MY
office-workflow
hold
5
hold
fabricated/MY/c000059/degraded/office-workflow
c000059
MY
raster
read_same
5
hold
fabricated/MY/c000059/degraded/raster
c000059
MY
shuffled-pages
hold
5
hold
fabricated/MY/c000059/degraded/shuffled-pages
c000059
MY
watermark
read_same
5
hold
fabricated/MY/c000059/degraded/watermark
c000060
MY
office-workflow
read_same
5
post
fabricated/MY/c000060/degraded/office-workflow
c000060
MY
photo
read_same
5
post
fabricated/MY/c000060/degraded/photo
c000060
MY
thermal-fade
best_effort
5
post
fabricated/MY/c000060/degraded/thermal-fade
c000068
MY
duplicate-page
read_same
2
hold
fabricated/MY/c000068/degraded/duplicate-page
c000068
MY
quantity-written-short
hold
1
hold
fabricated/MY/c000068/degraded/quantity-written-short
c000068
MY
redacted
best_effort
1
hold
fabricated/MY/c000068/degraded/redacted
c000077
MY
annotated
read_same
5
post
fabricated/MY/c000077/degraded/annotated
c000077
MY
approval-stamps
read_same
5
post
fabricated/MY/c000077/degraded/approval-stamps
c000077
MY
multi-signatures
read_same
5
post
fabricated/MY/c000077/degraded/multi-signatures
c000001
PH
duplicate-page
read_same
7
hold
fabricated/PH/c000001/degraded/duplicate-page
c000001
PH
folded
read_same
6
hold
fabricated/PH/c000001/degraded/folded
c000001
PH
missing-page
hold
5
hold
fabricated/PH/c000001/degraded/missing-page
c000001
PH
shuffled-pages
hold
6
hold
fabricated/PH/c000001/degraded/shuffled-pages
c000001
PH
stamp-over-total
hold
6
hold
fabricated/PH/c000001/degraded/stamp-over-total
c000005
PH
cropped
hold
1
hold
fabricated/PH/c000005/degraded/cropped
c000005
PH
paid-stamp
hold
1
hold
fabricated/PH/c000005/degraded/paid-stamp
c000005
PH
stained
best_effort
1
post
fabricated/PH/c000005/degraded/stained
c000016
PH
altered-total
hold
1
hold
fabricated/PH/c000016/degraded/altered-total
c000016
PH
crossed-out-line
hold
1
hold
fabricated/PH/c000016/degraded/crossed-out-line
c000016
PH
stamp-over-total
hold
1
hold
fabricated/PH/c000016/degraded/stamp-over-total
c000017
PH
blank-page
read_same
3
post
fabricated/PH/c000017/degraded/blank-page
End of preview.

Fabricated business documents

Every document here is fabricated: companies, people, addresses, tax numbers, bank accounts and amounts are invented. Tax numbers have the shape of the real ones but are drawn at random, so one may by chance equal a real number; e-mail addresses end in .example. No real company's document was used. They are made by a seeded generator (fabricate/ in the Tern repository), so the same arguments give the same files.

What is in a chain

One buying-and-selling story per folder, from a request for quotation to the invoice and its match:

rfq → quotation → purchase_order (with its approval levels) → sales_order → proforma_invoice (some) → delivery_note → proof_of_delivery (signed) → goods_receipt → invoice → tax_invoice (Indonesia) → credit_note (some), and the packet the supplier sends: the invoice alone, with a tax invoice, with a proof of delivery, or bundled with the order, the delivery note and an attachment, as one multi-page PDF.

Layout

fabricated/<REGION>/<chain id>/
  chain.json            the whole story: parties, every document, the expected match, notes on what was done to it
  packet.pdf            what the supplier sends, one file
  packet.json           the parts of packet.pdf with their page ranges, and the expected decision (post or hold, with reasons)
  documents/NN-<kind>.pdf   each document on its own
  documents/NN-<kind>.json  its ground truth: every field and every line
  degraded/<scenario>/packet.pdf    the same packet as a bad document: scan, fax, phone photo, blur, scratches, stains, stamps, a struck-through line, an altered total, a missing page...
  degraded/<scenario>/degrade.json  what was done to which page, where each page came from, and the right handling (read_same, best_effort or hold, with reasons)
manifest.jsonl          one line per chain
degraded.jsonl          one line per bad-document packet
real/                   reserved; empty unless real files have been cleared for publication, and then marked as such

Regions and what varies

Indonesia, Singapore, Malaysia, Thailand, Philippines, Vietnam, United States. Five page layouts, long tables over several pages with the sum carried forward, labels in the local language or English, local number formats, tax rules (a separate tax invoice in Indonesia, zero-rated cross-border sales in USD). About a quarter of the invoices have something wrong on purpose: a price above the order, more billed than received, a missing order number, a total that does not add up, a notice changing the bank account, an order missing its last approval. packet.json says what is wrong, so a system's decision can be scored.

Limits

These are drawn documents. The degraded/ packets add drawn damage (blur, stains, strikes, stamps, missing pages), which is not the same as photographs of real paper: real stamps, handwriting and odd layouts are the hard part and are only partly here. This release: 100 chains, 18 of them with something wrong, 40 with an invoice of more than 12 lines, packet types full_bundle, single, invoice_support, invoice_tax.

Downloads last month
243