銀行對帳單解析基準測試
首個公開衡量銀行對帳單擷取準確度的基準測試。15 份合成對帳單、36 個解析挑戰、涵蓋 12 個國家，共 3 個難度等級。
快速開始
三步驟即可開始。
curl -O https://bankstatemently.com/benchmark/statements/bsb-001-statement.pdf
your-parser bsb-001-statement.pdf > result.jsonHASH=$(shasum -a 256 bsb-001-statement.pdf | cut -d' ' -f1)
curl -X POST https://api.bankstatemently.com/v1/benchmark/evaluate \
-H "Content-Type: application/json" \
-H "X-API-Key: your-api-key" \
-d "{\"contentHash\": \"$HASH\", \"transactions\": $(cat result.json)}"為什麼銀行對帳單解析如此困難
真實世界的銀行對帳單包含數十種格式怪癖，足以讓陽春解析器出錯。
Missing Year in Dates
Statements omit the year from transaction dates (e.g., "03/15" instead of "03/15/2025"), which makes it impossible to sort or reconcile without guessing.
出現於 4 份對帳單Blank Dates for Same-Day Transactions
Statements only print the date on the first transaction of each day — the rest have blank date cells.
出現於 1 份對帳單Plus/Minus Sign After the Amount
Statements place the plus or minus sign after the amount (e.g., "123.45-" instead of "-123.45").
出現於 1 份對帳單Currency Symbols Breaking Numeric Parsing
Statements embed currency symbols (e.g., "$1,234.56") directly in amount cells, which breaks numeric parsing in Excel and most CSV importers.
出現於 2 份對帳單No Visible Table Lines or Borders
Statements sometimes lack clear row or column lines.
出現於 1 份對帳單Separate Credit & Debit Columns
Statements use separate Credit and Debit columns in some formats, while others use a single Amount column with signs.
出現於 10 份對帳單Page Headers & Footers Mixed with Data
Page headers and footers contain bank logos, page numbers, and disclaimers that can interfere with transaction extraction.
出現於 1 份對帳單No Selectable Text (Scanned PDF)
Statements are often image-based PDFs with no selectable text, so copy-paste and standard PDF extractors return nothing useful.
出現於 3 份對帳單資料集
bsb-001（Straits Capital (Singapore)）已於 2026-08-11 重新發行。 Reissued to fix a benchmark-generator reproducibility bug (non-deterministic PDF metadata: creation date, embedded font-subset tags, image object ordering). Printed content and every ground-truth answer are unchanged. 原始 PDF sha256： 7f96da7316b2b540f2f8ecfc4151cd242a501be3ed84204a15bfbbe3c355531b. 已由此取代： 8a5b0e296d5d9ffaccfe6e8d99e0b52f7f60db76db54132c8c9ce6b8cbd52720. 內容未變更 — 針對原始 PDF 產生的解析或分數仍然有效，但其索引使用的是已被取代的雜湊值。請重新對應索引，而非重新解析。
bsb-002（Liberty National (United States)）已於 2026-08-11 重新發行。 Reissued to fix a benchmark-generator reproducibility bug (non-deterministic PDF metadata: creation date, embedded font-subset tags, image object ordering). Printed content and every ground-truth answer are unchanged. 原始 PDF sha256： f9947911a399ff0ccb7e8afde6e3380e3f7a1f61a5809b68022ce0b9ba449b06. 已由此取代： 7632b2a50b3ab172e18f03ea110241660c220a4ef738bbfb683eba4cccbbc324. 內容未變更 — 針對原始 PDF 產生的解析或分數仍然有效，但其索引使用的是已被取代的雜湊值。請重新對應索引，而非重新解析。
bsb-003（Continental Trust (Netherlands)）已於 2026-08-11 重新發行。 Reissued to fix a benchmark-generator reproducibility bug (non-deterministic PDF metadata: creation date, embedded font-subset tags, image object ordering). Printed content and every ground-truth answer are unchanged. 原始 PDF sha256： d605ff5e2c803f531c18d295354f135017c56082538f41e53d6c37452604a566. 已由此取代： 4bad73296bdbcf7142b62c8af421527f88d1d3a6865617e9fea48e6f51c704a1. 內容未變更 — 針對原始 PDF 產生的解析或分數仍然有效，但其索引使用的是已被取代的雜湊值。請重新對應索引，而非重新解析。
bsb-004（Silk Road Banking (Hong Kong)）已於 2026-08-07 重新發行。 Reissued to correct errors in the original statement. The second account is now denominated in USD, with a disclosed FX rate. 原始 PDF sha256： e93cacdb23195adefdf13ef4ba9528f55454c6fb2dc83624fe9c5077c5bc6455. 已由此取代： 522f8c9c92cb567580ffca4ac5ff9d4caea7d831764671de0ab8ae0878e513c2. 針對原始 PDF 產生的解析或分數不可比較。請重新解析已重新發行的 PDF。
bsb-004（Silk Road Banking (Hong Kong)）已於 2026-08-11 重新發行。 Reissued to fix the same generator reproducibility bug as the other released statements, plus a generator fix to how ground-truth account identity is derived: the account name no longer bakes in the account number, which lets a parser that correctly componentizes the printed heading exact-match cross-engine. Printed content (including the heading) is unchanged. 原始 PDF sha256： 522f8c9c92cb567580ffca4ac5ff9d4caea7d831764671de0ab8ae0878e513c2. 已由此取代： 0a4201b0e1a9c9ed834a2dce898a97ef8bac2b959ae6e3eb041c59328e0b19d1. 內容未變更 — 針對原始 PDF 產生的解析或分數仍然有效，但其索引使用的是已被取代的雜湊值。請重新對應索引，而非重新解析。
bsb-005（Harbour Bank (Canada)）已於 2026-08-11 重新發行。 Reissued to fix a benchmark-generator reproducibility bug (non-deterministic PDF metadata: creation date, embedded font-subset tags, image object ordering). Printed content and every ground-truth answer are unchanged. 原始 PDF sha256： 01a26c767b9488e45b358b329e2c9ea3d9fba05878a1963da06fde899827debb. 已由此取代： 8919aa5abdb06d377898e5cc78751661fcc80a25a89a477af49ceb0f0d7fe143. 內容未變更 — 針對原始 PDF 產生的解析或分數仍然有效，但其索引使用的是已被取代的雜湊值。請重新對應索引，而非重新解析。
bsb-001（Straits Capital (Singapore)）已於 2026-08-14 重新發行。 Reissued to fix account numbers that were previously a static, non-random digit string shared by every synthetic statement for that banking tag. Account numbers now come from a country-keyed format table (#4931). 原始 PDF sha256： 8a5b0e296d5d9ffaccfe6e8d99e0b52f7f60db76db54132c8c9ce6b8cbd52720. 已由此取代： d3d5071b90a33272f31bab5c5660d78b2ae29d2c6e1720477f5e8b167f73e843. 針對原始 PDF 產生的解析或分數不可比較。請重新解析已重新發行的 PDF。
bsb-002（Liberty National (United States)）已於 2026-08-14 重新發行。 Reissued to fix account numbers that were previously a static, non-random digit string shared by every synthetic statement for that banking tag. Account numbers now come from a country-keyed format table (#4931). 原始 PDF sha256： 7632b2a50b3ab172e18f03ea110241660c220a4ef738bbfb683eba4cccbbc324. 已由此取代： 3700a073ac1262173cbabbd5667cd3a59e04cf0b14e3a865ae25ce21089ab97d. 針對原始 PDF 產生的解析或分數不可比較。請重新解析已重新發行的 PDF。
bsb-003（Continental Trust (Netherlands)）已於 2026-08-14 重新發行。 Reissued to fix the account number: NL is an IBAN country, but the old template randomized a two-letter "country code" instead of emitting NL's literal IBAN prefix, and the checksum was never a valid ISO 7064 mod-97-10 IBAN check digit. Account numbers now come from a country-keyed format table that emits a real, check-digit-valid IBAN for NL (#4931). 原始 PDF sha256： 4bad73296bdbcf7142b62c8af421527f88d1d3a6865617e9fea48e6f51c704a1. 已由此取代： 4945bcda107e6d7845850e728b90c4dba1cb11de5a13356b0794df349e5527c2. 針對原始 PDF 產生的解析或分數不可比較。請重新解析已重新發行的 PDF。
bsb-004（Silk Road Banking (Hong Kong)）已於 2026-08-14 重新發行。 Reissued to fix account numbers that were previously a static, non-random digit string shared by every synthetic statement for that banking tag. Hong Kong has no single documented national account-number scheme, so the new generic digit format is honestly noted as such rather than citing a fabricated standard. Account numbers now come from a country-keyed format table (#4931). 原始 PDF sha256： 0a4201b0e1a9c9ed834a2dce898a97ef8bac2b959ae6e3eb041c59328e0b19d1. 已由此取代： 67a5b862696fc06e675db53867be6fcc71a4ecc13998193e8b45e2cd63371051. 針對原始 PDF 產生的解析或分數不可比較。請重新解析已重新發行的 PDF。
bsb-005（Harbour Bank (Canada)）已於 2026-08-14 重新發行。 Reissued to fix the account number: Canada has no IBAN, but the old template printed a French IBAN (inherited from the base `fr.json` dictionary — format was keyed by language, not jurisdiction). Account numbers are now keyed by country, so `fr-CA` prints Canada's 5-digit-transit + 3-digit-institution + 7-digit-account format instead of France's IBAN (#4931). 原始 PDF sha256： 8919aa5abdb06d377898e5cc78751661fcc80a25a89a477af49ceb0f0d7fe143. 已由此取代： a5cc2a7d210bc657c55b44ac159c5d10b2d71553f3646ce17a64051f918fd960. 針對原始 PDF 產生的解析或分數不可比較。請重新解析已重新發行的 PDF。
基礎5 份對帳單
中階5 份對帳單
MX銀行對帳單1 頁
CA信用卡對帳單1 頁
AU銀行對帳單1 頁
GB銀行對帳單1 頁
IN銀行對帳單27 頁進階5 份對帳單
HK銀行對帳單3 頁
SG信用卡對帳單1 頁
KZ銀行對帳單3 頁
TH銀行對帳單1 頁
MY信用卡對帳單1 頁評分方式
您的分數以兩個面向衡量 — 與我們用來 獨立測試商業轉換工具的框架相同.
擷取準確度
將日期、金額、描述與餘額逐欄與真實資料比對。
對帳單完整性
餘額核對、總額驗證，以及整份對帳單的逐列對齊。