Datasets:

mic7ch
/

manchu-2025-0033

im imagewidth (px) 480 480	roman stringlengths 0 22	manchu stringlengths 1 20
	FIKŠEHE	ᠸᡳᡴᡧᡝᡥᡝ
	SABSIRA	ᠰᠠᠪᠰᡳᡵᠠ
	ACINGGIYABUHANGGE	ᠠᠴᡳᠩᡤᡳᠶᠠᠪᡠᡥᠠᠩᡤᡝ
	EDUNGGIYEBUTELE	ᡝᡩᡠᠩᡤᡳᠶᡝᠪᡠᡨᡝᠯᡝ
	SUJUBURAKŪNGGE	ᠰᡠᠵᡠᠪᡠᡵᠠᡴᡡᠩᡤᡝ
	TAŠARABURAKŪNGGE	ᡨᠠᡧᠠᡵᠠᠪᡠᡵᠠᡴᡡᠩᡤᡝ
	PORPONOHOI	ᡦᠣᡵᡦᠣᠨᠣᡥᠣᡳ
	CIFELEMAHABI	ᠴᡳᡶᡝᠯᡝᠮᠠᡥᠠᠪᡳ
	LIYELIYEŠERAKŪ	ᠯᡳᠶᡝᠯᡳᠶᡝᡧᡝᡵᠠᡴᡡ
	UNUHAKŪNGGE	ᡠᠨᡠᡥᠠᡴᡡᠩᡤᡝ
	DEMNIYEMBIHE	ᡩᡝᠮᠨᡳᠶᡝᠮᠪᡳᡥᡝ
	TENGGELJEHEBIHE	ᡨᡝᠩᡤᡝᠯᠵᡝᡥᡝᠪᡳᡥᡝ
	EBICINA	ᡝᠪᡳᠴᡳᠨᠠ
	JERGICELEME	ᠵᡝᡵᡤᡳᠴᡝᠯᡝᠮᡝ
	ISINJIMBIHE	ᡳᠰᡳᠨᠵᡳᠮᠪᡳᡥᡝ
	DURULEKI	ᡩᡠᡵᡠᠯᡝᡴᡳ
	FERGUWEHEBIHE	ᡶᡝᡵᡤᡠᠸᡝᡥᡝᠪᡳᡥᡝ
	FANGŠAME	ᡶᠠᠩᡧᠠᠮᡝ
	BUYENDURAKŪ	ᠪᡠᠶᡝᠨᡩᡠᡵᠠᡴᡡ
	ETUHE	ᡝᡨᡠᡥᡝ
	URUHAKŪNGGE	ᡠᡵᡠᡥᠠᡴᡡᠩᡤᡝ
	YARU	ᠶᠠᡵᡠ
	WERDEHEBI	ᠸᡝᡵᡩᡝᡥᡝᠪᡳ
	MERGEDEHAKŪ	ᠮᡝᡵᡤᡝᡩᡝᡥᠠᡴᡡ
	TABCILANJIRANGGE	ᡨᠠᠪᠴᡳᠯᠠᠨᠵᡳᡵᠠᠩᡤᡝ
	TEDERI	ᡨᡝᡩᡝᡵᡳ
	LESEREMAHABI	ᠯᡝᠰᡝᡵᡝᠮᠠᡥᠠᠪᡳ
	BAITALABURAKŪ	ᠪᠠᡳᡨᠠᠯᠠᠪᡠᡵᠠᡴᡡ
	ŠUKUKINI	ᡧᡠᡴᡠᡴᡳᠨᡳ
	GAIHABIHE	ᡤᠠᡳᡥᠠᠪᡳᡥᡝ
	INJEBUHEBIHE	ᡳᠨᠵᡝᠪᡠᡥᡝᠪᡳᡥᡝ
	GŪWANUME	ᡤᡡᠸᠠᠨᡠᠮᡝ
	GEDUREHE	ᡤᡝᡩᡠᡵᡝᡥᡝ
	AINAMBI	ᠠᡳᠨᠠᠮᠪᡳ
	ACINGGIYABUCI	ᠠᠴᡳᠩᡤᡳᠶᠠᠪᡠᠴᡳ
	JASIRA	ᠵᠠᠰᡳᡵᠠ
	TOHAKŪNGGE	ᡨᠣᡥᠠᡴᡡᠩᡤᡝ
	GIYALABURANGGE	ᡤᡳᠶᠠᠯᠠᠪᡠᡵᠠᠩᡤᡝ
	GURUNECI	ᡤᡠᡵᡠᠨᡝᠴᡳ
	JURCENJEMBI	ᠵᡠᡵᠴᡝᠨᠵᡝᠮᠪᡳ
	IKŪRŠAHABI	ᡳᡴᡡᡵᡧᠠᡥᠠᠪᡳ
	JUWABURAKŪ	ᠵᡠᠸᠠᠪᡠᡵᠠᡴᡡ
	YABILANGGALA	ᠶᠠᠪᡳᠯᠠᠩᡤᠠᠯᠠ
	FOSOBURONGGE	ᠸᠣᠰᠣᠪᡠᡵᠣᠩᡤᡝ
	SUCUNUHEBIHE	ᠰᡠᠴᡠᠨᡠᡥᡝᠪᡳᡥᡝ
	MAHŪLABURANGGE	ᠮᠠᡥᡡᠯᠠᠪᡠᡵᠠᠩᡤᡝ
	JAFUNUBUFI	ᠵᠠᠸᡠᠨᡠᠪᡠᠸᡳ
	SUKJIBURAKŪNGGE	ᠰᡠᡴᠵᡳᠪᡠᡵᠠᡴᡡᠩᡤᡝ
	ŠURGABUHABI	ᡧᡠᡵᡤᠠᠪᡠᡥᠠᠪᡳ
	WECE	ᠸᡝᠴᡝ
	HUWEKIYETELE	ᡥᡠᠸᡝᡴᡳᠶᡝᡨᡝᠯᡝ
	MIDAHA	ᠮᡳᡩᠠᡥᠠ
	ŠORGINACI	ᡧᠣᡵᡤᡳᠨᠠᠴᡳ
	MOSELABURANGGE	ᠮᠣᠰᡝᠯᠠᠪᡠᡵᠠᠩᡤᡝ
	EDUNGGE	ᡝᡩᡠᠩᡤᡝ
	TUMPANAHA	ᡨᡠᠮᡦᠠᠨᠠᡥᠠ
	WAHIYAKI	ᠸᠠᡥᡳᠶᠠᡴᡳ
	LOKSOBUFI	ᠯᠣᡴᠰᠣᠪᡠᠸᡳ
	EDUHAKŪ	ᡝᡩᡠᡥᠠᡴᡡ
	SAHARARAKŪNGGE	ᠰᠠᡥᠠᡵᠠᡵᠠᡴᡡᠩᡤᡝ
	TAMALIRAKŪ	ᡨᠠᠮᠠᠯᡳᡵᠠᡴᡡ
	TAKSIBUHAKŪ	ᡨᠠᡴᠰᡳᠪᡠᡥᠠᡴᡡ
	TEYEHEBI	ᡨᡝᠶᡝᡥᡝᠪᡳ
	SONORAKŪNGGE	ᠰᠣᠨᠣᡵᠠᡴᡡᠩᡤᡝ
	HALANJIRA	ᡥᠠᠯᠠᠨᠵᡳᡵᠠ
	FURUNAKINI	ᠸᡠᡵᡠᠨᠠᡴᡳᠨᡳ
	IDARŠATAI	ᡳᡩᠠᡵᡧᠠᡨᠠᡳ
	JALBARI	ᠵᠠᠯᠪᠠᡵᡳ
	SUWALIYABUCINA	ᠰᡠᠸᠠᠯᡳᠶᠠᠪᡠᠴᡳᠨᠠ
	SURGI	ᠰᡠᡵᡤᡳ
	FATABUHABI	ᡶᠠᡨᠠᠪᡠᡥᠠᠪᡳ
	ŠEŠEHEBIHE	ᡧᡝᡧᡝᡥᡝᠪᡳᡥᡝ
	WAIKIYAHAKŪNGGE	ᠸᠠᡳᡴᡳᠶᠠᡥᠠᡴᡡᠩᡤᡝ
	HOIFALAHA	ᡥᠣᡳᡶᠠᠯᠠᡥᠠ
	HAFIRŠANUHABI	ᡥᠠᠸᡳᡵᡧᠠᠨᡠᡥᠠᠪᡳ
	AŠUHAI	ᠠᡧᡠᡥᠠᡳ
	SUNTALACINA	ᠰᡠᠨᡨᠠᠯᠠᠴᡳᠨᠠ
	KOKIRANDUCI	ᡴᠣᡴᡳᡵᠠᠨᡩᡠᠴᡳ
	TEMŠEBUHAKŪ	ᡨᡝᠮᡧᡝᠪᡠᡥᠠᡴᡡ
	URHUBUMBIHE	ᡠᡵᡥᡠᠪᡠᠮᠪᡳᡥᡝ
	KIMCICINA	ᡴᡳᠮᠴᡳᠴᡳᠨᠠ
	JUMANGGILARANGGE	ᠵᡠᠮᠠᠩᡤᡳᠯᠠᡵᠠᠩᡤᡝ
	UHELEHEBI	ᡠᡥᡝᠯᡝᡥᡝᠪᡳ
	WENIYEBUHEBIHE	ᠸᡝᠨᡳᠶᡝᠪᡠᡥᡝᠪᡳᡥᡝ
	TARHŪKAN	ᡨᠠᡵᡥᡡᡴᠠᠨ
	DAŠAN	ᡩᠠᡧᠠᠨ
	HAFIRAHA	ᡥᠠᠸᡳᡵᠠᡥᠠ
	FUJIYANG	ᠸᡠᠵᡳᠶᠠᠩ
	MELJEHEBI	ᠮᡝᠯᠵᡝᡥᡝᠪᡳ
	ULEJEME	ᡠᠯᡝᠵᡝᠮᡝ
	DEOCILEBUMBIHE	ᡩᡝᠣᠴᡳᠯᡝᠪᡠᠮᠪᡳᡥᡝ
	BIYARIŠACINA	ᠪᡳᠶᠠᡵᡳᡧᠠᠴᡳᠨᠠ
	ŠARHŪHABI	ᡧᠠᡵᡥᡡᡥᠠᠪᡳ
	DUILEBUHENGGE	ᡩᡠᡳᠯᡝᠪᡠᡥᡝᠩᡤᡝ
	DEDUNEMBI	ᡩᡝᡩᡠᠨᡝᠮᠪᡳ
	KEIKEDECI	ᡴᡝᡳᡴᡝᡩᡝᠴᡳ
	GOHONGGE	ᡤᠣᡥᠣᠩᡤᡝ
	DOKSIDACINA	ᡩᠣᡴᠰᡳᡩᠠᠴᡳᠨᠠ
	MONJIBUME	ᠮᠣᠨᠵᡳᠪᡠᠮᡝ
	WAIKURŠAHANGGE	ᠸᠠᡳᡴᡠᡵᡧᠠᡥᠠᠩᡤᡝ

End of preview. Expand in Data Studio

Dataset Card

Dataset Summary

This dataset is a word-level Manchu OCR corpus combining:

Synthetic training and validation splits derived from the open-source ManchuOCR project by Dr. Zhang Zhuohui, and
A real-world historical test split curated from Manchu block-prints, archival manuscripts, and palace memorials from the 17th–19th centuries.

Each example consists of:

im: a word-level image patch (Manchu script, vertical writing),
roman: the corresponding romanization, and
manchu: the Unicode Manchu string.

Dataset Structure

Features

im: image — word-level image crops (PNG/JPEG, normalized during preprocessing).
roman: string — romanized Manchu (Latin script).
manchu: string — Manchu text in Unicode.

Splits

train: 60,000 synthetic word images
validation: 15,000 synthetic word images
test: 753 real historical word images

The training and validation splits are synthetic but font- and style-diverse; the test split is entirely real-world data.

Source Data

Training and Validation Splits (Synthetic)

The training and validation splits are derived from the open-source ManchuOCR synthetic dataset created by Dr. Zhang Zhuohui:

https://github.com/tyotakuki/ManchuOCR

The original ManchuOCR dataset contains approximately 7,500,000 synthetic word images generated from multiple fonts and augmentation pipelines.
From this corpus, we randomly sampled 75,000 examples, which were then split into:
- 60,000 examples for train, and
- 15,000 examples for validation.
We applied substantial additional image processing:

No manual changes were made to the underlying text; modifications are limited to sampling and preprocessing.

Test Split (Historical Sources)

The test split (753 examples) is constructed from real historical Manchu sources, including block-prints, veritable records, archival compilations, and palace memorials from the 17th–19th centuries. Word images were manually extracted from high-resolution scans, and then preprocessed and segmented with a semi-automated pipeline.

The following works and archival collections are represented:

1. Baqi tongzhi chuji 八旗通志初集 (General Gazetteer of the Eight Banners, 1739)

General Gazetteer of the Eight Banners (1739). General gazetteer of the eight banners (han i araha jakûn gûsai tung j’i bithe, 《八旗通志初集》). Kyoto Neifu block-print edition; Berlin State Library – Prussian Cultural Heritage Foundation digital edition (2016). Vol. 178.

2. Qing shilu 清實錄 — Veritable Records of the Taizong Emperor (Manchu Edition)

Veritable Records of the Taizong Emperor, Manchu Edition (n.d.). Daicing gurun i taidzung genggiyen šu hūwangdi i yargiyan kooli (《大清太宗文皇帝滿文實錄卷五：天聰三年正月至十二月》; accessed 20 November 2025). Manchu-language text, vol. 5 (册1), section for Tiancong 3rd year, 8th month. National Palace Museum archival ID 故官012620 (件10), digital file K4A012620-009.pdf; pp. 2–4.

3. Neige cangben Manwen laodang 內閣藏本滿文老檔 (Old Manchu Archives from the Grand Secretariat, 20 vols.)

Wu, Y., Y. Zhang, and Editorial Committee of Neige cangben Manwen laodang (Eds.) (2009). Old Manchu Archives from the Grand Secretariat (內閣藏本滿文老檔). Shenyang: Liaoning minzu chubanshe. 20 vols.

4. Meadows (1849), Translations from the Manchu

Meadows, T. T. (1849). Translations from the Manchu: With the Original Texts, Prefaced by an Essay on the Language (《清文敘略》 Manju gisun be majige gisurehe bithe). Canton: Press of S. Wells Williams. Digital edition available via Internet Archive; accessed 20 November 2025.

5. Manchu Palace Memorials — Qianlong reign (乾隆朝)

Disangga (1736). Manchu palace memorials, Qianlong reign (《宮中檔滿文奏摺‧乾隆朝》). Memorial by Disangga, titled 〈奏報雍正十三年比較堪看來年乾隆元年農祥雨水收成摺〉, dated Yongzheng 13.12.29 (10 February 1736). NPM archival ID: Gu-gong 157489, item 1, digital file K4D157489-0.pdf; accessed 20 November 2025.

6. Manchu Palace Memorials — Xianfeng reign (咸豐朝)

Peiceng (1849). Manchu palace memorials, Xianfeng reign (《宮中檔滿文奏摺‧咸豐朝》). Memorial by Peiceng, titled 〈奏謝天恩並報到任日期〉, dated Daoguang 29.08.24 (10 October 1849). NPM archival ID: Gu-gong 157901, item 1, digital file K4D157901-0.pdf; accessed 20 November 2025.

7. Manchu Palace Memorials — Guangxu catalog assignment (光緒朝; Yongzheng-era memorial)

Furdan (1728). Manchu palace memorials, Guangxu reign (《宮中檔滿文奏摺‧光緒朝》). Memorial by Furdan, titled 〈奏謝天恩賞戴二眼花翎〉, dated Yongzheng 05.12.04 (14 January 1728). Although catalogued under the Guangxu reign, the memorial date corresponds to the Yongzheng period; this discrepancy reflects metadata in the digital archive. NPM archival ID: Gu-gong 158156, item 1, digital file K4D158156-0.pdf; accessed 20 November 2025.