引用与致谢

本资源由 Khan 选编、数字化与翻译，汇集公有领域梵语辞书、OCR 校理的文献、开放语料与语法分析（DCS、Héritage、ByT5、CBETA、SuttaCentral 等），辅以 MITRA 机器翻译、以及自制教学胶囊与合成音频。下方 colophon 逐项列出每个第三方来源及其许可。

如何引用本资源

若您在研究或教学中使用了本站，敬请按下列格式引用：

Khan (2026). Kosa (वाणी) [Digital Sanskrit corpus & lexical database]. https://kosa.asphodelus.me

梵英／梵德／梵法等辞书（Monier-Williams、Apte、Böhtlingk-Roth 等）的数字化文本来自科隆数字梵语词典（Cologne Digital Sanskrit Dictionaries, CDSL，科隆大学）。
Héritage du Sanskrit 梵语词典，以及变格／变位语法引擎，系 Gérard Huet（sanskrit.inria.fr）之作品。本站的英、泰、中译本为在其基础上的派生翻译。
各部源辞书（19 世纪至 20 世纪初）本身属公有领域学术成果，其原作者已标注于词典库各条目。

下列各项各依其自身许可，未被本站原创许可重新许可；标「须署名／相同方式共享／非商业／禁改」者请按其条款使用。

Cologne Digital Sanskrit Dictionaries (CDSL), Universität zu Köln — © The Sanskrit Library & Thomas Malten — CC BY-NC-SA 3.0 — sanskrit-lexicon.uni-koeln.de
The CDSL licence covers Cologne’s digitisation of public-domain lexica (Monier-Williams, Apte 1890, Böhtlingk-Roth, Cappeller, Macdonell, etc.). Individual dictionaries that are themselves still in copyright are listed under “Restricted” below, not under this badge.
Sanskrit Heritage — Héritage du Sanskrit & declension/conjugation engine — © Gérard Huet — LGPLLR — sanskrit.inria.fr
Public-domain source lexica (pre-1929): Monier-Williams 1872/1899, Apte 1890, Macdonell 1893, Böhtlingk-Roth, Cappeller, Benfey, Wilson, Yates, classical kośas (Amara, Hemacandra, Halāyudha, Śabdakalpadruma, Vācaspatyam) — public domain — original authors credited per entry.
Monier-Williams 1899 → Thai translation — original translation by Khan of a public-domain base — © Khan, CC BY-NC-SA 4.0.

Digital Corpus of Sanskrit (DCS), Oliver Hellwig — CC BY 4.0 — github.com/OliverHellwig/sanskrit
Per-word lemma & morphological analysis for ~270 texts and the Bhagavadgītā / Nala readers; data reformatted for display.
ByT5-Sanskrit (Nehrdich, Hellwig & Keutzer, UC Berkeley BALab) — machine-generated segmentation, lemmatisation & morphosyntactic tagging for texts without DCS gold annotation, and the analyser behind the /ai-demo grammar-analysis panel — repository licence not stated by upstream; cite the paper below — github.com (byt5-sanskrit-analyzers)
Nehrdich, S., Hellwig, O., & Keutzer, K. (2024). One Model is All You Need: ByT5-Sanskrit, a Unified Model for Sanskrit NLP Tasks. Findings of EMNLP 2024. — machine-generated, non-gold-standard; labelled as such wherever shown, never presented as equivalent to DCS gold annotation.
M. R. Kale, A Higher Sanskrit Grammar, 7th ed., Bombay: Gopāl Nārāyen & Co., 1931 — public domain
作者身故后才出版（标题页署「Late」），已于泰国（殁后 50 年）、印度（60 年）等法域进入公有领域；本站的泰语阐述以 Kale 的分节为引用骨架自行撰写，非其散文之翻译。
GRETIL, Göttingen — scholarly, non-commercial use — gretil.sub.uni-goettingen.de
Underlying e-texts and gap-verse fill; courtesy credit.
Underlying Sanskrit texts (Ṛgveda, Mahābhārata, Rāmāyaṇa, purāṇas, śāstra, kāvya…) — public domain (ancient/medieval).

MITRA / Dharmamitra parallel dataset (BALab, UC Berkeley — Nehrdich, Keutzer et al.) — CC BY-SA 4.0 · ShareAlike — kept as an isolated module — dharmamitra.org
MITRA machine-translation output (self-hosted Gemma-2-MITRA) — Gemma Terms of Use, applied to public-domain source text.
SuttaCentral — Bhikkhu Sujato translations + Mahāsaṅgīti Pali — CC0 1.0 — suttacentral.net
CBETA — Chinese Buddhist Electronic Text Association — CC BY-NC-SA 3.0 TW — cbeta.org
Itihasa MBh/Rāmāyaṇa EN–SA parallel corpus (Aralikatte et al.) — Apache-2.0; underlying M. N. Dutt translation public domain.
Vidyut (Ambuda) — MIT · VedaWeb (Cologne) — per-source CC.

English translation of the Mūlamadhyamakakārikā — Stephen Batchelor (via WisdomLib) — CC BY-NC-SA 3.0
Recitation audio — BodhiSvara (reciter: Prof. Kashinath Nyaupane) — CC BY-NC-ND 4.0 · NoDerivatives — streamed unmodified, per chapter — bodhisvara.com

Synthesised recitations (Bhagavadgītā, Nala, lesson sentences, akṣara sounds) — generated with Matcha-TTS (© 2023 Shivam Mehta, MIT) on public-domain Sanskrit text — © Khan.

Noto Serif / Noto Sans (Latin & KR), Noto Serif / Sans Devanagari, Noto Serif / Sans Thai, and the Noto script fonts used by the reader's extended transliterations (Tibetan, Myanmar, Khmer, Balinese, Javanese, Grantha, Sharada, Siddham, Brahmi, Mongolian — Serif or Sans as published) — Google — SIL Open Font License 1.1
Gentium Plus (self-hosted on the Learn pages) — SIL — SIL Open Font License 1.1

Aksharamukha (transliteration) — Vinodh Rajan — AGPL-3.0 — aksharamukha.appspot.com
Devanagari stroke-order data — derived from Wikimedia Commons animations (Opiaterein CC BY 3.0; JackPotte CC BY-SA 3.0/GFDL) and Noto glyph outlines (OFL 1.1) · ShareAlike portions kept isolated.

以下材料仍属其作者与出版社版権，未经开放许可；仅对注册用户限量检索／有限访问，不提供再分发，各归其权利人。

Modern reference works indexed for search (/refs) and certain in-copyright library scans remain © their authors and publishers — used as limited, access-controlled search/reading for registered users only; not offered for download or redistribution.
In-copyright dictionaries within the collection — Apte Practical Sanskrit-English Dictionary revised ed. 1957 (Prasad Prakashan); Edgerton Buddhist Hybrid Sanskrit 1953 (Yale UP); Stchoupak–Nitti–Renou 1959; Mani Purāṇic Encyclopaedia 1975 (Motilal); Meulenbeld; Sircar 1966; Mehendale 1993; Sharma 1978 — © their authors/publishers. Cologne’s digitisation licence does not cover these underlying works; treated as restricted, not open.
Apte 1957 → Thai translation — a derivative of the in-copyright Apte 1957; requires the rights-holder’s permission and is not covered by this site’s licence.

本站的原创部分（网站、OCR 校理文本、多语翻译）以 CC BY-NC-SA 4.0 协议发布：可自由分享与改作，唯须署名、不得商用、衍生作品须采用相同协议。上游材料（CDSL、Héritage 等）仍受其各自许可约束。 CC BY-NC-SA 4.0 ↗

← 返回