This page records how the Works Cited page is now generated from the translation memory, what the switch changed, and the questions that still need a curator's decision.
The switch has been made. The hand-maintained version is archived in full at Texts Cited in Śikṣāsamuccaya/Archive 2026-08 — nothing was lost, and the identifications recorded there are preserved.
Please read What we need from you at the end.
Where the data comes from
The Śikṣāsamuccaya has been segmented and aligned in the Translation
Memory namespace, one wiki page per segment, under
Translation Memory:016-Tsadra-SSC-Root/.
Each segment records the Tibetan, Wylie, Sanskrit and English of one passage. Crucially, when a segment is a quotation from another text, it is marked as such and the source is named:
{{TranslationMemory
|SegmentOrder=44
|Tibetan=མི་ཁོམ་པ་བརྒྱད་ལས་ལྡོག་པ་ཡང་རྙེད་པར་དཀའ། …
|English=It is difficult to come back from the eight bad rebirths. …
|SegmentFormat=Quote
|QuoteSource=Gaṇḍavyūhasūtra (RKTSK 44-45)
…
}}
Those last two parameters are what make this possible. SegmentFormat=Quote
marks the passage as a quotation, and QuoteSource names the work
being quoted, usually with its RKTS catalogue number — RKTSK for a Kangyur
text, RKTST for a Tengyur text.
Look at every segment of the Śikṣāsamuccaya alignment. Wherever a segment is a quotation and names its source, collect that source. Group the results by work, and count how many times each work is quoted.
In query form that is simply:
{{#ask: [[~Translation Memory:016-Tsadra-SSC-Root/*]][[QuoteSource::+]]
|?QuoteSource
}}
Run against the wiki as it stands today, this finds:
A live sample
The table below is generated live by the query described above, limited to the first six results. It is real data, pulled from the wiki as you load this page — not a mock-up.
| Segment | Quoted source | Ch. | Folio |
|---|---|---|---|
| Translation Memory:016-Tsadra-SSC-Root/44 | Texts/Gaṇḍavyūhasūtra | 1 | 3b |
| Translation Memory:016-Tsadra-SSC-Root/50 | Texts/Ratnolkādhāraṇīsūtra | 1 | 3b;4a;4b |
| Translation Memory:016-Tsadra-SSC-Root/52 | Texts/Ratnolkādhāraṇīsūtra | 1 | 4b |
| Translation Memory:016-Tsadra-SSC-Root/54 | Texts/Daśadharmakasūtra | 1 | 5a |
| Translation Memory:016-Tsadra-SSC-Root/56 | Texts/Lalitavistarasūtra Texts/Siṃhaparipṛcchāsūtra | 1 | 5a |
| Translation Memory:016-Tsadra-SSC-Root/59 | Texts/Siṃhaparipṛcchāsūtra | 1 | 5a |
| ... further results | |||
Note what this gives us that the present page does not: a citation count per work. The most-quoted sources in the Śikṣāsamuccaya turn out to be the Pitāputrasamāgamanasūtra (62 quotations), the Aṣṭasāhasrikāprajñāpāramitā (45), the Ratnameghasūtra (34) and the Dharmasaṃgītisūtra (30). That is a scholarly data point the hand-written page had no way of showing, and it now fills the empty left-hand column of the existing layout.
Counting quotations, not segments
One quotation is often split across several consecutive segments. A long
passage is segmented for alignment, and each piece carries the same
QuoteSource. Counting segments therefore overstates the citations
badly.
The Pitāputrasamāgamanasūtra is the clearest case: 62 quote segments, but 14 quotations, and one of its runs is 36 consecutive segments. Across the whole Śikṣāsamuccaya, 633 quote segments are 322 quotations.
The list groups runs of consecutive SegmentOrder that share a
source, and anchors each quotation to the lowest segment number, so a link
lands at the start of the passage. A change of
TranslationChapter breaks a run — two adjacent quotes from one work
in different chapters are two quotations — though no run in the current data
actually crosses a chapter boundary.
Two details that are easy to get wrong, both found by testing against the real data:
- A segment can name more than one source, semicolon-separated, when a passage is shared:
Bodhisattvapratimokṣa… (RKTSK 248);Vinayaviniścayopāli… (RKTSK 68). Every id has to be read, or one work takes the whole passage and the other shows nothing. - The catalogue number needs a boundary. Matching
RKTSK 12loosely also catchesRKTSK 127andRKTSK 129. Every source value closes its bracket right after the number, so(RKTSK 12)is exact.
Size is measured in words, not segments. A "long quotation" note appears past 400 words of translated text, which is about 8% of them. Segment count says nothing useful about length — the single longest quotation in the corpus, 8,330 words of the Ratnolkādhāraṇī, is one segment.
What the switch changed
The hand-maintained page listed 78 works. The generated list finds 61, and the two do not simply overlap:
- 6 works the hand list did not have now appear, including the Vajradhvajapariṇāmanā with 25 quotations — a substantial omission.
- 24 works the hand list had are not tagged in the translation memory, so they cannot be counted automatically.
Those 24 are not dropped. They sit in a clearly-labelled second section of the page, with the curator's own descriptions and "Also cited as" notes intact. As segments are tagged, entries move up to the generated list on their own.
Eleven of the 24 have no RKTS catalogue number at all — only a title — so they could never be matched automatically even once tagged. Those need a catalogue identification first, and are the most useful thing to fix.
Matching sources to wiki pages
QuoteSource holds a display string, not a link — for example
Gaṇḍavyūhasūtra (RKTSK 44-45). Each of those strings has to be matched
to a text page on this wiki.
Two ways of doing that, used together:
- By RKTS catalogue number. Extract
RKTSK 44-45and match it against theDrlPageNameproperty. This is the more reliable route, because it is unaffected by spelling and it keeps Kangyur and Tengyur texts distinct. - By title. Look up the name as a page, following redirects.
Together these currently resolve 66 of the 77 works, covering roughly 96% of the 633 individual quotations. The eleven that remain are listed below.
We have already created 26 redirects from the alternate titles recorded in
the Also cited as notes on the existing page, which is where a good
deal of this matching came from in the first place. That curatorial work is
preserved.
Titles needing a decision
These are the QuoteSource values that do not currently resolve
to a page. Candidates are mechanical suggestions only — they are based on
spelling similarity, not on scholarship, and none should be acted on without
review. Red links indicate pages that do not exist.
Probable spelling variants
| Quotations | QuoteSource value |
Suggested target | Difference |
|---|---|---|---|
| 2 | Mahākaruṇasūtra | Mahākaruṇāsūtra | ṇa / ṇā
|
| 1 | Jñānavaipūlya sūtra | Jñānavaipulyasūtra | ū / u, and a space
|
| 1 | Saptamaithunasaṃyukta | Saptamaithunasaṃyuktasūtra | target adds sūtra
|
| 1 | Āryāvalokiteśvaravimokṣa | Avalokiteśvaravimokṣa | target drops the Ārya honorific
|
Genuine judgement calls
| Quotations | QuoteSource value |
Possible target | Why we are unsure |
|---|---|---|---|
| 11 | Saddharmasmṛtyupasthānakārikā (RKTSK 947) | Saddharmasmṛtyupasthāna or Saddharmasmṛtyupasthānasūtra | A kārikā may be a separate verse work from the sūtra. Two candidate pages exist. |
| 9 | Bhaiṣajyaguruvaidūryaprabharājasūtra | Bhaiṣajyaguruvaiḍūryaprabhasūtra | Two differences at once (vaidūrya/vaiḍūrya, prabharāja/prabha), and there are several Bhaiṣajyaguru sūtras.
|
| 5 | Mahāratnakūṭasūtra | Ratnakūṭasūtra | Mahāratnakūṭa is the name of the whole collection. The candidate page already redirects to the Kāśyapaparivarta. These may not be the same thing at all. |
| 4 | Trisamayarājasādhana (RKTST 2220 and 2221) | Trisamayarāja | Two different Tengyur sādhanas share this name, while the candidate is a Kangyur tantra. Probably distinct works. |
Cited but not yet on this wiki
No page on this wiki matches these, under any spelling we could find.
That does not necessarily mean we lack the text. Several of these may well exist in the Research Library (DRL) archive and simply never have been created here, having been judged not directly relevant to the Bodhicitta project. If so, bringing them across may be straightforward — but that is a decision for the curator, not an automatic import.
As a first check we searched the DRL database for the titles below and did not find matches for Kāmāvavādakasūtra, Lokanāthavyākaraṇa or Anupūrvasamudgataparivarta. That search was by Sanskrit title only, so it is not conclusive — these works may be catalogued there under their Tibetan or Wylie titles instead.
| Quotations | QuoteSource value |
RKTS number |
|---|---|---|
| 5 | Kāmāvavādakasūtra | RKTSK 914 |
| 1 | Lokanāthavyākaraṇa | — |
| 1 | Anupūrvasamudgataparivarta | — |
What the generated list gives us
- It stays current. The hand-written page had already fallen behind by six works, one of them quoted 25 times. A query cannot.
- Quotation counts. How often each work is quoted, in the column that previously held a decorative vajra.
- The same figure on the text pages. Each cited work now carries a "Cited n times in the Śikṣāsamuccaya" panel of its own, listing every passage with its folio and a link into the bilingual layout. The two are generated from the same rule, so they cannot disagree.
- Far less markup. 790 lines of hand-written entries became about 100, most of which is the 24 untagged works.
- One source of truth. Titles, Tibetan and translations come from the text pages, so a correction made once appears everywhere.
What the switch did not solve
- The
Also cited asnotes are scholarship, not formatting. They record identifications that exist nowhere else on the wiki — that the Candrapradīpasūtra is the Samādhirājasūtra, that the Ratnakūṭasūtra cited here is the Kāśyapaparivarta. They survive in the archive and on the 24 untagged entries, but they are not yet data: they should become redirects or a property before the archive is ever tidied away. - Genre labels. The hand-written page distinguished Sūtra, Dhāraṇī, Tantra, Avadāna, Aspiration prayer and others. That is curatorial judgement, stored nowhere as data, so the generated list falls back to
Classification. Recovering it needs a property of its own. - The unresolved titles below still need a decision either way.
Quality audit of the alignment
We ran an automated pass over all 3349 segments of the alignment, comparing the Tibetan recorded in each segment against the full Tibetan text, and looking for quotations that may have been missed or marked inconsistently.
This is a machine pass, not a scholarly reading. It is good at finding things that look irregular and cannot judge whether an irregularity is an error. Every item below needs a human decision.
Coverage of the Tibetan
Of the 1,414 segments that carry Tibetan, 1,252 appear verbatim in the full text once spacing, tsheg, shad and editorial note markers are normalised. Of the remaining 122, closer checking showed:
- 85 are present but with small textual variants — the text is there, differing in a word or a particle. These are the ordinary differences between witnesses and are probably not errors.
- 31 matched completely once the comparison ignored where the segment starts — i.e. the passage is in the text but the segment boundary falls differently.
- 6 could not be found anywhere in the full text at all.
The six are short (27–44 characters) and are mostly structural, so they are most likely editorial headers rather than missing scripture:
| Segment | Marked as | Length |
|---|---|---|
| 133 | Quote | 40 |
| 322 | Commentary | 43 |
| 327 | Title | 34 |
| 553 | Root | 42 |
| 686 | Commentary | 44 |
| 1265 | Title | 27 |
Quotations with no source recorded
13 segments are marked SegmentFormat=Quote but carry no
QuoteSource. These are quotations the alignment has identified as
such, but whose source was never filled in — so they are invisible to the works-cited
query. This is the most clear-cut finding of the audit.
| Segment | Opening of the Tibetan |
|---|---|
| 133 | བྱང་ཆུབ་སེམས་དཔའི་སྡོམ་པ་ནི། ། རྒྱས་པར་ཐེག་པ་ཆེ་ལས་འབྱུང་། ། |
| 816 | འདོད་པ་དག་ལ་བརྟེན་པའི་མི། ། ལྡོངས་འགྱུར་ཕ་མ་གསོད་པར་འགྱུར། ། |
| 960 | གང་དག་དགའ་དང་མི་དགའ་ཅུང་ཟད་མེད། ། |
| 962 | ཉེས་པའི་ཚོགས་མེད་ཐམས་ཅད་སྤངས་ནས་སུ། ། |
| 964 | གྲོང་དང་གྲོང་ཁྱེར་དག་ལ་དགའ་སྤངས་ལ། ། |
| 966 | སྐྱོན་གྱིས་གཟིངས་པ་མཐའ་ཡས་ལྡན་པའི་ཁྱིམ་སྤངས་ |
| 969 | འོད་སྲུང་འདི་ལྟ་སྟེ་དཔེར་ན་མི་ལ་ལ་ཞིག་ |
| 971 | གནས་གང་དག་ན་བསོད་སྙོམས་བླང་བའི་སྤྱོད་ཡུལ་ |
| 972 | གལ་ཏེ་དགེ་སློང་དགོན་པ་པའི་དྲུང་དུ་རྒྱལ་པོའམ་ |
| 1180 | དེ་བཞིན་དུ་སྦྱར་ཏེ། དེ་དག་ནི་ཐ་སྙད་ཙམ་ |
| 1218 | མཚན་མར་འཛིན་པ་ལས་བྱུང་བའི། ། |
| 1458 | བློ་གྲོས་རྒྱ་མཚོ་སྔགས་ཀྱི་ཚིག་འདི་དག་ནི་ |
| 1476 | བདག་གཞན་མཉམ་ཉིད་གོམས་བྱས་པས། ། |
Several of these (960–972) are consecutive, which suggests a single passage where the source was simply not carried down the run.
Possible quotations marked as commentary
We looked for segments not marked as quotations that nevertheless both
open by naming a sūtra (…མདོ་ལས།, "from the … sūtra") and
close with a citation formula (…ཞེས་གསུངས་སོ།). 23 segments
match both conditions and are currently marked Commentary.
This matters because the texts they name are already tracked elsewhere in the alignment — the Ugraparipṛcchā has 15 recorded quotations, the Dharmasaṃgīti 30, the Ākāśagarbha 22. So these look like inconsistent marking rather than unknown sources. 13 of the 23 sit directly beside a segment that is marked as a quotation.
| Segment | Source named in the Tibetan | Already tracked? |
|---|---|---|
| 82 | Śūraṅgamasamādhisūtra | yes |
| 208 | Gaganagañjaparipṛcchā / Ākāśagarbha | yes |
| 219 | Vīradattagṛhapatiparipṛcchāsūtra | yes |
| 283 | Gaganagañjaparipṛcchā / Ākāśagarbha | yes |
| 302 | Ratnakūṭa | yes |
| 362 | "from that same text" — refers back | — |
| 560 | Ugraparipṛcchāsūtra | yes |
| 584 | Dharmasaṃgītisūtra | yes |
| 594 | Brahmaparipṛcchāsūtra | yes |
| 602 | Gaganagañjaparipṛcchā / Ākāśagarbha | yes |
| 627 | Ugraparipṛcchāsūtra | yes |
| 634 | Ratnameghasūtra | yes |
| 693 | Ugraparipṛcchāsūtra | yes |
| 708 | Ratnakūṭa | yes |
| 718 | "from that same text" — refers back | — |
| 745 | Mañjuśrīvikrīḍitasūtra | yes |
| 746 | Gaṇḍavyūhasūtra | yes |
| 749 | "from that same text" — refers back | — |
| 771 | Śraddhābalādhānāvatāramudrāsūtra | yes |
| 904 | Candrapradīpa / Samādhirāja | yes |
| 1333 | Ugraparipṛcchāsūtra | yes |
| 1398 | Ugraparipṛcchāsūtra | yes |
| 1453 | "from that same text" — refers back | — |
There is a legitimate reason some of these might stay as they are. Śāntideva frequently paraphrases or introduces a source before quoting it, and the alignment may deliberately mark the introduction as commentary and the quotation proper as the following segment. That is exactly the kind of distinction a machine cannot make.
What we did not find
- No segment carries a
QuoteSourcewithout also being marked as a quotation — so there is no case of a source attached to something that is not a quote. - No duplicated or out-of-order segment numbers.
- No quotation appears to have been invented — every marked quotation with Tibetan was found in the full text, apart from the six short segments noted above.
On the evidence of this pass the alignment looks sound. The findings are about completeness of the metadata — sources not filled in, and a marking distinction applied unevenly — rather than about the text itself.
What we need from you
- The eleven untagged works with no RKTS number — these can never be matched automatically as things stand. A catalogue identification for each is the single most useful thing to add.
- The other thirteen untagged works — they have catalogue numbers but no segment names them. Should the alignment be extended to cover them, or were they cited too loosely to tag?
- The four probable spelling variants — may we create redirects for these?
- The four judgement calls — are these the same work or not? These need a specialist eye.
- The three works not yet on this wiki — do these exist in the DRL archive? If so, should they be brought across, or were they deliberately left out as not relevant to BCA?
- Genre labels — worth a property on the text pages, so the Sūtra / Dhāraṇī / Avadāna distinctions come back?
And on the audit:
- The 13 quotations with no source — can these be filled in? They are currently invisible to any works-cited listing.
- The 23 commentary segments that name a sūtra — should these be marked as quotations, or is the present marking deliberate?
- The six segments not found in the full text — are these editorial headers, as they appear to be?
