Use cases

These scenarios are illustrative, not exhaustive. Each pairs “what happens today” with “what TextRefs gives you” so the value is concrete.

Researcher citing a passage across editions

Today. You write “Plato, Rep. 514a” in your manuscript. Your reader either knows the Stephanus convention and resolves the citation themselves, or they don’t. If they want to follow up, they pick whichever edition is at hand; there is no shared link.

With TextRefs. You drop in https://textrefs.org/id/ref/... (or paste the human-readable alias Plato/Republic/514a). The reader resolves it to a landing page listing every curated mapping — Perseus, Scaife, Loeb (where licensed), a Wikisource transcript — plus a JSON-LD record they can pipe into Zotero or Hypothes.is. Switching editions is a click, not a search.

Digital edition project linking to and from canonical references

Today. You publish a critical edition with project-local IDs. Other projects link to your URLs; six months later you reorganise the site and every inbound link breaks. You email three colleagues and apologise.

With TextRefs. You mint a TextRefs ID for each canonical reference your edition surfaces and alias your project-local IDs to it. Inbound links target the TextRefs URI; you control your internal URL scheme without breaking the citation graph.

Library or institutional repository indexing scholarly works

Today. Your full-text indexer extracts citations from a PDF — “Aristotle, Eth. Nic. 1094a1” — as a string. It can match other strings literally but it can’t cluster, can’t infer the work, can’t expand abbreviations to Nicomachean Ethics.

With TextRefs. The indexer parses the citation against the bekker CitationSystem and stores the resulting TextRefs ID. Now you have FRBR-style work clustering for free, cross-corpus passage search, and authority alignment with Wikidata via mappings.

Theologian or legal scholar working with traditional reference systems

Today. Stephanus pagination, Bekker numbering, Homeric book-and-line references, and biblical book-chapter-verse — every tradition has its own implicit, untyped notation. There is no machine-readable contract for what is a valid citation in each system.

With TextRefs. Each tradition is a CitationSystem with a documented locator regex, a normalisation rule, and a list of valid reference types. A parser can validate “Vulg. Gen. 1:1” or reject “ST I-II.300.99” because no registered canonical reference exists for that locator. The reference identifier is independent of any single edition.

AI/LLM grounding and retrieval

Today. Language models cite “Plato, Republic 514a” verbatim from training data, including hallucinated passages. Retrieval-augmented systems have nothing to retrieve against at the passage level — only documents.

With TextRefs. Each canonical reference is a structured anchor. Training pipelines can tag occurrences in source material; retrievers can ground generations against https://textrefs.org/id/ref/... URIs; verification tools can compare the model’s claim to a known reference and flag drift.

Citation-managing tools and scholarly markdown

Today. Citation managers store free-text “Rep. 514a” — searchable, but neither typed nor linkable. Hypothes.is annotations on canonical passages are tied to a specific edition’s URL.

With TextRefs. Citation managers store a TextRefs URI as the primary key; the human-readable label is just a display string. Annotation tools can normalise edition-bound URLs to TextRefs IDs so a marginal note on Plato 514a follows the passage rather than the edition.