Field dictionary

These are the fields of a work's record and a source's record. A work's record.json on this site has the work fields, and so does each line of works.jsonl.gz in a release; each line of a release's sources.jsonl has the source fields. I generate both tables from each record's JSON Schema, which the build derives from the library's data models.

Works

Fields of a work record (scroll sideways for every column)
Field	Type	What it is
`id`	string, required	The work's identifier, such as `m7x2pd`. If I retire it, the record that replaces it lists it under `aliases`; if nothing replaces it, the record moves to the retired records and gets a `retired` block.
`source`	string	The identifier of the source that holds the work's text. Absent for a cataloged work, since I don't hold a copy.
`state`	string	Whether I hold a copy of the work. A held work has a source and a locator; a cataloged work is one I list without holding a copy. Values: `held`, `cataloged`.
`aliases`	array of objects	Identifiers this work used to have. Each one has the date I retired it, the reason, and whether its old address redirects here.
`locator`	object	Where the work sits in its source. It has the first and last page, the character offsets of the work in the source's served text, the SHA-256 of that stretch of text once I've recorded it, and whether a person has confirmed the boundary and, if so, who confirmed it and when. Absent for a cataloged work.
`author`	string, required	The author's surname, lowercased. For a court writing it's the judge's key in the project's list of judges.
`relationship`	string, required	How the work relates to Gorsuch. A work he authored is one he wrote. Values: `authored`, `joined`, `about`.
`genre`	string, required	What kind of writing the work is. Values: `judicial-opinion`, `book`, `law-review-article`, `working-paper`, `thesis`, `speech`, `testimony`, `interview`, `other`.
`posture`	string	For a judicial opinion, the author's role in the case. Values: `majority`, `concurrence`, `concurrence-in-judgment`, `dissent`, `partial`, `plurality`, `partial-majority`, `per-curiam`, `statement`, `concurrence-in-part-and-judgment`.
`publication`	string	For a court writing, how the court issued it. A writing the court issued with an order, such as a statement respecting the denial of certiorari, has the value `order`. Values: `published`, `unpublished`, `order`.
`crk_ordinal`	integer	A number I set by hand, 2 or higher, only when two works would otherwise share a court record key.
`crk_replaced`	array of objects	Court record keys an earlier release published for this work that a correction later replaced, each with the reason.
`venue`	string, required	The court the work comes from. Writing outside a court has `n/a`. Values: `scotus`, `ca10`, `other-court`, `n/a`.
`era`	string, required	The period of his career the work belongs to. Values: `pre-judicial`, `ca10`, `scotus`.
`decided`	string (date)	For an opinion, the decision date. For other writing, the publication date.
`word_count`	integer	The number of words in the work's served text.
`case`	object	For a judicial opinion, the case name and its reporter citations. A citation for a slip opinion whose page isn't assigned yet has no page, and a citation records where I got it when I know.
`scdb_case_id`	string	The case's identifier in the Supreme Court Database, where I've matched one.
`excerpt`	object	For an excerpt-only work, the pages and character span of the passage the site shows, inside the work's locator. The dataset never has the passage's text.
`bibliographic`	object	For writing outside a court, a bibliographic record modeled on CSL-JSON, with authors, title, container, publisher, year, pages and identifiers such as a DOI or ISBN. Every cataloged work has one.
`rights`	object	The work's own rights, when they differ from its source's. Absent when the work takes its source's rights.
`retired`	object	Present only on a record I've retired with no successor. It has the date, the reason, and whether the old address redirects to the changes page.
`record_url`	string (uri), required	The work's permanent address on this site, by identifier and without the title in it. Cite this one.
`release`	string or null, required	The version of the dataset release the record belongs to. In a `record.json` on the site, it's the newest release when I built the site, or null before the first one.
`rights_tier`	string, required	The rights tier that governs the work. It's the work's own tier if it has one, metadata-only for a cataloged work, and otherwise its source's.
`crk`	string	The court record key, a readable key for a judicial opinion as the court first issued it. Absent for a work that isn't a judicial opinion, and for an opinion I can't key yet.
`doctrine`	array of strings	Terms from the doctrine vocabulary that the work's tag file lists, without their confidence scores. Absent when the work has no tag file.
`method`	array of strings	Terms from the interpretive method vocabulary that the work's tag file lists, without their confidence scores. Absent when the work has no tag file.
`text`	string	The work's full served text. Present only for a public-domain or open-licensed work whose text the site shows in full.

Sources

Fields of a source record (scroll sideways for every column)
Field	Type	What it is
`id`	string, required	The source's identifier, such as `k3v9q2`. If I retire it, the record that replaces it lists it under `aliases`; if nothing replaces it, the record moves to the retired records and gets a `retired` block.
`title`	string, required	The document's title.
`role`	string	Whether the document is his. A primary source is a document with his byline, and most primary sources have works; `no_work_reason` says why one doesn't. A reference source is writing about him that I keep for context and provenance, and it never has works. Values: `primary`, `reference`.
`document_type`	string	Which copy of the writing this is. Values: `slip-opinion`, `page-proof`, `bound-volume`, `court-opinion`, `order-list`, `book`, `chapter`, `article`, `report`, `web-page`.
`venue`	string, required	The court the document comes from. A document from outside a court has `n/a`. Values: `scotus`, `ca10`, `other-court`, `n/a`.
`docket`	string	The lead docket number.
`other_dockets`	array of strings	The other docket numbers of a consolidated case.
`decided`	string (date)	The decision date, for a court document.
`status`	string	Where the source stands in acquisition and extraction. The `status_detail` field explains any status other than `acquired`. Values: `acquired`, `pending-acquisition`, `acquisition-failed`, `extraction-failed`, `superseded`.
`status_detail`	string	Why, when the status isn't `acquired`.
`superseded_by`	string	The identifier of the source that replaced this one, set only when the status is `superseded`. The validate command checks that following it ends at a source that isn't superseded.
`official_url`	string	Where to find the court's own copy, which isn't always where I got the file.
`no_work_reason`	string	Why a primary source has no work, as a short code or a sentence.
`aliases`	array of objects	Identifiers this source used to have. Each one has the date I retired it, the reason, and whether its old address redirects here.
`acquisition`	object	How I got the file. It has the method, when I retrieved the file, the address I got it from when there is one, and the file's size, media type and SHA-256. Absent until I've acquired it.
`rights`	object, required	The rights tier and the reason it applies, with the rights holder where there is one and any cap on excerpt length for this source.
`extraction`	object	How I extracted the text. It has the extractor, its version and backend, the options and models it ran with, when it ran, the page count and its confidence scores. Where I've recorded them, it also has SHA-256 pins of the extraction, of the word positions and of the served text, and any pages a reviewer accepted as blank or unaligned. Absent until I've extracted the source.
`retired`	object	Present only on a record I've retired with no successor. It has the date, the reason, and whether the old address redirects to the changes page.
`original_url`	string (uri)	The address of my copy of the original file on the data host. Present only for a public-domain source I've acquired.
`page_url`	string (uri)	The source's page on this site.