Config.Metadata.Author

Path: build/Config/Metadata/Author.hs | Language: Haskell | Lines: ~3,438

Configuration data for author name canonicalization and linking

Overview

Config.Metadata.Author is a large configuration module containing lookup tables for author name normalization and hyperlinking. It holds no logic—only data structures that define how variant author names (Twitter handles, abbreviated names, misspellings, non-ASCII variants) map to canonical forms, and how canonical names map to biographical URLs.

The module enables a "write once, link everywhere" model for author metadata: define "Eliezer Yudkowsky" → https://yudkowsky.net once, and every annotation mentioning that author automatically gets a hyperlink. Author backlinks then emerge naturally—the author's bio page shows every annotation where they appear.

The design separates configuration data from logic. All the actual processing (canonicalization, link insertion, collapse formatting) lives in Metadata.Author. This module provides the raw mappings consumed by that logic.

Public API

`canonicals :: M.Map String String`

Maps alternative author names to their canonical form. Contains ~770 entries covering:

Twitter handles → real names: "karpathy" → "Andrej Karpathy"
Name variants → standard form: "E. Yudkowsky" → "Eliezer Yudkowsky"
Pseudonyms → real names: "Student" → "William Sealy Gosset"
Typos and encoding issues: "Jurgen Schmidhuber" → "Jürgen Schmidhuber"

Called by: Metadata.Author.authorDB Used for: Pre-link canonicalization of author fields

`canonicalsWithInitials :: [String]`

List of ~200 full names that should have abbreviation variants auto-generated. Rather than manually listing "K. Ong", "K. J. Ong", etc., the name "Ken K. Ong" here causes Metadata.Author.name2Abbreviations to generate all plausible abbreviations automatically.

Called by: Metadata.Author.authorDB Used for: Generating exhaustive initial-based variants

`authorLinkDB :: M.Map T.Text T.Text`

Maps canonical author names to biographical URLs. Contains ~1,400+ entries combining:

Auto-generated Wikipedia URLs from authorWpLinkDB
Override URLs for authors with better homepages than Wikipedia

Examples:

("Gwern", "/index#abstract")
("Scott Alexander", "https://www.astralcodexten.com/")
("Michael Nielsen", "https://michaelnielsen.org/")

Disambiguation via anchor syntax:

("George Washington#SS", "https://en.wikipedia.org/wiki/SS_George_Washington")

Called by: Metadata.Author.linkify, link-prioritize.hs Used for: Converting author names to hyperlinks

`authorWpLinkDB :: [T.Text]`

List of ~1,200 author names that have Wikipedia articles. Converted to Wikipedia URLs via Interwiki.toWikipediaEnURL. This is the bulk source for authorLinkDB.

Called by: authorLinkDB (internal) Used for: Auto-generating Wikipedia author links

`cleanAuthorsRegexps :: [(String, String)]`

Regex-based rewrite rules for normalizing author name formatting. Handles:

Missing spaces after commas: "Foo Bar,Quuz" → "Foo Bar, Quuz"
Inverted name order: "Smith, J." → "J. Smith"
Collapsed initials: "A.B. Smith" → "A. B. Smith"

Called by: Metadata.Author.cleanAuthors Used for: Pre-canonicalization cleanup

`cleanAuthorsFixedRewrites :: [(String, String)]`

Fixed string replacements for author cleanup. Removes:

Academic titles: "Dr ", " PhD", " MD", "Prof "
Affiliation markers: " MA,", " MSc,"
Formatting noise: " et al", "Jr." → "Junior"

Called by: Metadata.Author.cleanAuthors Used for: Stripping academic credentials from names

`authorLinkBlacklist :: [T.Text]`

Names that should never be auto-linked due to ambiguity or being too generic. Includes single letters ("a" through "z"), years ("1922"), and common ambiguous surnames ("Anonymous", "Qi", "Rau").

Called by: Metadata.Author.authorPrioritize Used for: Filtering link candidates

`authorWhitelist :: [String]`

Author names that look like errors but are legitimate. Used by LinkMetadata.readLinkMetadataAndCheck to suppress false-positive warnings for names like "K. U." or "0xType".

Called by: LinkMetadata.readLinkMetadataAndCheck Used for: Suppressing validation warnings

`authorCollapseTestCases :: [(String, [Inline])]`

Test cases for the author list collapse/display logic. Verifies that long author lists get properly wrapped in collapsible <span class="collapse"> elements.

Called by: Test.hs Used for: Unit testing

`extractTwitterUsernameTestSuite :: [(String, String)]`

Test cases for Twitter/X URL → username extraction.

Called by: Test.hs Used for: Unit testing

Internal Architecture

Data Flow

Raw author string from annotation
    ↓
cleanAuthors (regex + fixed rewrites)
    ↓
authorsCanonicalize (lookup in authorDB)
    ↓
linkify (lookup in authorLinkDB)
    ↓
Pandoc Inline with optional Link

The Two-Database Pattern

Canonicalization uses a merged database (authorDB in Metadata.Author):

canonicals — hand-written variant → canonical mappings
Generated abbreviations from canonicalsWithInitials via name2Abbreviations

These must be disjoint; overlaps trigger a runtime error to catch semantic conflicts.

Disambiguation Syntax

Author names can include #anchor suffixes for disambiguation:

"George Washington#SS" links to the ship, not the president
"John Smith#genetics" vs "John Smith#VC" for same-name authors

The # suffix is stripped before display but used for link lookup.

Key Patterns

Wikipedia as Default

The bulk of author links come from Wikipedia. The authorWpLinkDB list contains names known to have Wikipedia articles; these are converted to URLs via toWikipediaEnURL. Non-Wikipedia URLs in authorLinkDB override these defaults.

Credential Stripping

Academic author names often include degrees and affiliations that clutter display:

"John Smith PhD, MD, FRCP" → "John Smith"

The cleanAuthorsFixedRewrites list removes 60+ such suffixes.

Initial Expansion

Rather than manually listing every abbreviation of "Ingrid Sigfrid Melle", the canonicalsWithInitials mechanism generates:

"I. Melle"
"I. S. Melle"
"I.S. Melle"
"Ingrid Melle"
"Ingrid S. Melle"
etc.

This reduces maintenance burden for frequently-cited authors.

Frequency-Based Prioritization

The authorPrioritize function (in Metadata.Author) identifies high-frequency unlinkified authors, weighted by backlink count. This guides manual effort toward the most impactful additions.

Configuration

All configuration is inline in this file as Haskell data literals. To add:

New canonical mapping:

canonicals = M.fromList [
  ...
  , ("twitterhandle", "Real Name")
  ...
]

New author link:

authorLinkDB = M.fromList $
  zip authorWpLinkDB (map toWikipediaEnURL authorWpLinkDB) ++
  [ ...
  , ("Author Name", "https://homepage.example/")
  ...
  ]

New auto-abbreviated name:

canonicalsWithInitials = [
  ...
  , "First Middle Last"
  ...
]

Integration Points

Consumed By

Module	What it uses
`Metadata.Author`	All exports (canonicalization, linking logic)
`LinkMetadata`	`authorLinkDB`, `authorWhitelist` (validation)
`link-prioritize.hs`	`authorLinkDB` (finding unlinked authors)
`Test.hs`	Test suites (validation)

Test Validation

Test.hs validates:

All map keys are unique
No canonicalization loops (A → B → A)
All regexes are syntactically valid
All authorLinkDB values are valid URLs

Overview​

Public API​

canonicals :: M.Map String String​

canonicalsWithInitials :: [String]​

authorLinkDB :: M.Map T.Text T.Text​

authorWpLinkDB :: [T.Text]​

cleanAuthorsRegexps :: [(String, String)]​

cleanAuthorsFixedRewrites :: [(String, String)]​

authorLinkBlacklist :: [T.Text]​

authorWhitelist :: [String]​

authorCollapseTestCases :: [(String, [Inline])]​

extractTwitterUsernameTestSuite :: [(String, String)]​

Internal Architecture​

Data Flow​

The Two-Database Pattern​

Disambiguation Syntax​

Key Patterns​

Wikipedia as Default​

Credential Stripping​

Initial Expansion​

Frequency-Based Prioritization​

Configuration​

Integration Points​

Consumed By​

Test Validation​

See Also​