v0.11.2

Patch Changes

• #199 6797408 Thanks @medelman17! - fix: suppress phantom citations emitted from numeric-prefixed party names (#196)

  Real-world NY caption like Board of Mgrs. of the 15 Union Sq. W. Condominium v. BCRE 15 Union St., LLC, 2025 NY Slip Op 00784 emitted two case
  citations: the real slip op plus a phantom 15 Union Sq. W. Condominium v. BCRE 15 extracted from inside the plaintiff's name. The phantom read 15
  as volume, Union Sq. W. Condominium v. BCRE as reporter, and 15 as page.

  Root cause. The state-reporter regex's non-greedy reporter capture
  ([A-Za-z.\d\s]+?) happily spanned the " v. " case-name separator and
  backtracked until a second number appeared. The downstream false-positive
  filter caught this only when reporters-db was loaded — which is opt-in for
  bundle-size reasons, so most consumers saw the phantom pass through.

  Fix. Added negative lookahead (?!\s+vs?\.\s) to both state-reporter
  and law-review patterns so the reporter/journal capture cannot span a
  " v. " or " vs. " token. No real US reporter or journal name contains
  that sequence. Applied to both patterns because a first-pass guard on just
  state-reporter surfaced the same phantom under law-review.

  Five new regression tests: the exact #196 text, a vs. variant, a cross-type
  guard (no phantom journal), and two adversarial controls.

v0.11.1

Patch Changes

• #197 925a719 Thanks @medelman17! - fix: align CASE_NAME_ABBREVS with reporters-db Bluebook T6 list + ampersand support

  After three consecutive bug reports (#187, #188, #193) exposing missing
  abbreviations, this change aligns CASE_NAME_ABBREVS with the canonical
  Bluebook T6 case-name abbreviation list maintained by Free Law Project
  (reporters-db/case_name_abbreviations.json).

  Three improvements:

  • Strip internal apostrophes in stem lookup. isLikelyAbbreviationPeriod
    previously kept inner apostrophes, so Nat'l. computed stem nat'l which
    no reasonable pure-letter set could match. Now normalized to natl, which
    matches the Bluebook's apostrophe-form abbreviations as pure-letter stems.
  • 41 new entries in CASE_NAME_ABBREVS. Period-forms (co, cmty,
    envtl, gend, par, prot, ref, sol, cty, adver) and
    apostrophe-forms (assn, dept, natl, intl, govt, commn, commr,
    contl, fedn, meml, pship, profl, secy, sholder, socy,
    commcn, engg, engr, entmt, envt, examr, invr, admr, admx,
    empr, empt, exr, exx, publg, publn, regl). co was the
    highest-impact gap — "Smith & Co. United States Corp." was silently
    truncated to "United States Corp." because "Co. U" fired the
    sentence-boundary scan.
  • & in isLikelyPartyName. Ampersand is ubiquitous in corporate
    captions ("Smith & Jones", "Goldman, Sachs & Co.") and previously caused
    the Priority-3 single-party fallback (#193) to reject such captions. Now
    treated as a valid standalone token.

  7 new regression tests covering period-forms, apostrophe-forms with trailing
  period, adversarial Dep't … caption, and ampersand patterns.

v0.11.0

Minor Changes

• #192 7f84d0c Thanks @medelman17! - feat: star-pagination (at *N) support on all pincite-bearing citation types (#191)

  Star-pagination pincites (at *1, at *2-4) were silently dropped on id,
  supra, shortFormCase, full case cites with slip-opinion reporters (NY Slip
  Op), and neutrals (Westlaw, Lexis). In real-world NY state-court briefs this
  meant a significant fraction of pincites came back undefined. Plain-integer
  pincites (at 465) continued to work.

  Changes:

  • parsePincite / PinciteInfo — accept optional * prefix; new
    starPage?: boolean flag distinguishes slip-opinion pages from reporter
    pages. Existing page: number still carries the numeric portion, so
    backward compatibility for consumers reading pincite as a number is
    preserved.
  • Full case cites — PINCITE_REGEX, LOOKAHEAD_PINCITE_REGEX, and
    PINCITE_SKIP_REGEX now accept an optional at keyword and * prefix.
    Pincite extraction also runs when no trailing parenthetical is present,
    so forms like 2020 NY Slip Op 00001 at *2 capture the pin even though
    there is no (Court YYYY) block.
  • Short-form citations — ID_PATTERN, IBID_PATTERN, SUPRA_PATTERN,
    STANDALONE_SUPRA_PATTERN, and SHORT_FORM_CASE_PATTERN now accept *?
    before the pincite digits. The matching extractors populate
    pinciteInfo.starPage and now expose pinciteInfo on IdCitation,
    SupraCitation, and ShortFormCaseCitation.
  • Neutral citations — NeutralCitation gains pincite?: number and
    pinciteInfo?: PinciteInfo fields. extractNeutral now accepts the
    cleaned source text and extracts a trailing , at *N / at *N pincite.
    Previously, numeric pincites on neutrals were also silently dropped;
    this change fixes that as a side effect.

  Known limitation: the second occurrence of a NY Slip Op short-form
  (2020 NY Slip Op 00001 at *2) is still classified as case rather than
  shortFormCase, because SHORT_FORM_CASE_PATTERN forbids a page between
  the reporter and at. The pincite data itself is captured correctly.
  Shortform classification for NY Slip Op will be addressed in a follow-up.

Patch Changes

• #195 f10c234 Thanks @medelman17! - fix: recognize single-party corporate captions and In the Matter of … prefix (#193)

  FullCaseCitation.caseName came back null for any caption that didn't
  contain v. or match the short procedural-prefix list (In re,
  Matter of, Estate of, Ex parte, etc.). Common NY patterns like
  Board of Mgrs. of the St. Tropez Condominium and Board of Directors of Hill Park silently lost their case names — downstream UI fell back
  to displaying the bare reporter triple.

  Two root causes:

  • Missing long-form procedural prefix. In the Matter of X was
    reduced to Matter of X because the short prefix matched mid-string
    before the long one could. Added In the Matter of to
    PROCEDURAL_PREFIX_REGEX and the extractPartyNames prefix list, both
    with priority over Matter of.

  • No generic fallback for single-party captions. When both V. and
    procedural-prefix scans fail, the backward scanner now uses the
    post-truncation precedingText itself as the caption, after stripping
    any leading signal word (See, cf., etc.) and validating via
    isLikelyPartyName + SENTENCE_INITIAL_WORDS. Because the truncation
    step already bounds precedingText by sentence/citation/paren-signal
    boundaries, sentence prose like "The court held that..." is not
    mis-matched.

    11 new regression tests cover corporate captions (Board of Mgrs. of,
    Board of Managers of, Board of Directors of, bare Corp.),
    In the Matter of priority over Matter of, sentence-prose safety, and
    pre-existing adversarial/Estate of/ex rel. controls.

v0.10.3

Patch Changes

• #189 7425448 Thanks @medelman17! - fix: close remaining case-name boundary gaps for NY-style citations (#187, #188)

  Two root causes behind remaining caseName failures on real-world NY briefs:

  • Missing geographic/street abbreviations. Is. (Island), Mt. (Mount),
    Ft. (Fort), Pt. (Point), Rt. (Route), St. (Saint/Street), Blvd.,
    Sq., Hwy., Pkwy., and Hts. were not in CASE_NAME_ABBREVS, so the
    backward scanner treated their periods as sentence boundaries and truncated
    names like Clark-Fitzpatrick, Inc. v. Long Is. R.R. Co. and
    Matter of Long Is. Power Auth. Hurricane Sandy Litig.. Added to the
    Bluebook T6/T10 set.
  • Missing paren signal words. quoted in, accord, and the
    citing, e.g., form were not recognized as hard boundaries, so backward
    scans of citations introduced by those signals overshot into the prior
    citation's trailing parenthetical. Extended PAREN_SIGNAL_BOUNDARY_REGEX.

v0.10.2

Patch Changes

• #185 e1a46d0 Thanks @medelman17! - fix: robust case-name boundary detection with Bluebook T6/T10 abbreviations (#182, #183, #184)

  Replace the narrow LEGAL_ABBREVS regex (~30 entries) with a comprehensive Bluebook-sourced abbreviation set (200+ entries from T6/T7/T10) backed by heuristics for single-letter initials and dotted initialisms. Add hard boundary detection for Id. markers and parenthetical signal words (quoting, citing, cited in). Fixes case names that were undefined, truncated, or overshot when party names contained abbreviation chains like "Cent. Sch. Dist.", "Mgt., Inc.", or "A.N.L.Y.H. Invs."

v0.10.1

Patch Changes

• #178 d72ba1e Thanks @medelman17! - Fix case name extraction still capturing sentence context in two scenarios: sentence-initial pronouns like "This" bypassing the trimming guard, and "In" prefix not being stripped from caseName after extractPartyNames removes it from plaintiff.

v0.10.0

Minor Changes

• #177 fc83dff Thanks @medelman17! - Add granular component spans for all citation types. Each citation now carries a spans record with per-component position data (volume, reporter, page, court, year, caseName, plaintiff, defendant, signal, etc.). Explanatory parentheticals gain a span field. New spanFromGroupIndex utility exported for power users. Closes #172, closes #171.

Patch Changes

• #175 76bd36d Thanks @medelman17! - Fix case name extraction capturing preceding sentence context as plaintiff. Add isLikelyPartyName validation with trimming when lowercase non-connector words are detected.

• #173 d96719a Thanks @medelman17! - Fix Id. resolution to correctly resolve through short-form, supra, and non-case citations. Remove dead allowNestedResolution option.

v0.9.0

Minor Changes

• #166 42fc27d Thanks @medelman17! - Add statute citation patterns for 31 additional US state jurisdictions

  Expands the abbreviated-code pattern family from 12 to 43 states using a data-driven
  regex generation approach. Each new state was verified against real citation formats from
  court opinions and legislative text.

  New jurisdictions: AK, AZ, AR, CT, DC, HI, IA, ID, KS, KY, LA, ME, MN, MO, MS, MT,
  ND, NE, NH, NM, NV, OK, OR, RI, SC, SD, TN, VT, WI, WV, WY.

v0.8.4

Patch Changes

• #164 449e1ea Thanks @medelman17! - Fix constitutional citation extraction gaps: support comma after "Const." separator, recognize "Amdt." abbreviation, and add low-confidence bare article pattern for standalone "Art. I, §8" references

v0.8.3

Patch Changes

• #162 919c8c9 Thanks @medelman17! - fix: prevent greedy false matches in position mapping lookahead (#161)

  The rebuildPositionMaps lookahead found the first matching character,
  not the correct one. When normalizeDashes expanded em-dashes (— → ---)
  near text containing hyphens (e.g., page ranges like "110-115"), the
  deletion lookahead grabbed the wrong "-", collapsing subsequent position
  mappings and producing zero-length original spans on extracted citations.

  The fix adds a confirmation check: a lookahead match is only accepted when
  at least 3 characters after the match point also align. Both deletion and
  insertion directions are searched simultaneously and the shorter confirmed
  match wins, preventing greedy false matches.