Repository navigation
Should a dotted title like "U.S. District Judge" read as a title, the way "US District Judge" does? #490
Description
Activity
Looked at this on 2026-10-02 against the 2.4 tree. Taking it out of 2.4: the benefit is small and nobody has asked for it. This records what we worked out so it doesn't need to be redone.
Why the dotted form misses.
US District Judge Marc Thomas Treadwellalready reads titleUS District Judge, becauseusis in TITLES.U.S. District Judge …reads givenU.S., middleDistrict Judge Marc Thomas, because title lookup keeps interior periods (u.s). Suffix-acronym lookup strips all of them, which is whyM.D.matches as a suffix andU.S.never matches as a title.Why a blanket period-strip is wrong. The interim 2.0 build that did this was reverted. 73 two- and three-letter titles would collide with ordinary initial pairs (
D.R.,M.R.,M.S.,S.T.,H.R., …), andU.S. Grantwould read titleU.S., familyGrant, the same wayUS Grantreads today. The title floor keeps a name word, but it gives back the last word, so the given name is lost.The candidate rule, if someone wants this. A token whose period-separated chunks are all single letters, and whose period-free spelling is in TITLES, joins the title run only if a title still follows it after the floor has given its word back:
Input Reading U.S. Grantunchanged: given U.S., familyGrantD.R. Kingunchanged: floor returns King, givenD.R.D.R. King Georgetitle D.R. King, givenGeorge(matchesDr King George)U.S. District Judge Marc Thomas Treadwelltitle U.S. District Judge, fixing the xfailThe accepted cost. Most short titles are one-word abbreviations (
dr,mr,lt,col) that nobody writes with a dot between every letter. Their dotted forms are always initials, soM.R. King Georgewould read titleM.R. King. The alternative is a curated subset of titles that can be dotted (us,uk,ceo,vc,bg,mg,pfc, …). That's more precise, but it's another list to maintain. We leaned toward accepting the cost and pinning that case in a test.No conflict with the 2.4 dotted-suffix switch.
Policy.unlisted_dotted_suffixes(rules.md#S3) acts only at the end of a name and explicitly excludes leading dotted runs. This rule acts only at the front, so the two never compete for the same token.Also considered, not pursued for now:
- Splitting
E.T. Smithinto givenE., middleT.(the xfailtest_two_initials_conflict_with_conjunction, which still cites this issue). Any split must happen after roles are assigned, or S3 would see two initials instead of one dotted token inJohn Smith E.T.. initials()on a dotted token. It gives one letter per token:J.R.R. Tolkien→J. T.,E.T. Smith→E. S.. This changes no roles, so it would be the safest piece to pick up first.
- Splitting
- changed the title
[-]Should dotted single letters read as initials when they collide with title and credential vocabulary? ("E.T. Smith", "U.S. District Judge")[/-][+]Should a dotted title like "U.S. District Judge" read as a title, the way "US District Judge" does?[/+]on Oct 3, 2026 - added a commit that references this issue
on Oct 8, 2026
Two long-standing xfails share one signal: letters separated by periods smell like initials.
E.T. Smithgivesgiven="E.T."butgiven="E.",middle="T."could be a nice optional correction.U.S. District Judge Marc Thomas Treadwellfails to readU.S.as a title in a chained title, but it could.The hazard is that the same shape is credential vocabulary — SUFFIX_ACRONYMS contains a specific list like
M.D.,Ph.D.— so a bare "dotted single letters = initials" rule collides withJohn Smith M.D.and the S-rules' acronym handling. But potentially anything at the end of a name that smells like initials could be a suffix candidate beyond what we have recorded in the constants.The question: is there a version of the signal that we could use in to inform decisions about first/middle initial split correction, potential title or suffix acronym?