Skip to content

tuomiokirjat_search

Full-text search over tuomiokirjat — 7.8 million machine-transcribed pages of Finnish lower-court records, 1610–1931, from 223 archives: the judgement books and minutes of the town courts (raastuvanoikeus), the district courts (kihlakunnanoikeus, by tuomiokunta), the town bailiffs' courts (kämnerinoikeus), the land-partition courts and two appeal courts. Each hit is one page of a court volume, led by its archival citation and linked to its image in Astia.

Parameters

parameter type default notes
keyword string (required) Search term in Swedish and period spelling. Swedish stemming and accent folding apply. Several words require all of them; "quoted words" require that exact phrase; bref\|breff matches either spelling. A trailing * is not available here — the corpus is too large for a vocabulary — so list spellings, or use fuzzy. Only the page text is indexed.
offset int ≥ 0 0 Pagination start.
limit int 1–100 25 Results per page.
collection string — The archive, as a case-insensitive substring over 223 Finnish names: Turun raastuvanoikeuden, Etelä-Pohjanmaan, Viipurin hovioikeuden; tuomiokunnan alone reaches every district court.
series string — The series, as a case-insensitive substring over 444 names. For the 17th–18th-century town courts it names the court (Turun raastuvanoikeuden tuomiokirjat); for the 19th–20th-century district courts the record type — Varsinaisten asioiden pöytäkirjat (cases) or Ilmoitusasioiden pöytäkirjat (registrations).
year_min int — Earliest year of the volume.
year_max int — Latest year of the volume.
fuzzy int 0-2 0 Edit distance per term. Spelling was never standardised, so 1 is the second attempt when a search looks thin. Pass a base form: a fuzzy term skips stemming.
match_all bool true Require every word. false matches any word — and the total then counts pages matching only one.
research_context string — A sentence on what the user is researching. Not searched; written to the server log beside the query, so an operator can see what the corpus is asked for. ra-mcp's tools take the same parameter.

How the filters behave

collection and series are substrings, case-insensitively, because 223 and 444 values are too many for a fixed choice and a court's name is how a caller reaches for them. The two work together: for a 19th-century district court the archive names the district and the series the record type, so collection="Lappeen" with series="Varsinaisten" is that district's court cases.

year_min / year_max test interval overlap over the volume's year range. 83 pages have no year and are left out whenever a year is set. 703 are catalogued with the end before the start; they are dated by whichever bound falls within the corpus — so the 139 pages catalogued 1984–1895 filter as 1895, and a volume reading 1899–1898 as 1899.

An inverted range is rejected up front rather than returning a silent empty result.

Examples

tuomiokirjat_search(keyword="hustru", series="Turun raastuvanoikeuden", year_min=1650, year_max=1660)
tuomiokirjat_search(keyword="Larsson", collection="Raastuvanoikeuksien renovoidut")
tuomiokirjat_search(keyword="lagfart", collection="Lappeen", series="Ilmoitusasioiden")
tuomiokirjat_search(keyword="\"then 3 Maji\"")                   # exact phrase
tuomiokirjat_search(keyword="breff", fuzzy=1, year_min=1620, year_max=1640)

Output

Tuomiokirjat search results for 'Larsson': showing 1 of 1 records (offset 0)

**Porin raastuvanoikeuden tuomiokirjat a:1 1622–1639, p. 66**
  Raastuvanoikeuksien renovoidut tuomiokirjat · page PpblIZcBCao99UPKvl9-
  Erich Larsson fick löftte förståndin Rätta 32 den 3 Februaryis heer Rådstugu och Smgård …
  https://astia.narc.fi/uusiastia/viewer/?fileId=5929607974&aineistoId=2317506414

Every hit leads with its citation — the series, the volume's signum, the year and the page — then the archive (and subseries, where there is one) and the page id, a snippet, and the page image in Astia. Pass the page id to tuomiokirjat_get_page for the whole page.

The total is a true count of matches up to 10,000, and reads 10000+ beyond it. On a corpus this size most common words pass the cap, so narrow by collection, series and years before paging: the total is then a number that means something.