settings/icu_tokenizer.yaml

   1 normalization:
   2     - ":: lower ()"
   3     - ":: Hans-Hant"
   4     - !include icu-rules/unicode-digits-to-decimal.yaml
   5     - "'№' > 'no'"
   6     - "'n°' > 'no'"
   7     - "'nº' > 'no'"
   8     - "ª > a"
   9     - "º > o"
  10     - "[[:Punctuation:][:Symbol:]\u02bc]  > ' '"
  11     - "ß > 'ss'" # German szet is unambiguously equal to double ss
  12     - "[^[:alnum:] [:Canonical_Combining_Class=Virama:] [:Space:]] >"
  13     - "[:Lm:] >"
  14     - ":: [[:Number:]] Latin ()"
  15     - ":: [[:Number:]] Ascii ();"
  16     - ":: [[:Number:]] NFD ();"
  17     - "[[:Nonspacing Mark:] [:Cf:]] >;"
  18     - "[:Space:]+ > ' '"
  19 transliteration:
  20     - ":: Latin ()"
  21     - !include icu-rules/extended-unicode-to-asccii.yaml
  22     - ":: Ascii ()"
  23     - ":: NFD ()"
  24     - ":: lower ()"
  25     - "[^a-z0-9[:Space:]] >"
  26     - ":: NFC ()"
  27     - "[:Space:]+ > ' '"
  28 sanitizers:
  29     - step: clean-housenumbers
  30       filter-kind:
  31         - housenumber
  32         - conscriptionnumber
  33         - streetnumber
  34       convert-to-name:
  35         - (\A|.*,)[^\d,]{3,}(,.*|\Z)
  36     - step: clean-postcodes
  37       convert-to-address: yes
  38       default-pattern: "[A-Z0-9- ]{3,12}"
  39     - step: clean-tiger-tags
  40     - step: split-name-list
  41     - step: strip-brace-terms
  42     - step: tag-analyzer-by-language
  43       filter-kind: [".*name.*"]
  44       whitelist: [bg,ca,cs,da,de,el,en,es,et,eu,fi,fr,gl,hu,it,ja,mg,ms,nl,no,pl,pt,ro,ru,sk,sl,sv,tr,uk,vi]
  45       use-defaults: all
  46       mode: append
  47 token-analysis:
  48     - analyzer: generic
  49     - id: "@housenumber"
  50       analyzer: housenumbers
  51     - id: "@postcode"
  52       analyzer: postcodes
  53     - id: bg
  54       analyzer: generic
  55       mode: variant-only
  56       variants:
  57           - !include icu-rules/variants-bg.yaml
  58     - id: ca
  59       analyzer: generic
  60       mode: variant-only
  61       variants:
  62           - !include icu-rules/variants-ca.yaml
  63     - id: cs
  64       analyzer: generic
  65       mode: variant-only
  66       variants:
  67           - !include icu-rules/variants-cs.yaml
  68     - id: da
  69       analyzer: generic
  70       mode: variant-only
  71       variants:
  72           - !include icu-rules/variants-da.yaml
  73     - id: de
  74       analyzer: generic
  75       mode: variant-only
  76       variants:
  77           - !include icu-rules/variants-de.yaml
  78       mutations:
  79           - pattern: ä
  80             replacements: ["ä", "ae"]
  81           - pattern: ö
  82             replacements: ["ö", "oe"]
  83           - pattern: ü
  84             replacements: ["ü", "ue"]
  85     - id: el
  86       analyzer: generic
  87       mode: variant-only
  88       variants:
  89           - !include icu-rules/variants-el.yaml
  90     - id: en
  91       analyzer: generic
  92       mode: variant-only
  93       variants:
  94           - !include icu-rules/variants-en.yaml
  95     - id: es
  96       analyzer: generic
  97       mode: variant-only
  98       variants:
  99           - !include icu-rules/variants-es.yaml
 100     - id: et
 101       analyzer: generic
 102       mode: variant-only
 103       variants:
 104           - !include icu-rules/variants-et.yaml
 105     - id: eu
 106       analyzer: generic
 107       mode: variant-only
 108       variants:
 109           - !include icu-rules/variants-eu.yaml
 110     - id: fi
 111       analyzer: generic
 112       mode: variant-only
 113       variants:
 114           - !include icu-rules/variants-fi.yaml
 115     - id: fr
 116       analyzer: generic
 117       mode: variant-only
 118       variants:
 119           - !include icu-rules/variants-fr.yaml
 120     - id: gl
 121       analyzer: generic
 122       mode: variant-only
 123       variants:
 124           - !include icu-rules/variants-gl.yaml
 125     - id: hu
 126       analyzer: generic
 127       mode: variant-only
 128       variants:
 129           - !include icu-rules/variants-hu.yaml
 130     - id: it
 131       analyzer: generic
 132       mode: variant-only
 133       variants:
 134           - !include icu-rules/variants-it.yaml
 135     - id: ja
 136       analyzer: generic
 137       mode: variant-only
 138       variants:
 139           - !include icu-rules/variants-ja.yaml
 140     - id: mg
 141       analyzer: generic
 142       mode: variant-only
 143       variants:
 144           - !include icu-rules/variants-mg.yaml
 145     - id: ms
 146       analyzer: generic
 147       mode: variant-only
 148       variants:
 149           - !include icu-rules/variants-ms.yaml
 150     - id: nl
 151       analyzer: generic
 152       mode: variant-only
 153       variants:
 154           - !include icu-rules/variants-nl.yaml
 155     - id: no
 156       analyzer: generic
 157       mode: variant-only
 158       variants:
 159           - !include icu-rules/variants-no.yaml
 160     - id: pl
 161       analyzer: generic
 162       mode: variant-only
 163       variants:
 164           - !include icu-rules/variants-pl.yaml
 165     - id: pt
 166       analyzer: generic
 167       mode: variant-only
 168       variants:
 169           - !include icu-rules/variants-pt.yaml
 170     - id: ro
 171       analyzer: generic
 172       mode: variant-only
 173       variants:
 174           - !include icu-rules/variants-ro.yaml
 175     - id: ru
 176       analyzer: generic
 177       mode: variant-only
 178       variants:
 179           - !include icu-rules/variants-ru.yaml
 180     - id: sk
 181       analyzer: generic
 182       mode: variant-only
 183       variants:
 184           - !include icu-rules/variants-sk.yaml
 185     - id: sl
 186       analyzer: generic
 187       mode: variant-only
 188       variants:
 189           - !include icu-rules/variants-sl.yaml
 190     - id: sv
 191       analyzer: generic
 192       mode: variant-only
 193       variants:
 194           - !include icu-rules/variants-sv.yaml
 195     - id: tr
 196       analyzer: generic
 197       mode: variant-only
 198       variants:
 199           - !include icu-rules/variants-tr.yaml
 200     - id: uk
 201       analyzer: generic
 202       mode: variant-only
 203       variants:
 204           - !include icu-rules/variants-uk.yaml
 205     - id: vi
 206       analyzer: generic
 207       mode: variant-only
 208       variants:
 209           - !include icu-rules/variants-vi.yaml