settings/icu_tokenizer.yaml

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