From ce901a887071c896eca0005f1574aac9fbd8db90 Mon Sep 17 00:00:00 2001 From: Marty Schoch Date: Sat, 29 Apr 2017 18:46:58 -0400 Subject: [PATCH] add a pure Go German analyzer this introduces new German light stemmer and moves the other pure Go german analysis components back in from the blevex package the libstemmer version of the stemmer will remain in blevex --- analysis/lang/de/analyzer_de.go | 61 +++++ analysis/lang/de/analyzer_de_test.go | 155 +++++++++++ analysis/lang/de/german_normalize.go | 95 +++++++ analysis/lang/de/german_normalize_test.go | 103 +++++++ analysis/lang/de/light_stemmer_de.go | 116 ++++++++ analysis/lang/de/stop_filter_de.go | 33 +++ analysis/lang/de/stop_words_de.go | 318 ++++++++++++++++++++++ 7 files changed, 881 insertions(+) create mode 100644 analysis/lang/de/analyzer_de.go create mode 100644 analysis/lang/de/analyzer_de_test.go create mode 100644 analysis/lang/de/german_normalize.go create mode 100644 analysis/lang/de/german_normalize_test.go create mode 100644 analysis/lang/de/light_stemmer_de.go create mode 100644 analysis/lang/de/stop_filter_de.go create mode 100644 analysis/lang/de/stop_words_de.go diff --git a/analysis/lang/de/analyzer_de.go b/analysis/lang/de/analyzer_de.go new file mode 100644 index 00000000..dcf4e8e9 --- /dev/null +++ b/analysis/lang/de/analyzer_de.go @@ -0,0 +1,61 @@ +// Copyright (c) 2017 Couchbase, Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package de + +import ( + "github.com/blevesearch/bleve/analysis" + "github.com/blevesearch/bleve/analysis/token/lowercase" + "github.com/blevesearch/bleve/analysis/tokenizer/unicode" + "github.com/blevesearch/bleve/registry" +) + +const AnalyzerName = "de" + +func AnalyzerConstructor(config map[string]interface{}, cache *registry.Cache) (*analysis.Analyzer, error) { + unicodeTokenizer, err := cache.TokenizerNamed(unicode.Name) + if err != nil { + return nil, err + } + toLowerFilter, err := cache.TokenFilterNamed(lowercase.Name) + if err != nil { + return nil, err + } + stopDeFilter, err := cache.TokenFilterNamed(StopName) + if err != nil { + return nil, err + } + normalizeDeFilter, err := cache.TokenFilterNamed(NormalizeName) + if err != nil { + return nil, err + } + lightStemmerDeFilter, err := cache.TokenFilterNamed(LightStemmerName) + if err != nil { + return nil, err + } + rv := analysis.Analyzer{ + Tokenizer: unicodeTokenizer, + TokenFilters: []analysis.TokenFilter{ + toLowerFilter, + stopDeFilter, + normalizeDeFilter, + lightStemmerDeFilter, + }, + } + return &rv, nil +} + +func init() { + registry.RegisterAnalyzer(AnalyzerName, AnalyzerConstructor) +} diff --git a/analysis/lang/de/analyzer_de_test.go b/analysis/lang/de/analyzer_de_test.go new file mode 100644 index 00000000..18e2683b --- /dev/null +++ b/analysis/lang/de/analyzer_de_test.go @@ -0,0 +1,155 @@ +// Copyright (c) 2017 Couchbase, Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package de + +import ( + "reflect" + "testing" + + "github.com/blevesearch/bleve/analysis" + "github.com/blevesearch/bleve/registry" +) + +func TestGermanAnalyzer(t *testing.T) { + tests := []struct { + input []byte + output analysis.TokenStream + }{ + { + input: []byte("Tisch"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("tisch"), + Position: 1, + Start: 0, + End: 5, + }, + }, + }, + { + input: []byte("Tische"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("tisch"), + Position: 1, + Start: 0, + End: 6, + }, + }, + }, + { + input: []byte("Tischen"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("tisch"), + Position: 1, + Start: 0, + End: 7, + }, + }, + }, + // german specials + { + input: []byte("Schaltflächen"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("schaltflach"), + Position: 1, + Start: 0, + End: 14, + }, + }, + }, + { + input: []byte("Schaltflaechen"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("schaltflach"), + Position: 1, + Start: 0, + End: 14, + }, + }, + }, + // tests added by marty to increase coverage + { + input: []byte("Blechern"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("blech"), + Position: 1, + Start: 0, + End: 8, + }, + }, + }, + { + input: []byte("Klecks"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("kleck"), + Position: 1, + Start: 0, + End: 6, + }, + }, + }, + { + input: []byte("Mindestens"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("mindest"), + Position: 1, + Start: 0, + End: 10, + }, + }, + }, + { + input: []byte("Kugelfest"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("kugelf"), + Position: 1, + Start: 0, + End: 9, + }, + }, + }, + { + input: []byte("Baldigst"), + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("baldig"), + Position: 1, + Start: 0, + End: 8, + }, + }, + }, + } + + cache := registry.NewCache() + analyzer, err := cache.AnalyzerNamed(AnalyzerName) + if err != nil { + t.Fatal(err) + } + for _, test := range tests { + actual := analyzer.Analyze(test.input) + if !reflect.DeepEqual(actual, test.output) { + t.Errorf("expected %v, got %v", test.output, actual) + } + } +} diff --git a/analysis/lang/de/german_normalize.go b/analysis/lang/de/german_normalize.go new file mode 100644 index 00000000..370efcbe --- /dev/null +++ b/analysis/lang/de/german_normalize.go @@ -0,0 +1,95 @@ +// Copyright (c) 2017 Couchbase, Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package de + +import ( + "bytes" + + "github.com/blevesearch/bleve/analysis" + "github.com/blevesearch/bleve/registry" +) + +const NormalizeName = "normalize_de" + +const ( + N = 0 /* ordinary state */ + V = 1 /* stops 'u' from entering umlaut state */ + U = 2 /* umlaut state, allows e-deletion */ +) + +type GermanNormalizeFilter struct { +} + +func NewGermanNormalizeFilter() *GermanNormalizeFilter { + return &GermanNormalizeFilter{} +} + +func (s *GermanNormalizeFilter) Filter(input analysis.TokenStream) analysis.TokenStream { + for _, token := range input { + term := normalize(token.Term) + token.Term = term + } + return input +} + +func normalize(input []byte) []byte { + state := N + runes := bytes.Runes(input) + for i := 0; i < len(runes); i++ { + switch runes[i] { + case 'a', 'o': + state = U + case 'u': + if state == N { + state = U + } else { + state = V + } + case 'e': + if state == U { + runes = analysis.DeleteRune(runes, i) + i-- + } + state = V + case 'i', 'q', 'y': + state = V + case 'ä': + runes[i] = 'a' + state = V + case 'ö': + runes[i] = 'o' + state = V + case 'ü': + runes[i] = 'u' + state = V + case 'ß': + runes[i] = 's' + i++ + runes = analysis.InsertRune(runes, i, 's') + state = N + default: + state = N + } + } + return analysis.BuildTermFromRunes(runes) +} + +func NormalizerFilterConstructor(config map[string]interface{}, cache *registry.Cache) (analysis.TokenFilter, error) { + return NewGermanNormalizeFilter(), nil +} + +func init() { + registry.RegisterTokenFilter(NormalizeName, NormalizerFilterConstructor) +} diff --git a/analysis/lang/de/german_normalize_test.go b/analysis/lang/de/german_normalize_test.go new file mode 100644 index 00000000..b95a7190 --- /dev/null +++ b/analysis/lang/de/german_normalize_test.go @@ -0,0 +1,103 @@ +// Copyright (c) 2017 Couchbase, Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package de + +import ( + "reflect" + "testing" + + "github.com/blevesearch/bleve/analysis" +) + +func TestGermanNormalizeFilter(t *testing.T) { + tests := []struct { + input analysis.TokenStream + output analysis.TokenStream + }{ + // Tests that a/o/u + e is equivalent to the umlaut form + { + input: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("Schaltflächen"), + }, + }, + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("Schaltflachen"), + }, + }, + }, + { + input: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("Schaltflaechen"), + }, + }, + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("Schaltflachen"), + }, + }, + }, + // Tests the specific heuristic that ue is not folded after a vowel or q. + { + input: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("dauer"), + }, + }, + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("dauer"), + }, + }, + }, + // Tests german specific folding of sharp-s + { + input: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("weißbier"), + }, + }, + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte("weissbier"), + }, + }, + }, + // empty + { + input: analysis.TokenStream{ + &analysis.Token{ + Term: []byte(""), + }, + }, + output: analysis.TokenStream{ + &analysis.Token{ + Term: []byte(""), + }, + }, + }, + } + + germanNormalizeFilter := NewGermanNormalizeFilter() + for _, test := range tests { + actual := germanNormalizeFilter.Filter(test.input) + if !reflect.DeepEqual(actual, test.output) { + t.Errorf("expected %#v, got %#v", test.output, actual) + t.Errorf("expected %s(% x), got %s(% x)", test.output[0].Term, test.output[0].Term, actual[0].Term, actual[0].Term) + } + } +} diff --git a/analysis/lang/de/light_stemmer_de.go b/analysis/lang/de/light_stemmer_de.go new file mode 100644 index 00000000..c1a093c5 --- /dev/null +++ b/analysis/lang/de/light_stemmer_de.go @@ -0,0 +1,116 @@ +// Copyright (c) 2017 Couchbase, Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package de + +import ( + "bytes" + + "github.com/blevesearch/bleve/analysis" + "github.com/blevesearch/bleve/registry" +) + +const LightStemmerName = "stemmer_de_light" + +type GermanLightStemmerFilter struct { +} + +func NewGermanLightStemmerFilter() *GermanLightStemmerFilter { + return &GermanLightStemmerFilter{} +} + +func (s *GermanLightStemmerFilter) Filter(input analysis.TokenStream) analysis.TokenStream { + for _, token := range input { + runes := bytes.Runes(token.Term) + runes = stem(runes) + token.Term = analysis.BuildTermFromRunes(runes) + } + return input +} + +func stem(input []rune) []rune { + + for i, r := range input { + switch r { + case 'ä', 'à', 'á', 'â': + input[i] = 'a' + case 'ö', 'ò', 'ó', 'ô': + input[i] = 'o' + case 'ï', 'ì', 'í', 'î': + input[i] = 'i' + case 'ü', 'ù', 'ú', 'û': + input[i] = 'u' + } + } + + input = step1(input) + return step2(input) +} + +func stEnding(ch rune) bool { + switch ch { + case 'b', 'd', 'f', 'g', 'h', 'k', 'l', 'm', 'n', 't': + return true + } + return false +} + +func step1(s []rune) []rune { + l := len(s) + if l > 5 && s[l-3] == 'e' && s[l-2] == 'r' && s[l-1] == 'n' { + return s[:l-3] + } + + if l > 4 && s[l-2] == 'e' { + switch s[l-1] { + case 'm', 'n', 'r', 's': + return s[:l-2] + } + } + + if l > 3 && s[l-1] == 'e' { + return s[:l-1] + } + + if l > 3 && s[l-1] == 's' && stEnding(s[l-2]) { + return s[:l-1] + } + + return s +} + +func step2(s []rune) []rune { + l := len(s) + if l > 5 && s[l-3] == 'e' && s[l-2] == 's' && s[l-1] == 't' { + return s[:l-3] + } + + if l > 4 && s[l-2] == 'e' && (s[l-1] == 'r' || s[l-1] == 'n') { + return s[:l-2] + } + + if l > 4 && s[l-2] == 's' && s[l-1] == 't' && stEnding(s[l-3]) { + return s[:l-2] + } + + return s +} + +func GermanLightStemmerFilterConstructor(config map[string]interface{}, cache *registry.Cache) (analysis.TokenFilter, error) { + return NewGermanLightStemmerFilter(), nil +} + +func init() { + registry.RegisterTokenFilter(LightStemmerName, GermanLightStemmerFilterConstructor) +} diff --git a/analysis/lang/de/stop_filter_de.go b/analysis/lang/de/stop_filter_de.go new file mode 100644 index 00000000..bcc423f0 --- /dev/null +++ b/analysis/lang/de/stop_filter_de.go @@ -0,0 +1,33 @@ +// Copyright (c) 2017 Couchbase, Inc. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +package de + +import ( + "github.com/blevesearch/bleve/analysis" + "github.com/blevesearch/bleve/analysis/token/stop" + "github.com/blevesearch/bleve/registry" +) + +func StopTokenFilterConstructor(config map[string]interface{}, cache *registry.Cache) (analysis.TokenFilter, error) { + tokenMap, err := cache.TokenMapNamed(StopName) + if err != nil { + return nil, err + } + return stop.NewStopTokensFilter(tokenMap), nil +} + +func init() { + registry.RegisterTokenFilter(StopName, StopTokenFilterConstructor) +} diff --git a/analysis/lang/de/stop_words_de.go b/analysis/lang/de/stop_words_de.go new file mode 100644 index 00000000..b71c8f70 --- /dev/null +++ b/analysis/lang/de/stop_words_de.go @@ -0,0 +1,318 @@ +package de + +import ( + "github.com/blevesearch/bleve/analysis" + "github.com/blevesearch/bleve/registry" +) + +const StopName = "stop_de" + +// this content was obtained from: +// lucene-4.7.2/analysis/common/src/resources/org/apache/lucene/analysis/snowball/ +// ` was changed to ' to allow for literal string + +var GermanStopWords = []byte(` | From svn.tartarus.org/snowball/trunk/website/algorithms/german/stop.txt + | This file is distributed under the BSD License. + | See http://snowball.tartarus.org/license.php + | Also see http://www.opensource.org/licenses/bsd-license.html + | - Encoding was converted to UTF-8. + | - This notice was added. + | + | NOTE: To use this file with StopFilterFactory, you must specify format="snowball" + + | A German stop word list. Comments begin with vertical bar. Each stop + | word is at the start of a line. + + | The number of forms in this list is reduced significantly by passing it + | through the German stemmer. + + +aber | but + +alle | all +allem +allen +aller +alles + +als | than, as +also | so +am | an + dem +an | at + +ander | other +andere +anderem +anderen +anderer +anderes +anderm +andern +anderr +anders + +auch | also +auf | on +aus | out of +bei | by +bin | am +bis | until +bist | art +da | there +damit | with it +dann | then + +der | the +den +des +dem +die +das + +daß | that + +derselbe | the same +derselben +denselben +desselben +demselben +dieselbe +dieselben +dasselbe + +dazu | to that + +dein | thy +deine +deinem +deinen +deiner +deines + +denn | because + +derer | of those +dessen | of him + +dich | thee +dir | to thee +du | thou + +dies | this +diese +diesem +diesen +dieser +dieses + + +doch | (several meanings) +dort | (over) there + + +durch | through + +ein | a +eine +einem +einen +einer +eines + +einig | some +einige +einigem +einigen +einiger +einiges + +einmal | once + +er | he +ihn | him +ihm | to him + +es | it +etwas | something + +euer | your +eure +eurem +euren +eurer +eures + +für | for +gegen | towards +gewesen | p.p. of sein +hab | have +habe | have +haben | have +hat | has +hatte | had +hatten | had +hier | here +hin | there +hinter | behind + +ich | I +mich | me +mir | to me + + +ihr | you, to her +ihre +ihrem +ihren +ihrer +ihres +euch | to you + +im | in + dem +in | in +indem | while +ins | in + das +ist | is + +jede | each, every +jedem +jeden +jeder +jedes + +jene | that +jenem +jenen +jener +jenes + +jetzt | now +kann | can + +kein | no +keine +keinem +keinen +keiner +keines + +können | can +könnte | could +machen | do +man | one + +manche | some, many a +manchem +manchen +mancher +manches + +mein | my +meine +meinem +meinen +meiner +meines + +mit | with +muss | must +musste | had to +nach | to(wards) +nicht | not +nichts | nothing +noch | still, yet +nun | now +nur | only +ob | whether +oder | or +ohne | without +sehr | very + +sein | his +seine +seinem +seinen +seiner +seines + +selbst | self +sich | herself + +sie | they, she +ihnen | to them + +sind | are +so | so + +solche | such +solchem +solchen +solcher +solches + +soll | shall +sollte | should +sondern | but +sonst | else +über | over +um | about, around +und | and + +uns | us +unse +unsem +unsen +unser +unses + +unter | under +viel | much +vom | von + dem +von | from +vor | before +während | while +war | was +waren | were +warst | wast +was | what +weg | away, off +weil | because +weiter | further + +welche | which +welchem +welchen +welcher +welches + +wenn | when +werde | will +werden | will +wie | how +wieder | again +will | want +wir | we +wird | will +wirst | willst +wo | where +wollen | want +wollte | wanted +würde | would +würden | would +zu | to +zum | zu + dem +zur | zu + der +zwar | indeed +zwischen | between + +`) + +func TokenMapConstructor(config map[string]interface{}, cache *registry.Cache) (analysis.TokenMap, error) { + rv := analysis.NewTokenMap() + err := rv.LoadBytes(GermanStopWords) + return rv, err +} + +func init() { + registry.RegisterTokenMap(StopName, TokenMapConstructor) +}