gibheer/bleve
0
0
Fork 0
Code

fixed issues with portuguese analyzer

Browse Source 
fixes #70
This commit is contained in:
Marty Schoch 2015-03-11 14:22:11 -04:00
parent 7970f42c29
commit 50bd082257
4 changed files with 609 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
analysis/language/pt/analyzer_pt.go
View File

@ -7,9 +7,6 @@
// either express or implied. See the License for the specific language governing permissions
// and limitations under the License.
// +build libstemmer full
// +build icu full
package pt
import (
@ -17,13 +14,13 @@ import (
"github.com/blevesearch/bleve/registry"
"github.com/blevesearch/bleve/analysis/token_filters/lower_case_filter"
"github.com/blevesearch/bleve/analysis/tokenizers/icu"
"github.com/blevesearch/bleve/analysis/tokenizers/unicode"
)
const AnalyzerName = "pt"
func AnalyzerConstructor(config map[string]interface{}, cache *registry.Cache) (*analysis.Analyzer, error) {
icuTokenizer, err := cache.TokenizerNamed(icu.Name)
tokenizer, err := cache.TokenizerNamed(unicode.Name)
if err != nil {
return nil, err
}
@ -35,12 +32,12 @@ func AnalyzerConstructor(config map[string]interface{}, cache *registry.Cache) (
if err != nil {
return nil, err
}
stemmerPtFilter, err := cache.TokenFilterNamed(StemmerName)
stemmerPtFilter, err := cache.TokenFilterNamed(LightStemmerName)
if err != nil {
return nil, err
}
rv := analysis.Analyzer{
Tokenizer: icuTokenizer,
Tokenizer: tokenizer,
TokenFilters: []analysis.TokenFilter{
toLowerFilter,
stopPtFilter,

37
analysis/language/pt/analyzer_pt_test.go
View File

@ -7,9 +7,6 @@
// either express or implied. See the License for the specific language governing permissions
// and limitations under the License.
// +build libstemmer full
// +build icu full
package pt
import (
@ -26,24 +23,22 @@ func TestPortugueseAnalyzer(t *testing.T) {
output analysis.TokenStream
}{
// stemming
// fails due to stemming discrepencies
// got quilométr instead of quilometric
// {
// input: []byte("quilométricas"),
// output: analysis.TokenStream{
// &analysis.Token{
// Term: []byte("quilometric"),
// },
// },
// },
// {
// input: []byte("quilométricos"),
// output: analysis.TokenStream{
// &analysis.Token{
// Term: []byte("quilometric"),
// },
// },
// },
{
input: []byte("quilométricas"),
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("quilometric"),
},
},
},
{
input: []byte("quilométricos"),
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("quilometric"),
},
},
},
// stop word
{
input: []byte("não"),

190
analysis/language/pt/light_stemmer_pt.go Normal file
View File

@ -0,0 +1,190 @@
// Copyright (c) 2015 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 pt
import (
"bytes"
"github.com/blevesearch/bleve/analysis"
"github.com/blevesearch/bleve/registry"
)
const LightStemmerName = "stemmer_pt_light"
type PortugueseLightStemmerFilter struct {
}
func NewPortugueseLightStemmerFilter() *PortugueseLightStemmerFilter {
return &PortugueseLightStemmerFilter{}
}
func (s *PortugueseLightStemmerFilter) 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 {
inputLen := len(input)
if inputLen < 4 {
return input
}
input = removeSuffix(input)
inputLen = len(input)
if inputLen > 3 && input[inputLen-1] == 'a' {
input = normFeminine(input)
inputLen = len(input)
}
if inputLen > 4 {
switch input[inputLen-1] {
case 'e', 'a', 'o':
input = input[0 : inputLen-1]
inputLen = len(input)
}
}
for i := 0; i < inputLen; i++ {
switch input[i] {
case 'à', 'á', 'â', 'ä', 'ã':
input[i] = 'a'
case 'ò', 'ó', 'ô', 'ö', 'õ':
input[i] = 'o'
case 'è', 'é', 'ê', 'ë':
input[i] = 'e'
case 'ù', 'ú', 'û', 'ü':
input[i] = 'u'
case 'ì', 'í', 'î', 'ï':
input[i] = 'i'
case 'ç':
input[i] = 'c'
}
}
return input
}
func removeSuffix(input []rune) []rune {
inputLen := len(input)
if inputLen > 4 && analysis.RunesEndsWith(input, "es") {
switch input[inputLen-3] {
case 'r', 's', 'l', 'z':
return input[0 : inputLen-2]
}
}
if inputLen > 3 && analysis.RunesEndsWith(input, "ns") {
input[inputLen-2] = 'm'
return input[0 : inputLen-1]
}
if inputLen > 4 && (analysis.RunesEndsWith(input, "eis") || analysis.RunesEndsWith(input, "éis")) {
input[inputLen-3] = 'e'
input[inputLen-2] = 'l'
return input[0 : inputLen-1]
}
if inputLen > 4 && analysis.RunesEndsWith(input, "ais") {
input[inputLen-2] = 'l'
return input[0 : inputLen-1]
}
if inputLen > 4 && analysis.RunesEndsWith(input, "óis") {
input[inputLen-3] = 'o'
input[inputLen-2] = 'l'
return input[0 : inputLen-1]
}
if inputLen > 4 && analysis.RunesEndsWith(input, "is") {
input[inputLen-1] = 'l'
return input
}
if inputLen > 3 &&
(analysis.RunesEndsWith(input, "ões") ||
analysis.RunesEndsWith(input, "ães")) {
input = input[0 : inputLen-1]
inputLen = len(input)
input[inputLen-2] = 'ã'
input[inputLen-1] = 'o'
return input
}
if inputLen > 6 && analysis.RunesEndsWith(input, "mente") {
return input[0 : inputLen-5]
}
if inputLen > 3 && input[inputLen-1] == 's' {
return input[0 : inputLen-1]
}
return input
}
func normFeminine(input []rune) []rune {
inputLen := len(input)
if inputLen > 7 &&
(analysis.RunesEndsWith(input, "inha") ||
analysis.RunesEndsWith(input, "iaca") ||
analysis.RunesEndsWith(input, "eira")) {
input[inputLen-1] = 'o'
return input
}
if inputLen > 6 {
if analysis.RunesEndsWith(input, "osa") ||
analysis.RunesEndsWith(input, "ica") ||
analysis.RunesEndsWith(input, "ida") ||
analysis.RunesEndsWith(input, "ada") ||
analysis.RunesEndsWith(input, "iva") ||
analysis.RunesEndsWith(input, "ama") {
input[inputLen-1] = 'o'
return input
}
if analysis.RunesEndsWith(input, "ona") {
input[inputLen-3] = 'ã'
input[inputLen-2] = 'o'
return input[0 : inputLen-1]
}
if analysis.RunesEndsWith(input, "ora") {
return input[0 : inputLen-1]
}
if analysis.RunesEndsWith(input, "esa") {
input[inputLen-3] = 'ê'
return input[0 : inputLen-1]
}
if analysis.RunesEndsWith(input, "na") {
input[inputLen-1] = 'o'
return input
}
}
return input
}
func PortugueseLightStemmerFilterConstructor(config map[string]interface{}, cache *registry.Cache) (analysis.TokenFilter, error) {
return NewPortugueseLightStemmerFilter(), nil
}
func init() {
registry.RegisterTokenFilter(LightStemmerName, PortugueseLightStemmerFilterConstructor)
}

399
analysis/language/pt/light_stemmer_pt_test.go Normal file
View File

@ -0,0 +1,399 @@
// Copyright (c) 2015 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 pt
import (
"reflect"
"testing"
"github.com/blevesearch/bleve/analysis"
"github.com/blevesearch/bleve/registry"
)
func TestPortugueseLightStemmer(t *testing.T) {
tests := []struct {
input analysis.TokenStream
output analysis.TokenStream
}{
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("doutores"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("doutor"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("doutor"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("doutor"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("homens"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("homem"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("homem"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("homem"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("papéis"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("papel"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("papel"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("papel"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("normais"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("normal"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("normal"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("normal"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("lencóis"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("lencol"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("lencol"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("lencol"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("barris"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("barril"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("barril"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("barril"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("botões"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("bota"),
},
},
},
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("botão"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("bota"),
},
},
},
// longer
{
input: analysis.TokenStream{
&analysis.Token{
Term: []byte("o"),
},
&analysis.Token{
Term: []byte("debate"),
},
&analysis.Token{
Term: []byte("político"),
},
&analysis.Token{
Term: []byte("pelo"),
},
&analysis.Token{
Term: []byte("menos"),
},
&analysis.Token{
Term: []byte("o"),
},
&analysis.Token{
Term: []byte("que"),
},
&analysis.Token{
Term: []byte("vem"),
},
&analysis.Token{
Term: []byte("a"),
},
&analysis.Token{
Term: []byte("público"),
},
&analysis.Token{
Term: []byte("parece"),
},
&analysis.Token{
Term: []byte("de"),
},
&analysis.Token{
Term: []byte("modo"),
},
&analysis.Token{
Term: []byte("nada"),
},
&analysis.Token{
Term: []byte("surpreendente"),
},
&analysis.Token{
Term: []byte("restrito"),
},
&analysis.Token{
Term: []byte("a"),
},
&analysis.Token{
Term: []byte("temas"),
},
&analysis.Token{
Term: []byte("menores"),
},
&analysis.Token{
Term: []byte("mas"),
},
&analysis.Token{
Term: []byte("há"),
},
&analysis.Token{
Term: []byte("evidentemente"),
},
&analysis.Token{
Term: []byte("grandes"),
},
&analysis.Token{
Term: []byte("questões"),
},
&analysis.Token{
Term: []byte("em"),
},
&analysis.Token{
Term: []byte("jogo"),
},
&analysis.Token{
Term: []byte("nas"),
},
&analysis.Token{
Term: []byte("eleições"),
},
&analysis.Token{
Term: []byte("que"),
},
&analysis.Token{
Term: []byte("se"),
},
&analysis.Token{
Term: []byte("aproximam"),
},
},
output: analysis.TokenStream{
&analysis.Token{
Term: []byte("o"),
},
&analysis.Token{
Term: []byte("debat"),
},
&analysis.Token{
Term: []byte("politic"),
},
&analysis.Token{
Term: []byte("pelo"),
},
&analysis.Token{
Term: []byte("meno"),
},
&analysis.Token{
Term: []byte("o"),
},
&analysis.Token{
Term: []byte("que"),
},
&analysis.Token{
Term: []byte("vem"),
},
&analysis.Token{
Term: []byte("a"),
},
&analysis.Token{
Term: []byte("public"),
},
&analysis.Token{
Term: []byte("parec"),
},
&analysis.Token{
Term: []byte("de"),
},
&analysis.Token{
Term: []byte("modo"),
},
&analysis.Token{
Term: []byte("nada"),
},
&analysis.Token{
Term: []byte("surpreendent"),
},
&analysis.Token{
Term: []byte("restrit"),
},
&analysis.Token{
Term: []byte("a"),
},
&analysis.Token{
Term: []byte("tema"),
},
&analysis.Token{
Term: []byte("menor"),
},
&analysis.Token{
Term: []byte("mas"),
},
&analysis.Token{
Term: []byte("há"),
},
&analysis.Token{
Term: []byte("evident"),
},
&analysis.Token{
Term: []byte("grand"),
},
&analysis.Token{
Term: []byte("questa"),
},
&analysis.Token{
Term: []byte("em"),
},
&analysis.Token{
Term: []byte("jogo"),
},
&analysis.Token{
Term: []byte("nas"),
},
&analysis.Token{
Term: []byte("eleica"),
},
&analysis.Token{
Term: []byte("que"),
},
&analysis.Token{
Term: []byte("se"),
},
&analysis.Token{
Term: []byte("aproximam"),
},
},
},
}
cache := registry.NewCache()
filter, err := cache.TokenFilterNamed(LightStemmerName)
if err != nil {
t.Fatal(err)
}
for _, test := range tests {
actual := filter.Filter(test.input)
if !reflect.DeepEqual(actual, test.output) {
t.Errorf("expected %s, got %s", test.output[0].Term, actual[0].Term)
}
}
}