0
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
1 package de.mpg.mpiwg.berlin.mpdl.donatus.analysis;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
2
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
3 import org.apache.log4j.Logger;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
4
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
5 import de.mpg.mpiwg.berlin.mpdl.donatus.xmlrpc.DonatusCache;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
6 import de.mpg.mpiwg.berlin.mpdl.donatus.xmlrpc.DonatusConstants;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
7 import de.mpg.mpiwg.berlin.mpdl.donatus.xmlrpc.DonatusLemma;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
8 import de.mpg.mpiwg.berlin.mpdl.exception.ApplicationException;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
9
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
10 public class DonatusStemmer {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
11 private String language = DonatusConstants.DEFAULT_LANGUAGE;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
12
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
13 protected void setLanguage(String language) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
14 this.language = language;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
15 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
16
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
17 /**
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
18 * Used for indexing documents and for querying
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
19 * @param term
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
20 * @return
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
21 */
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
22 protected String stem(String term) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
23 String stem = null;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
24 term = term.toLowerCase();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
25 // try to find the stem by the DonatusCache
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
26 DonatusLemma donatusLemma = null;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
27 try {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
28 DonatusCache donatusCache = DonatusCache.getInstance();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
29 donatusLemma = donatusCache.getLemmaByVariantForm(language, term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
30 } catch (ApplicationException e) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
31 // nothing, do not disturb
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
32 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
33 if (donatusLemma != null)
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
34 stem = donatusLemma.getForm();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
35 // if not found by Donatus try to use Snowball (or later other language specific stemmers)
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
36 if (stem == null) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
37 stem = stemBySnowball(term, language);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
38 // if term is not equal to the base form and also the stem is not too short (> 2 characters) then add this Snowball variant to the lemmas in cache
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
39 if ((! stem.equals(term)) && stem.length() > 2) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
40 try {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
41 DonatusCache donatusCache = DonatusCache.getInstance();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
42 if (donatusCache.getMode() == DonatusCache.DOCUMENT_MODE) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
43 donatusCache.addVariant(language, stem, DonatusConstants.TYPE_SNOWBALL, term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
44 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
45 } catch (ApplicationException e) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
46 Logger.getLogger(DonatusStemmer.class).warn("DonatusCache: an exception was caught while indexing a document: " + e.getMessage(), e);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
47 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
48 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
49 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
50 /* TODO if Snowball is too bad (for some languages) use Lucene analyzers
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
51 if (stem == null) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
52 stem = stemByLanguageStemmers(term, this.language);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
53 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
54 */
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
55 return stem;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
56 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
57
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
58 private String stemBySnowball(String term, String language) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
59 String stem = null;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
60 if (language.equals("de")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
61 net.sf.snowball.ext.GermanStemmer stemmer = new net.sf.snowball.ext.GermanStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
62 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
63 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
64 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
65 } else if (language.equals("en")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
66 net.sf.snowball.ext.EnglishStemmer stemmer = new net.sf.snowball.ext.EnglishStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
67 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
68 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
69 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
70 } else if (language.equals("nl")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
71 net.sf.snowball.ext.DutchStemmer stemmer = new net.sf.snowball.ext.DutchStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
72 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
73 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
74 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
75 } else if (language.equals("fi")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
76 net.sf.snowball.ext.FinnishStemmer stemmer = new net.sf.snowball.ext.FinnishStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
77 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
78 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
79 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
80 } else if (language.equals("fr")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
81 net.sf.snowball.ext.FrenchStemmer stemmer = new net.sf.snowball.ext.FrenchStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
82 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
83 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
84 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
85 } else if (language.equals("it")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
86 net.sf.snowball.ext.ItalianStemmer stemmer = new net.sf.snowball.ext.ItalianStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
87 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
88 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
89 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
90 } else if (language.equals("no")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
91 net.sf.snowball.ext.NorwegianStemmer stemmer = new net.sf.snowball.ext.NorwegianStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
92 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
93 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
94 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
95 } else if (language.equals("pt")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
96 net.sf.snowball.ext.PortugueseStemmer stemmer = new net.sf.snowball.ext.PortugueseStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
97 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
98 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
99 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
100 } else if (language.equals("ru")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
101 net.sf.snowball.ext.RussianStemmer stemmer = new net.sf.snowball.ext.RussianStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
102 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
103 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
104 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
105 } else if (language.equals("es")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
106 net.sf.snowball.ext.SpanishStemmer stemmer = new net.sf.snowball.ext.SpanishStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
107 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
108 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
109 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
110 } else if (language.equals("sv")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
111 net.sf.snowball.ext.SwedishStemmer stemmer = new net.sf.snowball.ext.SwedishStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
112 stemmer.setCurrent(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
113 stemmer.stem();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
114 stem = stemmer.getCurrent();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
115 } else {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
116 stem = term; // if no language fits deliver the term itself as the stem form
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
117 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
118 return stem;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
119 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
120
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
121 /*
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
122 private String stemByLanguageStemmers(String term, String language) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
123 // TODO provide other languages
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
124 String stem = null;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
125 if (language.equals("br")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
126 BrazilianStemmer stemmer = new BrazilianStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
127 stem = stemmer.stem(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
128 } else if (language.equals("de")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
129 GermanStemmer stemmer = new GermanStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
130 stem = stemmer.stem(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
131 } else if (language.equals("fr")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
132 FrenchStemmer stemmer = new FrenchStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
133 stem = stemmer.stem(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
134 } else if (language.equals("nl")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
135 DutchStemmer stemmer = new DutchStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
136 stem = stemmer.stem(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
137 } else if (language.equals("ru")) {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
138 RussianStemmer stemmer = new RussianStemmer();
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
139 stem = stemmer.stem(term);
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
140 } else {
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
141 stem = term; // if no language fits deliver the term itself as the stem form
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
142 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
143 return stem;
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
144 }
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
145 */
|
Josef Willenborg <jwillenborg@mpiwg-berlin.mpg.de>
parents:
diff
changeset
|
146 }
|