KickJava   Java API By Example, From Geeks To Geeks.

Java > Open Source Codes > org > apache > lucene > analysis > PorterStemmer


1 package org.apache.lucene.analysis;
2
3 /**
4  * Copyright 2004 The Apache Software Foundation
5  *
6  * Licensed under the Apache License, Version 2.0 (the "License");
7  * you may not use this file except in compliance with the License.
8  * You may obtain a copy of the License at
9  *
10  * http://www.apache.org/licenses/LICENSE-2.0
11  *
12  * Unless required by applicable law or agreed to in writing, software
13  * distributed under the License is distributed on an "AS IS" BASIS,
14  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15  * See the License for the specific language governing permissions and
16  * limitations under the License.
17  */
18
19 /*
20
21    Porter stemmer in Java. The original paper is in
22
23        Porter, 1980, An algorithm for suffix stripping, Program, Vol. 14,
24        no. 3, pp 130-137,
25
26    See also http://www.tartarus.org/~martin/PorterStemmer/index.html
27
28    Bug 1 (reported by Gonzalo Parra 16/10/99) fixed as marked below.
29    Tthe words 'aed', 'eed', 'oed' leave k at 'a' for step 3, and b[k-1]
30    is then out outside the bounds of b.
31
32    Similarly,
33
34    Bug 2 (reported by Steve Dyrdahl 22/2/00) fixed as marked below.
35    'ion' by itself leaves j = -1 in the test for 'ion' in step 5, and
36    b[j] is then outside the bounds of b.
37
38    Release 3.
39
40    [ This version is derived from Release 3, modified by Brian Goetz to
41      optimize for fewer object creations. ]
42
43 */
44
45
46 import java.io.*;
47
48 /**
49  *
50  * Stemmer, implementing the Porter Stemming Algorithm
51  *
52  * The Stemmer class transforms a word into its root form. The input
53  * word can be provided a character at time (by calling add()), or at once
54  * by calling one of the various stem(something) methods.
55  */
56
57 class PorterStemmer
58 {
59   private char[] b;
60   private int i, /* offset into b */
61     j, k, k0;
62   private boolean dirty = false;
63   private static final int INC = 50; /* unit of size whereby b is increased */
64   private static final int EXTRA = 1;
65
66   public PorterStemmer() {
67     b = new char[INC];
68     i = 0;
69   }
70
71   /**
72    * reset() resets the stemmer so it can stem another word. If you invoke
73    * the stemmer by calling add(char) and then stem(), you must call reset()
74    * before starting another word.
75    */
76   public void reset() { i = 0; dirty = false; }
77
78   /**
79    * Add a character to the word being stemmed. When you are finished
80    * adding characters, you can call stem(void) to process the word.
81    */
82   public void add(char ch) {
83     if (b.length <= i + EXTRA) {
84       char[] new_b = new char[b.length+INC];
85       for (int c = 0; c < b.length; c++)
86         new_b[c] = b[c];
87       b = new_b;
88     }
89     b[i++] = ch;
90   }
91
92   /**
93    * After a word has been stemmed, it can be retrieved by toString(),
94    * or a reference to the internal buffer can be retrieved by getResultBuffer
95    * and getResultLength (which is generally more efficient.)
96    */
97   public String JavaDoc toString() { return new String JavaDoc(b,0,i); }
98
99   /**
100    * Returns the length of the word resulting from the stemming process.
101    */
102   public int getResultLength() { return i; }
103
104   /**
105    * Returns a reference to a character buffer containing the results of
106    * the stemming process. You also need to consult getResultLength()
107    * to determine the length of the result.
108    */
109   public char[] getResultBuffer() { return b; }
110
111   /* cons(i) is true <=> b[i] is a consonant. */
112
113   private final boolean cons(int i) {
114     switch (b[i]) {
115     case 'a': case 'e': case 'i': case 'o': case 'u':
116       return false;
117     case 'y':
118       return (i==k0) ? true : !cons(i-1);
119     default:
120       return true;
121     }
122   }
123
124   /* m() measures the number of consonant sequences between k0 and j. if c is
125      a consonant sequence and v a vowel sequence, and <..> indicates arbitrary
126      presence,
127
128           <c><v> gives 0
129           <c>vc<v> gives 1
130           <c>vcvc<v> gives 2
131           <c>vcvcvc<v> gives 3
132           ....
133   */
134
135   private final int m() {
136     int n = 0;
137     int i = k0;
138     while(true) {
139       if (i > j)
140         return n;
141       if (! cons(i))
142         break;
143       i++;
144     }
145     i++;
146     while(true) {
147       while(true) {
148         if (i > j)
149           return n;
150         if (cons(i))
151           break;
152         i++;
153       }
154       i++;
155       n++;
156       while(true) {
157         if (i > j)
158           return n;
159         if (! cons(i))
160           break;
161         i++;
162       }
163       i++;
164     }
165   }
166
167   /* vowelinstem() is true <=> k0,...j contains a vowel */
168
169   private final boolean vowelinstem() {
170     int i;
171     for (i = k0; i <= j; i++)
172       if (! cons(i))
173         return true;
174     return false;
175   }
176
177   /* doublec(j) is true <=> j,(j-1) contain a double consonant. */
178
179   private final boolean doublec(int j) {
180     if (j < k0+1)
181       return false;
182     if (b[j] != b[j-1])
183       return false;
184     return cons(j);
185   }
186
187   /* cvc(i) is true <=> i-2,i-1,i has the form consonant - vowel - consonant
188      and also if the second c is not w,x or y. this is used when trying to
189      restore an e at the end of a short word. e.g.
190
191           cav(e), lov(e), hop(e), crim(e), but
192           snow, box, tray.
193
194   */
195
196   private final boolean cvc(int i) {
197     if (i < k0+2 || !cons(i) || cons(i-1) || !cons(i-2))
198       return false;
199     else {
200       int ch = b[i];
201       if (ch == 'w' || ch == 'x' || ch == 'y') return false;
202     }
203     return true;
204   }
205
206   private final boolean ends(String JavaDoc s) {
207     int l = s.length();
208     int o = k-l+1;
209     if (o < k0)
210       return false;
211     for (int i = 0; i < l; i++)
212       if (b[o+i] != s.charAt(i))
213         return false;
214     j = k-l;
215     return true;
216   }
217
218   /* setto(s) sets (j+1),...k to the characters in the string s, readjusting
219      k. */
220
221   void setto(String JavaDoc s) {
222     int l = s.length();
223     int o = j+1;
224     for (int i = 0; i < l; i++)
225       b[o+i] = s.charAt(i);
226     k = j+l;
227     dirty = true;
228   }
229
230   /* r(s) is used further down. */
231
232   void r(String JavaDoc s) { if (m() > 0) setto(s); }
233
234   /* step1() gets rid of plurals and -ed or -ing. e.g.
235
236            caresses -> caress
237            ponies -> poni
238            ties -> ti
239            caress -> caress
240            cats -> cat
241
242            feed -> feed
243            agreed -> agree
244            disabled -> disable
245
246            matting -> mat
247            mating -> mate
248            meeting -> meet
249            milling -> mill
250            messing -> mess
251
252            meetings -> meet
253
254   */
255
256   private final void step1() {
257     if (b[k] == 's') {
258       if (ends("sses")) k -= 2;
259       else if (ends("ies")) setto("i");
260       else if (b[k-1] != 's') k--;
261     }
262     if (ends("eed")) {
263       if (m() > 0)
264         k--;
265     }
266     else if ((ends("ed") || ends("ing")) && vowelinstem()) {
267       k = j;
268       if (ends("at")) setto("ate");
269       else if (ends("bl")) setto("ble");
270       else if (ends("iz")) setto("ize");
271       else if (doublec(k)) {
272         int ch = b[k--];
273         if (ch == 'l' || ch == 's' || ch == 'z')
274           k++;
275       }
276       else if (m() == 1 && cvc(k))
277         setto("e");
278     }
279   }
280
281   /* step2() turns terminal y to i when there is another vowel in the stem. */
282
283   private final void step2() {
284     if (ends("y") && vowelinstem()) {
285       b[k] = 'i';
286       dirty = true;
287     }
288   }
289
290   /* step3() maps double suffices to single ones. so -ization ( = -ize plus
291      -ation) maps to -ize etc. note that the string before the suffix must give
292      m() > 0. */
293
294   private final void step3() {
295     if (k == k0) return; /* For Bug 1 */
296     switch (b[k-1]) {
297     case 'a':
298       if (ends("ational")) { r("ate"); break; }
299       if (ends("tional")) { r("tion"); break; }
300       break;
301     case 'c':
302       if (ends("enci")) { r("ence"); break; }
303       if (ends("anci")) { r("ance"); break; }
304       break;
305     case 'e':
306       if (ends("izer")) { r("ize"); break; }
307       break;
308     case 'l':
309       if (ends("bli")) { r("ble"); break; }
310       if (ends("alli")) { r("al"); break; }
311       if (ends("entli")) { r("ent"); break; }
312       if (ends("eli")) { r("e"); break; }
313       if (ends("ousli")) { r("ous"); break; }
314       break;
315     case 'o':
316       if (ends("ization")) { r("ize"); break; }
317       if (ends("ation")) { r("ate"); break; }
318       if (ends("ator")) { r("ate"); break; }
319       break;
320     case 's':
321       if (ends("alism")) { r("al"); break; }
322       if (ends("iveness")) { r("ive"); break; }
323       if (ends("fulness")) { r("ful"); break; }
324       if (ends("ousness")) { r("ous"); break; }
325       break;
326     case 't':
327       if (ends("aliti")) { r("al"); break; }
328       if (ends("iviti")) { r("ive"); break; }
329       if (ends("biliti")) { r("ble"); break; }
330       break;
331     case 'g':
332       if (ends("logi")) { r("log"); break; }
333     }
334   }
335
336   /* step4() deals with -ic-, -full, -ness etc. similar strategy to step3. */
337
338   private final void step4() {
339     switch (b[k]) {
340     case 'e':
341       if (ends("icate")) { r("ic"); break; }
342       if (ends("ative")) { r(""); break; }
343       if (ends("alize")) { r("al"); break; }
344       break;
345     case 'i':
346       if (ends("iciti")) { r("ic"); break; }
347       break;
348     case 'l':
349       if (ends("ical")) { r("ic"); break; }
350       if (ends("ful")) { r(""); break; }
351       break;
352     case 's':
353       if (ends("ness")) { r(""); break; }
354       break;
355     }
356   }
357
358   /* step5() takes off -ant, -ence etc., in context <c>vcvc<v>. */
359
360   private final void step5() {
361     if (k == k0) return; /* for Bug 1 */
362     switch (b[k-1]) {
363     case 'a':
364       if (ends("al")) break;
365       return;
366     case 'c':
367       if (ends("ance")) break;
368       if (ends("ence")) break;
369       return;
370     case 'e':
371       if (ends("er")) break; return;
372     case 'i':
373       if (ends("ic")) break; return;
374     case 'l':
375       if (ends("able")) break;
376       if (ends("ible")) break; return;
377     case 'n':
378       if (ends("ant")) break;
379       if (ends("ement")) break;
380       if (ends("ment")) break;
381       /* element etc. not stripped before the m */
382       if (ends("ent")) break;
383       return;
384     case 'o':
385       if (ends("ion") && j >= 0 && (b[j] == 's' || b[j] == 't')) break;
386       /* j >= 0 fixes Bug 2 */
387       if (ends("ou")) break;
388       return;
389       /* takes care of -ous */
390     case 's':
391       if (ends("ism")) break;
392       return;
393     case 't':
394       if (ends("ate")) break;
395       if (ends("iti")) break;
396       return;
397     case 'u':
398       if (ends("ous")) break;
399       return;
400     case 'v':
401       if (ends("ive")) break;
402       return;
403     case 'z':
404       if (ends("ize")) break;
405       return;
406     default:
407       return;
408     }
409     if (m() > 1)
410       k = j;
411   }
412
413   /* step6() removes a final -e if m() > 1. */
414
415   private final void step6() {
416     j = k;
417     if (b[k] == 'e') {
418       int a = m();
419       if (a > 1 || a == 1 && !cvc(k-1))
420         k--;
421     }
422     if (b[k] == 'l' && doublec(k) && m() > 1)
423       k--;
424   }
425
426
427   /**
428    * Stem a word provided as a String. Returns the result as a String.
429    */
430   public String JavaDoc stem(String JavaDoc s) {
431     if (stem(s.toCharArray(), s.length()))
432       return toString();
433     else
434       return s;
435   }
436
437   /** Stem a word contained in a char[]. Returns true if the stemming process
438    * resulted in a word different from the input. You can retrieve the
439    * result with getResultLength()/getResultBuffer() or toString().
440    */
441   public boolean stem(char[] word) {
442     return stem(word, word.length);
443   }
444
445   /** Stem a word contained in a portion of a char[] array. Returns
446    * true if the stemming process resulted in a word different from
447    * the input. You can retrieve the result with
448    * getResultLength()/getResultBuffer() or toString().
449    */
450   public boolean stem(char[] wordBuffer, int offset, int wordLen) {
451     reset();
452     if (b.length < wordLen) {
453       char[] new_b = new char[wordLen + EXTRA];
454       b = new_b;
455     }
456     for (int j=0; j<wordLen; j++)
457       b[j] = wordBuffer[offset+j];
458     i = wordLen;
459     return stem(0);
460   }
461
462   /** Stem a word contained in a leading portion of a char[] array.
463    * Returns true if the stemming process resulted in a word different
464    * from the input. You can retrieve the result with
465    * getResultLength()/getResultBuffer() or toString().
466    */
467   public boolean stem(char[] word, int wordLen) {
468     return stem(word, 0, wordLen);
469   }
470
471   /** Stem the word placed into the Stemmer buffer through calls to add().
472    * Returns true if the stemming process resulted in a word different
473    * from the input. You can retrieve the result with
474    * getResultLength()/getResultBuffer() or toString().
475    */
476   public boolean stem() {
477     return stem(0);
478   }
479
480   public boolean stem(int i0) {
481     k = i - 1;
482     k0 = i0;
483     if (k > k0+1) {
484       step1(); step2(); step3(); step4(); step5(); step6();
485     }
486     // Also, a word is considered dirty if we lopped off letters
487 // Thanks to Ifigenia Vairelles for pointing this out.
488 if (i != k+1)
489       dirty = true;
490     i = k+1;
491     return dirty;
492   }
493
494   /** Test program for demonstrating the Stemmer. It reads a file and
495    * stems each word, writing the result to standard out.
496    * Usage: Stemmer file-name
497    */
498   public static void main(String JavaDoc[] args) {
499     PorterStemmer s = new PorterStemmer();
500
501     for (int i = 0; i < args.length; i++) {
502       try {
503         InputStream in = new FileInputStream(args[i]);
504         byte[] buffer = new byte[1024];
505         int bufferLen, offset, ch;
506
507         bufferLen = in.read(buffer);
508         offset = 0;
509         s.reset();
510
511         while(true) {
512           if (offset < bufferLen)
513             ch = buffer[offset++];
514           else {
515             bufferLen = in.read(buffer);
516             offset = 0;
517             if (bufferLen < 0)
518               ch = -1;
519             else
520               ch = buffer[offset++];
521           }
522
523           if (Character.isLetter((char) ch)) {
524             s.add(Character.toLowerCase((char) ch));
525           }
526           else {
527              s.stem();
528              System.out.print(s.toString());
529              s.reset();
530              if (ch < 0)
531                break;
532              else {
533                System.out.print((char) ch);
534              }
535            }
536         }
537
538         in.close();
539       }
540       catch (IOException e) {
541         System.out.println("error reading " + args[i]);
542       }
543     }
544   }
545 }
546
547
Popular Tags