2 * Copyright (C) 1994-1997, Index Data I/S
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.22 1997-12-18 10:54:25 adam
8 * New method result set method rs_hits that returns the number of
9 * hits in result-set (if known). The ranked result set returns real
10 * number of hits but only when not combined with other operands.
12 * Revision 1.21 1997/11/18 10:05:08 adam
13 * Changed character map facility so that admin can specify character
14 * mapping files for each register type, w, p, etc.
16 * Revision 1.20 1997/10/31 12:37:55 adam
17 * Code calls xfree() instead of free().
19 * Revision 1.19 1997/10/01 11:44:06 adam
20 * Small improvement of new ranking.
22 * Revision 1.18 1997/09/24 13:36:41 adam
23 * More work on new ranking algorithm.
25 * Revision 1.17 1997/09/22 12:39:07 adam
26 * Added get_pos method for the ranked result sets.
28 * Revision 1.16 1997/09/17 12:19:23 adam
29 * Zebra version corresponds to YAZ version 1.4.
30 * Changed Zebra server so that it doesn't depend on global common_resource.
32 * Revision 1.15 1997/09/09 13:38:16 adam
33 * Partial port to WIN95/NT.
35 * Revision 1.14 1996/11/08 11:15:58 adam
36 * Compressed isam fully supported.
38 * Revision 1.13 1996/10/29 13:55:26 adam
39 * Include of zebrautl.h instead of alexutil.h.
41 * Revision 1.12 1996/10/08 13:00:40 adam
42 * Bug fix: result sets with ranked operands in boolean operations weren't
45 * Revision 1.11 1996/10/07 16:05:29 quinn
48 * Revision 1.9 1995/12/11 09:15:26 adam
49 * New set types: sand/sor/snot - ranked versions of and/or/not in
50 * ranked/semi-ranked result sets.
51 * Note: the snot not finished yet.
52 * New rset member: flag.
53 * Bug fix: r_delete in rsrel.c did free bad memory block.
55 * Revision 1.8 1995/12/05 11:25:45 adam
56 * Doesn't include math.h.
58 * Revision 1.7 1995/10/12 12:41:57 adam
59 * Private info (buf) moved from struct rset_control to struct rset.
60 * Bug fixes in relevance.
62 * Revision 1.6 1995/10/10 14:00:04 adam
63 * Function rset_open changed its wflag parameter to general flags.
65 * Revision 1.5 1995/10/06 14:38:06 adam
66 * New result set method: r_score.
67 * Local no (sysno) and score is transferred to retrieveCtrl.
69 * Revision 1.4 1995/09/14 07:48:56 adam
70 * Other score calculation.
72 * Revision 1.3 1995/09/11 15:23:40 adam
73 * More work on relevance search.
75 * Revision 1.2 1995/09/11 13:09:41 adam
76 * More work on relevance feedback.
78 * Revision 1.1 1995/09/08 14:52:42 adam
79 * Work on relevance feedback.
93 static void *r_create(const struct rset_control *sel, void *parms,
95 static RSFD r_open (RSET ct, int flag);
96 static void r_close (RSFD rfd);
97 static void r_delete (RSET ct);
98 static void r_rewind (RSFD rfd);
99 static int r_count (RSET ct);
100 static int r_hits (RSET ct, void *oi);
101 static int r_read (RSFD rfd, void *buf);
102 static int r_write (RSFD rfd, const void *buf);
103 static int r_score (RSFD rfd, int *score);
105 static const rset_control control =
120 const rset_control *rset_kind_relevance = &control;
122 struct rset_rel_info {
126 int hits; /* hits count */
127 int (*cmp)(const void *p1, const void *p2);
128 int (*get_pos)(const void *p);
129 char *key_buf; /* key buffer */
130 float *score_buf; /* score buffer */
131 int *sort_idx; /* score sorted index */
132 int *sysno_idx; /* sysno sorted index (ring buffer) */
133 struct rset_rel_rfd *rfd_list;
136 struct rset_rel_rfd {
140 struct rset_rel_rfd *next;
141 struct rset_rel_info *info;
144 static void add_rec (struct rset_rel_info *info, double score, void *key)
149 for (i = 0; i<info->no_rec; i++)
151 idx = info->sort_idx[i];
152 if (score <= info->score_buf[idx])
155 if (info->no_rec < info->max_rec)
156 { /* there is room for this entry */
157 for (j = info->no_rec; j > i; --j)
158 info->sort_idx[j] = info->sort_idx[j-1];
159 idx = info->sort_idx[j] = info->no_rec;
163 return; /* score too low */
166 idx = info->sort_idx[0]; /* remove this entry */
169 for (j = 0; j < i; ++j) /* make room */
170 info->sort_idx[j] = info->sort_idx[j+1];
171 info->sort_idx[j] = idx; /* allocate sort entry */
173 memcpy (info->key_buf + idx*info->key_size, key, info->key_size);
174 info->score_buf[idx] = score;
178 static struct rset_rel_info *qsort_info;
180 static int qcomp (const void *p1, const void *p2)
185 return qsort_info->cmp (qsort_info->key_buf + i1*qsort_info->key_size,
186 qsort_info->key_buf + i2*qsort_info->key_size);
189 #define NEW_RANKING 0
191 #define SCORE_SHOW 0.0 /* base score for showing up */
192 #define SCORE_COOC 0.3 /* component dependent on co-oc */
193 #define SCORE_DYN (1-(SCORE_SHOW+SCORE_COOC)) /* dynamic component of score */
195 static void relevance (struct rset_rel_info *info, rset_relevance_parms *parms)
205 char *isam_prev_buf = NULL;
208 ISPT *isam_pt = NULL;
209 ISAMC_PP *isamc_pp = NULL;
212 logf (LOG_DEBUG, "relevance");
213 isam_buf = xmalloc (parms->no_isam_positions * sizeof(*isam_buf));
214 isam_r = xmalloc (sizeof (*isam_r) * parms->no_isam_positions);
216 isam_pt = xmalloc (sizeof (*isam_pt) * parms->no_isam_positions);
218 isamc_pp = xmalloc (sizeof (*isamc_pp) * parms->no_isam_positions);
221 logf (LOG_FATAL, "No isamc or isam in rs_rel");
224 isam_tmp_buf = xmalloc (info->key_size);
225 max_tf = xmalloc (sizeof (*max_tf) * parms->no_terms);
226 tf = xmalloc (sizeof (*tf) * parms->no_terms);
228 for (i = 0; i<parms->no_terms; i++)
230 for (i = 0; i<parms->no_isam_positions; i++)
232 isam_buf[i] = xmalloc (info->key_size);
235 isam_pt[i] = is_position (parms->is, parms->isam_positions[i]);
236 max_tf [parms->term_no[i]] = is_numkeys (isam_pt[i]);
237 isam_r[i] = is_readkey (isam_pt[i], isam_buf[i]);
241 isamc_pp[i] = isc_pp_open (parms->isc, parms->isam_positions[i]);
242 max_tf [parms->term_no[i]] = isc_pp_num (isamc_pp[i]);
243 isam_r[i] = isc_pp_read (isamc_pp[i], isam_buf[i]);
245 logf (LOG_DEBUG, "max tf %d = %d", i, max_tf[i]);
247 switch (parms->method)
254 for (i = 0; i<parms->no_isam_positions; i++)
257 (r = (*parms->cmp)(isam_buf[i], isam_buf[min])) < 1))
261 pos_tf = xmalloc (sizeof(*pos_tf) * parms->no_isam_positions);
262 isam_prev_buf = xmalloc (info->key_size);
263 fact1 = 100000/parms->no_isam_positions;
265 (parms->no_isam_positions*parms->no_isam_positions);
267 no_occur = score_sum = 0;
268 memcpy (isam_prev_buf, isam_buf[min], info->key_size);
269 for (i = 0; i<parms->no_isam_positions; i++)
273 (*parms->cmp)(isam_buf[min], isam_prev_buf) > 1)
275 logf (LOG_LOG, "final occur = %d ratio=%d",
276 no_occur, score_sum / no_occur);
277 add_rec (info, score_sum / (10000.0*no_occur), isam_prev_buf);
280 no_occur = score_sum = 0;
281 memcpy (isam_prev_buf, isam_buf[min], info->key_size);
282 for (i = 0; i<parms->no_isam_positions; i++)
285 pos = (*parms->get_pos)(isam_buf[min]);
286 logf (LOG_LOG, "pos=%d", pos);
287 for (i = 0; i<parms->no_isam_positions; i++)
289 int d = pos - pos_tf[i];
292 if (pos_tf[i] < 0 && i != min)
297 score_sum += fact2 / d;
299 score_sum += fact1 / d;
302 logf (LOG_LOG, "score_sum = %d", score_sum);
305 isam_r[i] = is_readkey (isam_pt[i], isam_buf[i]);
307 isam_r[i] = isc_pp_read (isamc_pp[i], isam_buf[i]);
309 xfree (isam_prev_buf);
317 int co_oc, last_term; /* Number of co-occurrences */
320 /* find min with lowest sysno */
321 for (i = 0; i<parms->no_isam_positions; i++)
325 || (r = (*parms->cmp)(isam_buf[i], isam_buf[min])) < 2))
330 else if (!r && last_term != parms->term_no[i])
331 co_oc++; /* new occurrence */
332 last_term = parms->term_no[i];
337 memcpy (isam_tmp_buf, isam_buf[min], info->key_size);
338 /* calculate for all with those sysno */
339 for (i = 0; i < parms->no_terms; i++)
341 for (i = 0; i<parms->no_isam_positions; i++)
346 r = (*parms->cmp)(isam_buf[i], isam_tmp_buf);
349 if (r <= 1 && r >= -1)
353 tf[parms->term_no[i]]++;
355 isam_r[i] = is_readkey (isam_pt[i], isam_buf[i]);
357 isam_r[i] = isc_pp_read (isamc_pp[i], isam_buf[i]);
358 } while (isam_r[i] &&
359 (*parms->cmp)(isam_buf[i], isam_tmp_buf) <= 1);
362 /* calculate relevance value */
364 for (i = 0; i<parms->no_terms; i++)
366 score += SCORE_SHOW + SCORE_COOC*co_oc/parms->no_terms +
367 SCORE_DYN*tf[i]/max_tf[i];
368 /* if value is in the top score, then save it - don't emit yet */
369 add_rec (info, score/parms->no_terms, isam_tmp_buf);
373 for (i = 0; i<info->no_rec; i++)
374 info->sysno_idx[i] = i;
376 qsort (info->sysno_idx, info->no_rec, sizeof(*info->sysno_idx), qcomp);
377 for (i = 0; i<parms->no_isam_positions; i++)
380 is_pt_free (isam_pt[i]);
382 isc_pp_close (isamc_pp[i]);
386 xfree (isam_tmp_buf);
394 static void *r_create (const struct rset_control *sel, void *parms,
397 rset_relevance_parms *r_parms = parms;
398 struct rset_rel_info *info;
400 *flags |= RSET_FLAG_RANKED;
401 info = xmalloc (sizeof(struct rset_rel_info));
402 info->key_size = r_parms->key_size;
403 assert (info->key_size > 1);
404 info->max_rec = r_parms->max_rec;
405 assert (info->max_rec > 1);
406 info->cmp = r_parms->cmp;
407 info->get_pos = r_parms->get_pos;
409 info->key_buf = xmalloc (info->key_size * info->max_rec);
410 info->score_buf = xmalloc (sizeof(*info->score_buf) * info->max_rec);
411 info->sort_idx = xmalloc (sizeof(*info->sort_idx) * info->max_rec);
412 info->sysno_idx = xmalloc (sizeof(*info->sysno_idx) * info->max_rec);
415 info->rfd_list = NULL;
417 relevance (info, r_parms);
421 static RSFD r_open (RSET ct, int flag)
423 struct rset_rel_rfd *rfd;
424 struct rset_rel_info *info = ct->buf;
426 if (flag & RSETF_WRITE)
428 logf (LOG_FATAL, "relevance set type is read-only");
431 rfd = xmalloc (sizeof(*rfd));
433 rfd->next = info->rfd_list;
435 info->rfd_list = rfd;
440 static void r_close (RSFD rfd)
442 struct rset_rel_info *info = ((struct rset_rel_rfd*)rfd)->info;
443 struct rset_rel_rfd **rfdp;
445 for (rfdp = &info->rfd_list; *rfdp; rfdp = &(*rfdp)->next)
448 *rfdp = (*rfdp)->next;
452 logf (LOG_FATAL, "r_close but no rfd match!");
456 static void r_delete (RSET ct)
458 struct rset_rel_info *info = ct->buf;
460 assert (info->rfd_list == NULL);
461 xfree (info->key_buf);
462 xfree (info->score_buf);
463 xfree (info->sort_idx);
464 xfree (info->sysno_idx);
468 static void r_rewind (RSFD rfd)
470 struct rset_rel_rfd *p = rfd;
471 struct rset_rel_info *info = p->info;
473 if (p->flag & RSETF_SORT_RANK)
474 p->position = info->no_rec;
479 static int r_count (RSET ct)
481 struct rset_rel_info *info = ct->buf;
486 static int r_hits (RSET ct, void *oi)
488 struct rset_rel_info *info = ct->buf;
493 static int r_read (RSFD rfd, void *buf)
495 struct rset_rel_rfd *p = rfd;
496 struct rset_rel_info *info = p->info;
498 if (p->flag & RSETF_SORT_RANK)
500 if (p->position <= 0)
503 p->last_read_pos = info->sort_idx[p->position];
507 if (p->position == info->no_rec)
509 p->last_read_pos = info->sysno_idx[p->position];
513 info->key_buf + info->key_size * p->last_read_pos,
518 static int r_score (RSFD rfd, int *score)
520 struct rset_rel_rfd *p = rfd;
521 struct rset_rel_info *info = p->info;
523 *score = (int) (1000*info->score_buf[p->last_read_pos]);
527 static int r_write (RSFD rfd, const void *buf)
529 logf (LOG_FATAL, "relevance set type is read-only");