src/relevance.c

   1 /* This file is part of Pazpar2.
   2    Copyright (C) 2006-2013 Index Data
   3
   4 Pazpar2 is free software; you can redistribute it and/or modify it under
   5 the terms of the GNU General Public License as published by the Free
   6 Software Foundation; either version 2, or (at your option) any later
   7 version.
   8
   9 Pazpar2 is distributed in the hope that it will be useful, but WITHOUT ANY
  10 WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11 FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12 for more details.
  13
  14 You should have received a copy of the GNU General Public License
  15 along with this program; if not, write to the Free Software
  16 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  17
  18 */
  19
  20 #if HAVE_CONFIG_H
  21 #include <config.h>
  22 #endif
  23
  24 #include <assert.h>
  25 #include <math.h>
  26 #include <stdlib.h>
  27
  28 #include "relevance.h"
  29 #include "session.h"
  30 #include "client.h"
  31 #include "settings.h"
  32
  33 #ifdef WIN32
  34 #define log2(x) (log(x)/log(2))
  35 #endif
  36
  37 struct relevance
  38 {
  39     int *doc_frequency_vec;
  40     int *term_frequency_vec_tmp;
  41     int *term_pos;
  42     int vec_len;
  43     struct word_entry *entries;
  44     pp2_charset_token_t prt;
  45     int rank_cluster;
  46     double follow_factor;
  47     double lead_decay;
  48     int length_divide;
  49     NMEM nmem;
  50     struct norm_client *norm;
  51 };
  52
  53 struct word_entry {
  54     const char *norm_str;
  55     const char *display_str;
  56     int termno;
  57     char *ccl_field;
  58     struct word_entry *next;
  59 };
  60
  61 // Structure to keep data for norm_client scores from one client
  62 struct norm_client
  63 {
  64     int num; // number of the client
  65     float max;
  66     float min;
  67     int count;
  68     const char *native_score;
  69     int scorefield;
  70     float a,b; // Rn = a*R + b
  71     struct client *client;
  72     struct norm_client *next;
  73     struct norm_record *records;
  74 };
  75
  76 const int scorefield_none = -1;  // Do not normalize anything, use tf/idf as is
  77   // This is the old behavior, and the default
  78 const int scorefield_internal = -2;  // use our tf/idf, but normalize it
  79 const int scorefield_position = -3;  // fake a score based on the position
  80
  81 // A structure for each (sub)record. There is one list for each client
  82 struct norm_record
  83 {
  84     struct record *record;
  85     float score;
  86     struct record_cluster *clust;
  87     struct norm_record *next;
  88 };
  89
  90 // Find the norm_client entry for this client, or create one if not there
  91 struct norm_client *findnorm( struct relevance *rel, struct client* client)
  92 {
  93     struct norm_client *n = rel->norm;
  94     struct session_database *sdb;
  95     while (n) {
  96         if (n->client == client )
  97             return n;
  98         n = n->next;
  99     }
 100     n = nmem_malloc(rel->nmem, sizeof(struct norm_client) );
 101     if ( rel->norm )
 102         n->num = rel->norm->num +1;
 103     else
 104         n->num = 1;
 105     n->count = 0;
 106     n->max = 0.0;
 107     n->min = 0.0;
 108     n->client = client;
 109     n->next = rel->norm;
 110     rel->norm = n;
 111     sdb = client_get_database(client);
 112     n->native_score = session_setting_oneval(sdb, PZ_NATIVE_SCORE);
 113     n->records = 0;
 114     n->scorefield = scorefield_none;
 115     yaz_log(YLOG_LOG,"Normalizing: Client %d uses '%s'", n->num, n->native_score );
 116     if ( ! n->native_score  || ! *n->native_score )  // not specified
 117         n->scorefield = scorefield_none;
 118     else if ( strcmp(n->native_score,"position") == 0 )
 119         n->scorefield = scorefield_position;
 120     else if ( strcmp(n->native_score,"internal") == 0 )
 121         n->scorefield = scorefield_internal;
 122     else
 123     { // Get the field index for the score
 124         struct session *se = client_get_session(client);
 125         n->scorefield = conf_service_metadata_field_id(se->service, n->native_score);
 126     }
 127     yaz_log(YLOG_LOG,"Normalizing: Client %d uses '%s' = %d",
 128             n->num, n->native_score, n->scorefield );
 129     return n;
 130 }
 131
 132
 133 // Add a record in the list for that client, for normalizing later
 134 static void setup_norm_record( struct relevance *rel,  struct record_cluster *clust)
 135 {
 136     struct record *record;
 137     for (record = clust->records; record; record = record->next)
 138     {
 139         struct norm_client *norm = findnorm(rel, record->client);
 140         struct norm_record *rp;
 141         if ( norm->scorefield == scorefield_none)
 142             break;  // not interested in normalizing this client
 143         rp = nmem_malloc(rel->nmem, sizeof(struct norm_record) );
 144         norm->count ++;
 145         rp->next = norm->records;
 146         norm->records = rp;
 147         rp->clust = clust;
 148         rp->record = record;
 149         if ( norm->scorefield == scorefield_position )
 150             rp->score = 1.0 / record->position;
 151         else if ( norm->scorefield == scorefield_internal )
 152             rp->score = clust->relevance_score; // the tf/idf for the whole cluster
 153               // TODO - Get them for each record, merge later!
 154         else
 155         {
 156             struct record_metadata *md = record->metadata[norm->scorefield];
 157             rp->score = md->data.fnumber;
 158         }
 159         yaz_log(YLOG_LOG,"Got score for %d/%d : %f ",
 160                 norm->num, record->position, rp->score );
 161         if ( norm->count == 1 )
 162         {
 163             norm->max = rp->score;
 164             norm->min = rp->score;
 165         } else {
 166             if ( rp->score > norm->max )
 167                 norm->max = rp->score;
 168             if ( rp->score < norm->min && abs(rp->score) < 1e-6 )
 169                 norm->min = rp->score;  // skip zeroes
 170         }
 171     }
 172 }
 173
 174 // Calculate the squared sum of residuals, that is the difference from
 175 // normalized values to the target curve, which is 1/n
 176 static double squaresum( struct norm_record *rp, double a, double b)
 177 {
 178     double sum = 0.0;
 179     for ( ; rp; rp = rp->next )
 180     {
 181         double target = 1.0 / rp->record->position;
 182         double normscore = rp->score * a + b;
 183         double diff = target - normscore;
 184         sum += diff * diff;
 185     }
 186     return sum;
 187 }
 188
 189 // For each client, normalize scores
 190 static void normalize_scores(struct relevance *rel)
 191 {
 192     const int maxiterations = 100;
 193     const double enough = 100.0;  // sets the number of decimals we are happy with
 194     struct norm_client *norm;
 195     for ( norm = rel->norm; norm; norm = norm->next )
 196     {
 197         yaz_log(YLOG_LOG,"Normalizing client %d: scorefield=%d count=%d",
 198                 norm->num, norm->scorefield, norm->count);
 199         norm->a = 1.0; // default normalizing factors, no change
 200         norm->b = 0.0;
 201         if ( norm->scorefield != scorefield_none &&
 202              norm->scorefield != scorefield_position )
 203         { // have something to normalize
 204             double range = norm->max - norm->min;
 205             int it = 0;
 206             double a,b;  // params to optimize
 207             double as,bs; // step sizes
 208             double chi;
 209             char dir = 'a';
 210             // initial guesses for the parameters
 211             if ( range < 1e-6 ) // practically zero
 212                 range = norm->max;
 213             a = 1.0 / range;
 214             b = abs(norm->min);
 215             as = a / 10;
 216             bs = b / 10;
 217             chi = squaresum( norm->records, a,b);
 218             while (it++ < maxiterations)  // safeguard against things not converging
 219             {
 220                 double aplus = squaresum(norm->records, a+as, b);
 221                 double aminus= squaresum(norm->records, a-as, b);
 222                 double bplus = squaresum(norm->records, a, b+bs);
 223                 double bminus= squaresum(norm->records, a, b-bs);
 224                 if ( aplus < chi && aplus < aminus && aplus < bplus && aplus < bminus)
 225                 {
 226                     a = a + as;
 227                     chi = aplus;
 228                     yaz_log(YLOG_LOG,"Fitting aplus it=%d: a=%f / %f  b=%f / %f  chi = %f",
 229                         it, a, as, b, bs, chi );
 230                 }
 231                 else if ( aminus < chi && aminus < aplus && aminus < bplus && aminus < bminus)
 232                 {
 233                     a = a - as;
 234                     chi = aminus;
 235                     yaz_log(YLOG_LOG,"Fitting aminus it=%d: a=%f / %f  b=%f / %f  chi = %f",
 236                         it, a, as, b, bs, chi );
 237                 }
 238                 else if ( bplus < chi && bplus < aplus && bplus < aminus && bplus < bminus)
 239                 {
 240                     b = b + bs;
 241                     chi = bplus;
 242                     yaz_log(YLOG_LOG,"Fitting bplus it=%d: a=%f / %f  b=%f / %f  chi = %f",
 243                         it, a, as, b, bs, chi );
 244                 }
 245                 else if ( bminus < chi && bminus < aplus && bminus < bplus && bminus < aminus)
 246                 {
 247                     b = b - bs;
 248                     chi = bminus;
 249                     yaz_log(YLOG_LOG,"Fitting bminus it=%d: a=%f / %f  b=%f / %f  chi = %f",
 250                         it, a, as, b, bs, chi );
 251                 }
 252                 else
 253                 {
 254                     if ( as > bs )
 255                     {
 256                         as = as / 2;
 257                         yaz_log(YLOG_LOG,"Fitting step a it=%d: a=%f / %f  b=%f / %f  chi = %f",
 258                             it, a, as, b, bs, chi );
 259                     }
 260                     else
 261                     {
 262                         bs = bs / 2;
 263                         yaz_log(YLOG_LOG,"Fitting step b it=%d: a=%f / %f  b=%f / %f  chi = %f",
 264                             it, a, as, b, bs, chi );
 265                     }
 266                 }
 267                 norm->a = a;
 268                 norm->b = b;
 269                 if ( fabs(as) * enough < fabs(a) &&
 270                      fabs(bs) * enough < fabs(b) ) {
 271                     yaz_log(YLOG_LOG,"Fitting done: stopping loop at %d" , it );
 272                     break;  // not changing much any more
 273
 274                 }
 275             }
 276             yaz_log(YLOG_LOG,"Fitting done: it=%d: a=%f / %f  b=%f / %f  chi = %f",
 277                         it-1, a, as, b, bs, chi );
 278             yaz_log(YLOG_LOG,"  a: %f < %f %d",
 279                     fabs(as)*enough, fabs(a), (fabs(as) * enough < fabs(a)) );
 280             yaz_log(YLOG_LOG,"  b: %f < %f %d",
 281                     fabs(bs)*enough, fabs(b), (fabs(bs) * enough < fabs(b)) );
 282         }
 283
 284         if ( norm->scorefield != scorefield_none )
 285         { // distribute the normalized scores to the records
 286             struct norm_record *nr = norm->records;
 287             for ( ; nr ; nr = nr->next ) {
 288                 double r = nr->score;
 289                 r = norm->a * r + norm -> b;
 290                 nr->clust->relevance_score = 10000 * r;
 291                 yaz_log(YLOG_LOG,"Normalized %f * %f + %f = %f",
 292                         nr->score, norm->a, norm->b, r );
 293                 // TODO - This keeps overwriting the cluster score in random order!
 294                 // Need to merge results better
 295             }
 296
 297         }
 298
 299     } // client loop
 300 }
 301
 302
 303 static struct word_entry *word_entry_match(struct relevance *r,
 304                                            const char *norm_str,
 305                                            const char *rank, int *weight)
 306 {
 307     int i = 1;
 308     struct word_entry *entries = r->entries;
 309     for (; entries; entries = entries->next, i++)
 310     {
 311         if (*norm_str && !strcmp(norm_str, entries->norm_str))
 312         {
 313             const char *cp = 0;
 314             int no_read = 0;
 315             sscanf(rank, "%d%n", weight, &no_read);
 316             rank += no_read;
 317             while (*rank == ' ')
 318                 rank++;
 319             if (no_read > 0 && (cp = strchr(rank, ' ')))
 320             {
 321                 if ((cp - rank) == strlen(entries->ccl_field) &&
 322                     memcmp(entries->ccl_field, rank, cp - rank) == 0)
 323                     *weight = atoi(cp + 1);
 324             }
 325             return entries;
 326         }
 327     }
 328     return 0;
 329 }
 330
 331 int relevance_snippet(struct relevance *r,
 332                       const char *words, const char *name,
 333                       WRBUF w_snippet)
 334 {
 335     int no = 0;
 336     const char *norm_str;
 337     int highlight = 0;
 338
 339     pp2_charset_token_first(r->prt, words, 0);
 340     while ((norm_str = pp2_charset_token_next(r->prt)))
 341     {
 342         size_t org_start, org_len;
 343         struct word_entry *entries = r->entries;
 344         int i;
 345
 346         pp2_get_org(r->prt, &org_start, &org_len);
 347         for (; entries; entries = entries->next, i++)
 348         {
 349             if (*norm_str && !strcmp(norm_str, entries->norm_str))
 350                 break;
 351         }
 352         if (entries)
 353         {
 354             if (!highlight)
 355             {
 356                 highlight = 1;
 357                 wrbuf_puts(w_snippet, "<match>");
 358                 no++;
 359             }
 360         }
 361         else
 362         {
 363             if (highlight)
 364             {
 365                 highlight = 0;
 366                 wrbuf_puts(w_snippet, "</match>");
 367             }
 368         }
 369         wrbuf_xmlputs_n(w_snippet, words + org_start, org_len);
 370     }
 371     if (highlight)
 372         wrbuf_puts(w_snippet, "</match>");
 373     if (no)
 374     {
 375         yaz_log(YLOG_DEBUG, "SNIPPET match: %s", wrbuf_cstr(w_snippet));
 376     }
 377     return no;
 378 }
 379
 380 void relevance_countwords(struct relevance *r, struct record_cluster *cluster,
 381                           const char *words, const char *rank,
 382                           const char *name)
 383 {
 384     int *w = r->term_frequency_vec_tmp;
 385     const char *norm_str;
 386     int i, length = 0;
 387     double lead_decay = r->lead_decay;
 388     struct word_entry *e;
 389     WRBUF wr = cluster->relevance_explain1;
 390     int printed_about_field = 0;
 391
 392     pp2_charset_token_first(r->prt, words, 0);
 393     for (e = r->entries, i = 1; i < r->vec_len; i++, e = e->next)
 394     {
 395         w[i] = 0;
 396         r->term_pos[i] = 0;
 397     }
 398
 399     assert(rank);
 400     while ((norm_str = pp2_charset_token_next(r->prt)))
 401     {
 402         int local_weight = 0;
 403         e = word_entry_match(r, norm_str, rank, &local_weight);
 404         if (e)
 405         {
 406             int res = e->termno;
 407             int j;
 408
 409             if (!printed_about_field)
 410             {
 411                 printed_about_field = 1;
 412                 wrbuf_printf(wr, "field=%s content=", name);
 413                 if (strlen(words) > 50)
 414                 {
 415                     wrbuf_xmlputs_n(wr, words, 49);
 416                     wrbuf_puts(wr, " ...");
 417                 }
 418                 else
 419                     wrbuf_xmlputs(wr, words);
 420                 wrbuf_puts(wr, ";\n");
 421             }
 422             assert(res < r->vec_len);
 423             w[res] += local_weight / (1 + log2(1 + lead_decay * length));
 424             wrbuf_printf(wr, "%s: w[%d] += w(%d) / "
 425                          "(1+log2(1+lead_decay(%f) * length(%d)));\n",
 426                          e->display_str, res, local_weight, lead_decay, length);
 427             j = res - 1;
 428             if (j > 0 && r->term_pos[j])
 429             {
 430                 int d = length + 1 - r->term_pos[j];
 431                 wrbuf_printf(wr, "%s: w[%d] += w[%d](%d) * follow(%f) / "
 432                              "(1+log2(d(%d));\n",
 433                              e->display_str, res, res, w[res],
 434                              r->follow_factor, d);
 435                 w[res] += w[res] * r->follow_factor / (1 + log2(d));
 436             }
 437             for (j = 0; j < r->vec_len; j++)
 438                 r->term_pos[j] = j < res ? 0 : length + 1;
 439         }
 440         length++;
 441     }
 442
 443     for (e = r->entries, i = 1; i < r->vec_len; i++, e = e->next)
 444     {
 445         if (length == 0 || w[i] == 0)
 446             continue;
 447         wrbuf_printf(wr, "%s: tf[%d] += w[%d](%d)", e->display_str, i, i, w[i]);
 448         switch (r->length_divide)
 449         {
 450         case 0:
 451             cluster->term_frequency_vecf[i] += (double) w[i];
 452             break;
 453         case 1:
 454             wrbuf_printf(wr, " / log2(1+length(%d))", length);
 455             cluster->term_frequency_vecf[i] +=
 456                 (double) w[i] / log2(1 + length);
 457             break;
 458         case 2:
 459             wrbuf_printf(wr, " / length(%d)", length);
 460             cluster->term_frequency_vecf[i] += (double) w[i] / length;
 461         }
 462         cluster->term_frequency_vec[i] += w[i];
 463         wrbuf_printf(wr, " (%f);\n", cluster->term_frequency_vecf[i]);
 464     }
 465
 466     cluster->term_frequency_vec[0] += length;
 467 }
 468
 469 static void pull_terms(struct relevance *res, struct ccl_rpn_node *n)
 470 {
 471     char **words;
 472     int numwords;
 473     char *ccl_field;
 474     int i;
 475
 476     switch (n->kind)
 477     {
 478     case CCL_RPN_AND:
 479     case CCL_RPN_OR:
 480     case CCL_RPN_NOT:
 481     case CCL_RPN_PROX:
 482         pull_terms(res, n->u.p[0]);
 483         pull_terms(res, n->u.p[1]);
 484         break;
 485     case CCL_RPN_TERM:
 486         nmem_strsplit(res->nmem, " ", n->u.t.term, &words, &numwords);
 487         for (i = 0; i < numwords; i++)
 488         {
 489             const char *norm_str;
 490
 491             ccl_field = nmem_strdup_null(res->nmem, n->u.t.qual);
 492
 493             pp2_charset_token_first(res->prt, words[i], 0);
 494             while ((norm_str = pp2_charset_token_next(res->prt)))
 495             {
 496                 struct word_entry **e = &res->entries;
 497                 while (*e)
 498                     e = &(*e)->next;
 499                 *e = nmem_malloc(res->nmem, sizeof(**e));
 500                 (*e)->norm_str = nmem_strdup(res->nmem, norm_str);
 501                 (*e)->ccl_field = ccl_field;
 502                 (*e)->termno = res->vec_len++;
 503                 (*e)->display_str = nmem_strdup(res->nmem, words[i]);
 504                 (*e)->next = 0;
 505             }
 506         }
 507         break;
 508     default:
 509         break;
 510     }
 511 }
 512 void relevance_clear(struct relevance *r)
 513 {
 514     if (r)
 515     {
 516         int i;
 517         for (i = 0; i < r->vec_len; i++)
 518             r->doc_frequency_vec[i] = 0;
 519     }
 520 }
 521
 522 struct relevance *relevance_create_ccl(pp2_charset_fact_t pft,
 523                                        struct ccl_rpn_node *query,
 524                                        int rank_cluster,
 525                                        double follow_factor, double lead_decay,
 526                                        int length_divide)
 527 {
 528     NMEM nmem = nmem_create();
 529     struct relevance *res = nmem_malloc(nmem, sizeof(*res));
 530
 531     res->nmem = nmem;
 532     res->entries = 0;
 533     res->vec_len = 1;
 534     res->rank_cluster = rank_cluster;
 535     res->follow_factor = follow_factor;
 536     res->lead_decay = lead_decay;
 537     res->length_divide = length_divide;
 538     res->norm = 0;
 539     res->prt = pp2_charset_token_create(pft, "relevance");
 540
 541     pull_terms(res, query);
 542
 543     res->doc_frequency_vec = nmem_malloc(nmem, res->vec_len * sizeof(int));
 544
 545     // worker array
 546     res->term_frequency_vec_tmp =
 547         nmem_malloc(res->nmem,
 548                     res->vec_len * sizeof(*res->term_frequency_vec_tmp));
 549
 550     res->term_pos =
 551         nmem_malloc(res->nmem, res->vec_len * sizeof(*res->term_pos));
 552
 553     relevance_clear(res);
 554     return res;
 555 }
 556
 557 void relevance_destroy(struct relevance **rp)
 558 {
 559     if (*rp)
 560     {
 561         pp2_charset_token_destroy((*rp)->prt);
 562         nmem_destroy((*rp)->nmem);
 563         *rp = 0;
 564     }
 565 }
 566
 567 void relevance_mergerec(struct relevance *r, struct record_cluster *dst,
 568                         const struct record_cluster *src)
 569 {
 570     int i;
 571
 572     for (i = 0; i < r->vec_len; i++)
 573         dst->term_frequency_vec[i] += src->term_frequency_vec[i];
 574
 575     for (i = 0; i < r->vec_len; i++)
 576         dst->term_frequency_vecf[i] += src->term_frequency_vecf[i];
 577 }
 578
 579 void relevance_newrec(struct relevance *r, struct record_cluster *rec)
 580 {
 581     int i;
 582
 583     // term frequency [1,..] . [0] is total length of all fields
 584     rec->term_frequency_vec =
 585         nmem_malloc(r->nmem,
 586                     r->vec_len * sizeof(*rec->term_frequency_vec));
 587     for (i = 0; i < r->vec_len; i++)
 588         rec->term_frequency_vec[i] = 0;
 589
 590     // term frequency divided by length of field [1,...]
 591     rec->term_frequency_vecf =
 592         nmem_malloc(r->nmem,
 593                     r->vec_len * sizeof(*rec->term_frequency_vecf));
 594     for (i = 0; i < r->vec_len; i++)
 595         rec->term_frequency_vecf[i] = 0.0;
 596 }
 597
 598 void relevance_donerecord(struct relevance *r, struct record_cluster *cluster)
 599 {
 600     int i;
 601
 602     for (i = 1; i < r->vec_len; i++)
 603         if (cluster->term_frequency_vec[i] > 0)
 604             r->doc_frequency_vec[i]++;
 605
 606     r->doc_frequency_vec[0]++;
 607 }
 608
 609
 610
 611 // Prepare for a relevance-sorted read
 612 void relevance_prepare_read(struct relevance *rel, struct reclist *reclist)
 613 {
 614     int i;
 615     float *idfvec = xmalloc(rel->vec_len * sizeof(float));
 616
 617     reclist_enter(reclist);
 618
 619     // Calculate document frequency vector for each term.
 620     for (i = 1; i < rel->vec_len; i++)
 621     {
 622         if (!rel->doc_frequency_vec[i])
 623             idfvec[i] = 0;
 624         else
 625         {
 626             /* add one to nominator idf(t,D) to ensure a value > 0 */
 627             idfvec[i] = log((float) (1 + rel->doc_frequency_vec[0]) /
 628                             rel->doc_frequency_vec[i]);
 629         }
 630     }
 631     // Calculate relevance for each document
 632     while (1)
 633     {
 634         int relevance = 0;
 635         WRBUF w;
 636         struct word_entry *e = rel->entries;
 637         struct record_cluster *rec = reclist_read_record(reclist);
 638         if (!rec)
 639             break;
 640         w = rec->relevance_explain2;
 641         wrbuf_rewind(w);
 642         wrbuf_puts(w, "relevance = 0;\n");
 643         for (i = 1; i < rel->vec_len; i++)
 644         {
 645             float termfreq = (float) rec->term_frequency_vecf[i];
 646             int add = 100000 * termfreq * idfvec[i];
 647
 648             wrbuf_printf(w, "idf[%d] = log(((1 + total(%d))/termoccur(%d));\n",
 649                          i, rel->doc_frequency_vec[0],
 650                          rel->doc_frequency_vec[i]);
 651             wrbuf_printf(w, "%s: relevance += 100000 * tf[%d](%f) * "
 652                          "idf[%d](%f) (%d);\n",
 653                          e->display_str, i, termfreq, i, idfvec[i], add);
 654             relevance += add;
 655             e = e->next;
 656         }
 657         if (!rel->rank_cluster)
 658         {
 659             struct record *record;
 660             int cluster_size = 0;
 661
 662             for (record = rec->records; record; record = record->next)
 663                 cluster_size++;
 664
 665             wrbuf_printf(w, "score = relevance(%d)/cluster_size(%d);\n",
 666                          relevance, cluster_size);
 667             relevance /= cluster_size;
 668         }
 669         else
 670         {
 671             wrbuf_printf(w, "score = relevance(%d);\n", relevance);
 672         }
 673         rec->relevance_score = relevance;
 674
 675         // Build the normalizing structures
 676         // List of (sub)records for each target
 677         setup_norm_record( rel, rec );
 678
 679         // TODO - Loop again, merge individual record scores into clusters
 680         // Can I reset the reclist, or can I leave and enter without race conditions?
 681
 682     } // cluster loop
 683
 684     normalize_scores(rel);
 685
 686     reclist_leave(reclist);
 687     xfree(idfvec);
 688
 689 }
 690
 691 /*
 692  * Local variables:
 693  * c-basic-offset: 4
 694  * c-file-style: "Stroustrup"
 695  * indent-tabs-mode: nil
 696  * End:
 697  * vim: shiftwidth=4 tabstop=8 expandtab
 698  */
 699