1 /* This file is part of Pazpar2.
2 Copyright (C) 2006-2012 Index Data
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
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
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
28 #include "relevance.h"
33 int *doc_frequency_vec;
34 int *term_frequency_vec_tmp;
37 struct word_entry *entries;
38 pp2_charset_token_t prt;
48 const char *display_str;
51 struct word_entry *next;
54 static struct word_entry *word_entry_match(struct relevance *r,
56 const char *rank, int *weight)
59 struct word_entry *entries = r->entries;
60 for (; entries; entries = entries->next, i++)
62 if (*norm_str && !strcmp(norm_str, entries->norm_str))
66 sscanf(rank, "%d%n", weight, &no_read);
70 if (no_read > 0 && (cp = strchr(rank, ' ')))
72 if ((cp - rank) == strlen(entries->ccl_field) &&
73 memcmp(entries->ccl_field, rank, cp - rank) == 0)
74 *weight = atoi(cp + 1);
82 void relevance_countwords(struct relevance *r, struct record_cluster *cluster,
83 const char *words, const char *rank,
86 int *w = r->term_frequency_vec_tmp;
89 double lead_decay = r->lead_decay;
91 WRBUF wr = cluster->relevance_explain1;
92 int printed_about_field = 0;
94 pp2_charset_token_first(r->prt, words, 0);
95 for (e = r->entries, i = 1; i < r->vec_len; i++, e = e->next)
102 while ((norm_str = pp2_charset_token_next(r->prt)))
104 int local_weight = 0;
105 e = word_entry_match(r, norm_str, rank, &local_weight);
111 if (!printed_about_field)
113 printed_about_field = 1;
114 wrbuf_printf(wr, "field=%s content=", name);
115 if (strlen(words) > 50)
117 wrbuf_xmlputs_n(wr, words, 49);
118 wrbuf_puts(wr, " ...");
121 wrbuf_xmlputs(wr, words);
122 wrbuf_puts(wr, ";\n");
124 assert(res < r->vec_len);
125 w[res] += local_weight / (1 + log2(1 + lead_decay * length));
126 wrbuf_printf(wr, "%s: w[%d] += w(%d) / "
127 "(1+log2(1+lead_decay(%f) * length(%d)));\n",
128 e->display_str, res, local_weight, lead_decay, length);
130 if (j > 0 && r->term_pos[j])
132 int d = length + 1 - r->term_pos[j];
133 wrbuf_printf(wr, "%s: w[%d] += w[%d](%d) * follow(%f) / "
135 e->display_str, res, res, w[res],
136 r->follow_factor, d);
137 w[res] += w[res] * r->follow_factor / (1 + log2(d));
139 for (j = 0; j < r->vec_len; j++)
140 r->term_pos[j] = j < res ? 0 : length + 1;
145 for (e = r->entries, i = 1; i < r->vec_len; i++, e = e->next)
147 if (length == 0 || w[i] == 0)
149 wrbuf_printf(wr, "%s: tf[%d] += w[%d](%d)", e->display_str, i, i, w[i]);
150 switch (r->length_divide)
153 cluster->term_frequency_vecf[i] += (double) w[i];
156 wrbuf_printf(wr, " / log2(1+length(%d))", length);
157 cluster->term_frequency_vecf[i] +=
158 (double) w[i] / log2(1 + length);
161 wrbuf_printf(wr, " / length(%d)", length);
162 cluster->term_frequency_vecf[i] += (double) w[i] / length;
164 cluster->term_frequency_vec[i] += w[i];
165 wrbuf_printf(wr, " (%f);\n", cluster->term_frequency_vecf[i]);
168 cluster->term_frequency_vec[0] += length;
171 static void pull_terms(struct relevance *res, struct ccl_rpn_node *n)
184 pull_terms(res, n->u.p[0]);
185 pull_terms(res, n->u.p[1]);
188 nmem_strsplit(res->nmem, " ", n->u.t.term, &words, &numwords);
189 for (i = 0; i < numwords; i++)
191 const char *norm_str;
193 ccl_field = nmem_strdup_null(res->nmem, n->u.t.qual);
195 pp2_charset_token_first(res->prt, words[i], 0);
196 while ((norm_str = pp2_charset_token_next(res->prt)))
198 struct word_entry **e = &res->entries;
201 *e = nmem_malloc(res->nmem, sizeof(**e));
202 (*e)->norm_str = nmem_strdup(res->nmem, norm_str);
203 (*e)->ccl_field = ccl_field;
204 (*e)->termno = res->vec_len++;
205 (*e)->display_str = nmem_strdup(res->nmem, words[i]);
215 struct relevance *relevance_create_ccl(pp2_charset_fact_t pft,
216 struct ccl_rpn_node *query,
218 double follow_factor, double lead_decay,
221 NMEM nmem = nmem_create();
222 struct relevance *res = nmem_malloc(nmem, sizeof(*res));
228 res->rank_cluster = rank_cluster;
229 res->follow_factor = follow_factor;
230 res->lead_decay = lead_decay;
231 res->length_divide = length_divide;
232 res->prt = pp2_charset_token_create(pft, "relevance");
234 pull_terms(res, query);
236 res->doc_frequency_vec = nmem_malloc(nmem, res->vec_len * sizeof(int));
237 for (i = 0; i < res->vec_len; i++)
238 res->doc_frequency_vec[i] = 0;
241 res->term_frequency_vec_tmp =
242 nmem_malloc(res->nmem,
243 res->vec_len * sizeof(*res->term_frequency_vec_tmp));
246 nmem_malloc(res->nmem, res->vec_len * sizeof(*res->term_pos));
251 void relevance_destroy(struct relevance **rp)
255 pp2_charset_token_destroy((*rp)->prt);
256 nmem_destroy((*rp)->nmem);
261 void relevance_newrec(struct relevance *r, struct record_cluster *rec)
263 if (!rec->term_frequency_vec)
267 // term frequency [1,..] . [0] is total length of all fields
268 rec->term_frequency_vec =
270 r->vec_len * sizeof(*rec->term_frequency_vec));
271 for (i = 0; i < r->vec_len; i++)
272 rec->term_frequency_vec[i] = 0;
274 // term frequency divided by length of field [1,...]
275 rec->term_frequency_vecf =
277 r->vec_len * sizeof(*rec->term_frequency_vecf));
278 for (i = 0; i < r->vec_len; i++)
279 rec->term_frequency_vecf[i] = 0.0;
283 void relevance_donerecord(struct relevance *r, struct record_cluster *cluster)
287 for (i = 1; i < r->vec_len; i++)
288 if (cluster->term_frequency_vec[i] > 0)
289 r->doc_frequency_vec[i]++;
291 r->doc_frequency_vec[0]++;
294 // Prepare for a relevance-sorted read
295 void relevance_prepare_read(struct relevance *rel, struct reclist *reclist)
298 float *idfvec = xmalloc(rel->vec_len * sizeof(float));
300 reclist_enter(reclist);
301 // Calculate document frequency vector for each term.
302 for (i = 1; i < rel->vec_len; i++)
304 if (!rel->doc_frequency_vec[i])
308 /* add one to nominator idf(t,D) to ensure a value > 0 */
309 idfvec[i] = log((float) (1 + rel->doc_frequency_vec[0]) /
310 rel->doc_frequency_vec[i]);
313 // Calculate relevance for each document
318 struct word_entry *e = rel->entries;
319 struct record_cluster *rec = reclist_read_record(reclist);
322 w = rec->relevance_explain2;
324 wrbuf_puts(w, "relevance = 0;\n");
325 for (i = 1; i < rel->vec_len; i++)
327 float termfreq = (float) rec->term_frequency_vecf[i];
328 int add = 100000 * termfreq * idfvec[i];
330 wrbuf_printf(w, "idf[%d] = log(((1 + total(%d))/termoccur(%d));\n",
331 i, rel->doc_frequency_vec[0],
332 rel->doc_frequency_vec[i]);
333 wrbuf_printf(w, "%s: relevance += 100000 * tf[%d](%f) * "
334 "idf[%d](%f) (%d);\n",
335 e->display_str, i, termfreq, i, idfvec[i], add);
339 if (!rel->rank_cluster)
341 struct record *record;
342 int cluster_size = 0;
344 for (record = rec->records; record; record = record->next)
347 wrbuf_printf(w, "score = relevance(%d)/cluster_size(%d);\n",
348 relevance, cluster_size);
349 relevance /= cluster_size;
353 wrbuf_printf(w, "score = relevance(%d);\n", relevance);
355 rec->relevance_score = relevance;
357 reclist_leave(reclist);
364 * c-file-style: "Stroustrup"
365 * indent-tabs-mode: nil
367 * vim: shiftwidth=4 tabstop=8 expandtab