geonmin-kim's picture
Upload folder using huggingface_hub
d6585f5
/*
Copyright (c) 2008 - Chris Buckley.
Permission is granted for use and modification of this file for
research, non-commercial purposes.
*/
#include "common.h"
#include "sysfunc.h"
#include "trec_eval.h"
#include "functions.h"
#include "trec_format.h"
double log2(double x);
static int
te_calc_ndcg_cut (const EPI *epi, const REL_INFO *rel_info,
const RESULTS *results, const TREC_MEAS *tm, TREC_EVAL *eval);
static long long_cutoff_array[] = {5, 10, 15, 20, 30, 100, 200, 500, 1000};
static PARAMS default_ndcg_cutoffs = {
NULL, sizeof (long_cutoff_array) / sizeof (long_cutoff_array[0]),
&long_cutoff_array[0]};
/* See trec_eval.h for definition of TREC_MEAS */
TREC_MEAS te_meas_ndcg_cut =
{"ndcg_cut",
" Normalized Discounted Cumulative Gain at cutoffs.\n\
Compute a traditional nDCG measure according to Jarvelin and\n\
Kekalainen (ACM ToIS v. 20, pp. 422-446, 2002) at cutoffs.\n\
See comments for ndcg.\n\
Gain values are the relevance values in the qrels file. For now, if you\n\
want different gains, change the qrels file appropriately.\n\
Cutoffs must be positive without duplicates\n\
Default params: -m ndcg_cut.5,10,15,20,30,100,200,500,1000\n\
Based on an implementation by Ian Soboroff\n",
te_init_meas_a_float_cut_long,
te_calc_ndcg_cut,
te_acc_meas_a_cut,
te_calc_avg_meas_a_cut,
te_print_single_meas_a_cut,
te_print_final_meas_a_cut,
(void *) &default_ndcg_cutoffs, -1};
static int
te_calc_ndcg_cut (const EPI *epi, const REL_INFO *rel_info,
const RESULTS *results, const TREC_MEAS *tm, TREC_EVAL *eval)
{
long *cutoffs = (long *) tm->meas_params->param_values;
long cutoff_index = 0;
RES_RELS res_rels;
double gain, sum;
double ideal_dcg; /* ideal discounted cumulative gain */
long cur_lvl, lvl_count;
long i;
if (UNDEF == te_form_res_rels (epi, rel_info, results, &res_rels))
return (UNDEF);
sum = 0.0;
for (i = 0; i < res_rels.num_ret; i++) {
if (i == cutoffs[cutoff_index]) {
/* Calculate previous cutoff threshold.
Note i guaranteed to be positive by init_meas */
eval->values[tm->eval_index + cutoff_index].value = sum;
if (++cutoff_index == tm->meas_params->num_params)
break;
if (epi->debug_level > 0)
printf("ndcg_cut: cutoff %ld dcg %6.4f\n", i, sum);
}
gain = res_rels.results_rel_list[i];
if (gain > 0) {
/* Note: i+2 since doc i has rank i+1 */
sum += gain / log2((double) (i+2));
if (epi->debug_level > 1)
printf("ndcg_cut:%ld %3.1f %6.4f\n", i, gain, sum);
}
}
/* calculate values for those cutoffs not achieved */
while (cutoff_index < tm->meas_params->num_params) {
eval->values[tm->eval_index + cutoff_index].value = sum;
if (epi->debug_level > 0)
printf("ndcg_cut: cutoff %ld dcg %6.4f\n",
cutoffs[cutoff_index], sum);
cutoff_index++;
}
/* Calculate ideal discounted cumulative gain for this topic, and
normalize previous sum by it */
cutoff_index = 0;
cur_lvl = res_rels.num_rel_levels - 1;
lvl_count = 0;
ideal_dcg = 0.0;
for (i = 0; 1; i++) {
lvl_count++;
while (cur_lvl > 0 && lvl_count > res_rels.rel_levels[cur_lvl]) {
cur_lvl--;
lvl_count = 1;
}
if (cur_lvl == 0)
break;
if (i == cutoffs[cutoff_index]) {
/* Calculate previous cutoff threshold.
Note i guaranteed to be positive by init_meas */
if (ideal_dcg > 0.0)
eval->values[tm->eval_index + cutoff_index].value /= ideal_dcg;
if (epi->debug_level > 0)
printf("ndcg_cut: cutoff %ld idcg %6.4f\n", i, ideal_dcg);
if (++cutoff_index == tm->meas_params->num_params)
break;
}
gain = cur_lvl;
ideal_dcg += gain / (double) log2((double)(i + 2));
if (epi->debug_level > 0)
printf("ndcg_cut:%ld %ld %3.1f %6.4f\n",i, cur_lvl, gain,ideal_dcg);
}
/* calculate values for those cutoffs not achieved */
while (cutoff_index < tm->meas_params->num_params) {
if (ideal_dcg > 0.0)
eval->values[tm->eval_index + cutoff_index].value /= ideal_dcg;
if (epi->debug_level > 0)
printf("ndcg_cut: cutoff %ld idcg %6.4f\n",
cutoffs[cutoff_index], ideal_dcg);
cutoff_index++;
}
return (1);
}