/**
* \ingroup search
*
* Locate an n-gram in the indexed corpus, return its locations as <sentId, offsetInSent> pairs
* SentID and offset are all 1-based
*
* Note:
* The offset of the n-gram in a sentence is represented as "char" in the returned structure S_SimplePhraseLocationElement
* To output it as a number, one needs to cast it to integer type for proper display
*
*
* Revision $Rev: 3794 $
* Last Modified $LastChangedDate: 2007-06-29 02:17:32 -0400 (Fri, 29 Jun 2007) $
**/
int main(int argc, char * argv[]){
//-----------------------------------------------------------------------------
//check parameter
if(argc<2){
fprintf(stderr,"\nOutput all the locations of an n-gram in an indexed corpus\n");
fprintf(stderr,"\nUsage:\n");
fprintf(stderr,"\n%s corpusFileNameStem < list of n-grams\n\n",argv[0]);
exit(-1);
}
//-----------------------------------------------------------------------------
C_SuffixArraySearchApplicationBase saObj;
//load the indexed corpus with vocabulary(noVoc=false) and with offset(noOffset=false)
saObj.loadData_forSearch(argv[1], false, false);
cerr<<"Input N-grams:\n";
char tmpString[10000];
while(!cin.eof()){
cin.getline(tmpString,10000,'\n');
if(strlen(tmpString)>0){
vector<S_SimplePhraseLocationElement> locations;
locations = saObj.locateExactPhraseInCorpus(tmpString);
if(locations.size()==0){
cout<<"No occurrences found.\n";
}
else{
for(int i=0;i<locations.size(); i++){
cout<<"SentId="<<locations[i].sentIdInCorpus<<" Pos="<<(int)locations[i].posInSentInCorpus<<endl;
}
}
cout<<endl;
}
}
return 0;
}
int main(int argc, char * argv[])
{
int c;
const char* efile=0;
const char* ffile=0;
int pfe_index = 2;
while ((c = getopt(argc, argv, "cpf:e:i:n:l:m:h")) != -1) {
switch (c) {
case 'e':
efile = optarg;
break;
case 'f':
ffile = optarg;
break;
case 'i': // index of pfe in phrase table
pfe_index = atoi(optarg);
break;
case 'n': // keep only the top n entries in phrase table sorted by p(f|e) (0=all)
pfe_filter_limit = atoi(optarg);
std::cerr << "P(f|e) filter limit: " << pfe_filter_limit << std::endl;
break;
case 'c':
print_cooc_counts = true;
break;
case 'p':
print_neglog_significance = true;
break;
case 'h':
hierarchical = true;
break;
case 'm':
max_cache = atoi(optarg);
break;
case 'l':
std::cerr << "-l = " << optarg << "\n";
if (strcmp(optarg,"a+e") == 0) {
sig_filter_limit = ALPHA_PLUS_EPS;
} else if (strcmp(optarg,"a-e") == 0) {
sig_filter_limit = ALPHA_MINUS_EPS;
} else {
char *x;
sig_filter_limit = strtod(optarg, &x);
if (sig_filter_limit < 0.0) {
std::cerr << "Filter limit (-l) must be either 'a+e', 'a-e' or a real number >= 0.0\n";
usage();
}
}
break;
default:
usage();
}
}
if (sig_filter_limit == 0.0) pef_filter_only = true;
//-----------------------------------------------------------------------------
if (optind != argc || ((!efile || !ffile) && !pef_filter_only)) {
usage();
}
//load the indexed corpus with vocabulary(noVoc=false) and with offset(noOffset=false)
if (!pef_filter_only) {
e_sa.loadData_forSearch(efile, false, false);
f_sa.loadData_forSearch(ffile, false, false);
size_t elines = e_sa.returnTotalSentNumber();
size_t flines = f_sa.returnTotalSentNumber();
if (elines != flines) {
std::cerr << "Number of lines in e-corpus != number of lines in f-corpus!\n";
usage();
} else {
std::cerr << "Training corpus: " << elines << " lines\n";
num_lines = elines;
}
p_111 = -log(fisher_exact(1,1,1));
std::cerr << "\\alpha = " << p_111 << "\n";
if (sig_filter_limit == ALPHA_MINUS_EPS) {
sig_filter_limit = p_111 - 0.001;
} else if (sig_filter_limit == ALPHA_PLUS_EPS) {
sig_filter_limit = p_111 + 0.001;
}
std::cerr << "Sig filter threshold is = " << sig_filter_limit << "\n";
} else {
std::cerr << "Filtering using P(e|f) only. n=" << pfe_filter_limit << std::endl;
}
char tmpString[10000];
std::string prev = "";
std::vector<PTEntry*> options;
size_t pt_lines = 0;
while(!cin.eof()) {
cin.getline(tmpString,10000,'\n');
if(++pt_lines%10000==0) {
std::cerr << ".";
prune_cache(esets);
prune_cache(fsets);
if(pt_lines%500000==0)
std::cerr << "[n:"<<pt_lines<<"]\n";
}
if(strlen(tmpString)>0) {
PTEntry* pp = new PTEntry(tmpString, pfe_index);
if (prev != pp->f_phrase) {
prev = pp->f_phrase;
if (!options.empty()) { // always true after first line
compute_cooc_stats_and_filter(options);
}
for (std::vector<PTEntry*>::iterator i=options.begin(); i != options.end(); ++i) {
std::cout << **i << std::endl;
delete *i;
}
options.clear();
options.push_back(pp);
} else {
options.push_back(pp);
}
// for(int i=0;i<locations.size(); i++){
// cout<<"SentId="<<locations[i].sentIdInCorpus<<" Pos="<<(int)locations[i].posInSentInCorpus<<endl;
// }
}
}
compute_cooc_stats_and_filter(options);
for (std::vector<PTEntry*>::iterator i=options.begin(); i != options.end(); ++i) {
std::cout << **i << std::endl;
delete *i;
}
float pfefper = (100.0*(float)nremoved_pfefilter)/(float)pt_lines;
float sigfper = (100.0*(float)nremoved_sigfilter)/(float)pt_lines;
std::cerr << "\n\n------------------------------------------------------\n"
<< " unfiltered phrases pairs: " << pt_lines << "\n"
<< "\n"
<< " P(f|e) filter [first]: " << nremoved_pfefilter << " (" << pfefper << "%)\n"
<< " significance filter: " << nremoved_sigfilter << " (" << sigfper << "%)\n"
<< " TOTAL FILTERED: " << (nremoved_pfefilter + nremoved_sigfilter) << " (" << (sigfper + pfefper) << "%)\n"
<< "\n"
<< " FILTERED phrase pairs: " << (pt_lines - nremoved_pfefilter - nremoved_sigfilter) << " (" << (100.0-sigfper - pfefper) << "%)\n"
<< "------------------------------------------------------\n";
return 0;
}