From 2e605eb2745e56619b16fdbcb8095e0a6543ab27 Mon Sep 17 00:00:00 2001 From: Patrick Simianer Date: Wed, 3 Aug 2011 01:29:52 +0200 Subject: refactoring, cleaning up --- dtrain/dtest.cc | 47 ++++++++++++++++++++++------------------------- 1 file changed, 22 insertions(+), 25 deletions(-) (limited to 'dtrain/dtest.cc') diff --git a/dtrain/dtest.cc b/dtrain/dtest.cc index 5ae473e6..d1ff30c0 100644 --- a/dtrain/dtest.cc +++ b/dtrain/dtest.cc @@ -1,6 +1,6 @@ -#include "dcommon.h" - - +#include "common.h" +#include "kbestget.h" +#include "util.h" /* @@ -14,10 +14,10 @@ init(int argc, char** argv, po::variables_map* conf) bool q; po::options_description opts( "Options" ); opts.add_options() - ( "decoder-config,c", po::value(), "configuration file for cdec" ) - ( "weights,w", po::value(), "weights file") - ( "ngrams,n", po::value(&N)->default_value(4), "N for Ngrams (default 5)" ) - ( "quiet,q", po::value(&q)->default_value(true), "do not output translations" ); + ( "decoder-config,c", po::value(), "configuration file for cdec" ) + ( "weights,w", po::value(), "weights file" ) + ( "ngrams,n", po::value(&N)->default_value(DTRAIN_DEFAULT_N), "N for Ngrams (default 5)" ) + ( "quiet,q", po::value(&q)->default_value(true), "do not output translations" ); po::options_description cmdline_options; cmdline_options.add(opts); po::store( parse_command_line(argc, argv, cmdline_options), *conf ); @@ -57,17 +57,17 @@ main(int argc, char** argv) vector strs, ref_strs; vector ref_ids; string in, psg; - size_t sid = 0; - double overall = 0.0; + size_t sn = 0; + double overall = 0.0; double overall1 = 0.0; double overall2 = 0.0; - cerr << "(a dot equals 100 lines of input)" << endl; + cerr << "(A dot equals " << DTRAIN_DOTOUT << " lines of input.)" << endl; while( getline(cin, in) ) { - if ( (sid+1) % 100 == 0 ) { + if ( (sn+1) % DTRAIN_DOTOUT == 0 ) { cerr << "."; - if ( (sid+1)%1000 == 0 ) cerr << endl; + if ( (sn+1) % (20*DTRAIN_DOTOUT) == 0 ) cerr << endl; } - //if ( sid > 5000 ) break; + //if ( sn > 5000 ) break; strs.clear(); boost::split( strs, in, boost::is_any_of("\t") ); // grammar @@ -80,25 +80,22 @@ main(int argc, char** argv) boost::split( ref_strs, strs[1], boost::is_any_of(" ") ); register_and_convert( ref_strs, ref_ids ); // scoring kbest - double score = 0.0; + double score = 0.0; double score1 = 0.0; double score2 = 0.0; NgramCounts counts = make_ngram_counts( ref_ids, kb->sents[0], 4 ); - score = smooth_bleu( counts, - ref_ids.size(), - kb->sents[0].size(), N ); - score1 = stupid_bleu( counts, ref_ids.size(), kb->sents[0].size(), N) ; - score2 = bleu( counts, ref_ids.size(), kb->sents[0].size(), N ); - //if ( ! quiet ) - cout << TD::GetString( kb->sents[0] ) << endl; + score = smooth_bleu( counts, ref_ids.size(), kb->sents[0].size(), N ); + score1 = stupid_bleu( counts, ref_ids.size(), kb->sents[0].size(), N ); + score2 = bleu( counts, ref_ids.size(), kb->sents[0].size(), N ); + if ( ! quiet ) cout << TD::GetString( kb->sents[0] ) << endl; overall += score; overall1 += score1; overall2 += score2; - sid += 1; + sn += 1; } - cerr << "Average score (smooth): " << overall/(double)(sid+1) << endl; - cerr << "Average score (stupid): " << overall1/(double)(sid+1) << endl; - cerr << "Average score (normal): " << overall2/(double)(sid+1) << endl; + cerr << "Average score (smooth) : " << overall/(double)(sn+1) << endl; + cerr << "Average score (stupid) : " << overall1/(double)(sn+1) << endl; + cerr << "Average score (vanilla): " << overall2/(double)(sn+1) << endl; cerr << endl; return 0; -- cgit v1.2.3