diff --git a/UD/template/maca_trans_tagger/Makefile b/UD/template/maca_trans_tagger/Makefile
index 884cd6959dc58ff34d0a3c4e19df87f8c6dc3c06..b4a2855d474f0ad5d6e60a2f1ed23e4a89265774 100644
--- a/UD/template/maca_trans_tagger/Makefile
+++ b/UD/template/maca_trans_tagger/Makefile
@@ -3,11 +3,15 @@ MCF_DEV=../data/treebank/dev.mcf
MCF_TEST=../data/treebank/test.mcf
CFF_TRAIN=train.cff
+CFF_FANN_TRAIN=train.fann.cff
+FANN_TRAIN=train.fann
CFF_CUTOFF_TRAIN=train.cutoff.cff
PERCEPTRON_ITERATIONS=9
CFF_CUTOFF=1
FEATURES_MODEL_FILENAME=../../fm/maca_trans_tagger.fm
+FEATURES_MODEL_FANN_FILENAME=../../fm/maca_trans_tagger_fann.fm
VOCABS_FILENAME=maca_trans_tagger.vocab
+VOCABS_FANN_FILENAME=maca_trans_parser_fann.vocab
MCD_FILENAME=../../mcd/maca_trans_tagger.mcd
MODEL_FILENAME=maca_trans_tagger.model
NUMBER_OF_SENTENCES=10000000
@@ -18,3 +22,5 @@ FORM_POS_FILENAME=../data/morpho-lexicon/fP
#include ./maca_trans_tagger.makefile
include ../../makefiles/maca_trans_tagger.makefile
+
+
diff --git a/UD/ud_template.tgz b/UD/ud_template.tgz
index 002ed8421208367c50605e9025a7dae8e2cb125d..f4d2ffaea32cb3e490c6e2457c01a76f60ce6928 100644
Binary files a/UD/ud_template.tgz and b/UD/ud_template.tgz differ
diff --git a/fm/maca_trans_parser.fm b/fm/maca_trans_parser.fm
new file mode 100644
index 0000000000000000000000000000000000000000..6e18e5ed6fec6180187615dd6eafd3ff5df848d7
--- /dev/null
+++ b/fm/maca_trans_parser.fm
@@ -0,0 +1,56 @@
+#b0m
+#s0m
+#b0m s0m
+#s0l s0m b0l b0m
+
+
+b0g
+s0g s0p
+s0g b0p
+s0g
+s0sf
+#s1g
+#s1sf
+s0l
+s0p
+s1p
+s2p
+b0l
+b0p
+b1l
+b1p
+b2p
+b3p
+ldep_s0r
+rdep_s0r
+ldep_s1r
+rdep_s1r
+ldep_b0r
+rdep_b0r
+s0l b0l
+
+s0p b0p
+b0p b0l
+b0p ldep_b0r
+s1p b1p
+b1p b2p
+s0p b0p b0l
+s0p ldep_s0r rdep_s0r
+s0p s0l b0p
+s0p b0p dist_s0_b0
+s1p s0p b0p
+b0p b1p b2p
+b1p b2p b3p
+s0p b0p b1p
+b1p b1l b2p b3p
+b1p b1l b2p b2l b3p
+t1
+#t2
+#t3
+#t4
+t1 t2
+#t2 t3
+t1 t2 t3
+
+bm1p
+bm2p
diff --git a/fm/maca_trans_parser_fann.fm b/fm/maca_trans_parser_fann.fm
new file mode 100644
index 0000000000000000000000000000000000000000..3fdb2ce20b7984daf982a5293ba17bb162c5485c
--- /dev/null
+++ b/fm/maca_trans_parser_fann.fm
@@ -0,0 +1,25 @@
+b0l
+b0p
+b1l
+b1p
+b2l
+b2p
+b3p
+bm1p
+bm2p
+dist_s0_b0
+ldep_b0r
+ldep_s0r
+ldep_s1r
+rdep_b0r
+rdep_s0r
+rdep_s1r
+s0g
+s0l
+s0p
+s0sf
+s1p
+s2p
+t1
+t2
+t3
diff --git a/fm/maca_trans_tagger.fm b/fm/maca_trans_tagger.fm
new file mode 100644
index 0000000000000000000000000000000000000000..50af67ebe5ff61fa2c71a524cdcd40401183c8df
--- /dev/null
+++ b/fm/maca_trans_tagger.fm
@@ -0,0 +1,37 @@
+b0U1
+b0len
+
+b0sgn
+b1sgn
+
+b1f
+b0f
+bm1f
+bm2f
+
+bm1p
+bm2p
+bm3p
+bm2p bm1p
+bm2p bm3p
+bm1p b0sgn
+
+b0s1
+b0s2
+b0s3
+b0s4
+b0s5
+b0s1 b0s2
+b0s1 b0s2 b0s3
+b0s1 b0s2 b0s3 b0s4
+
+b0p1
+b0p2
+b0p3
+b0p4
+b0p5
+b0p1 b0p2
+b0p1 b0p2 b0p3
+b0p1 b0p2 b0p3 b0p4
+
+
diff --git a/fm/maca_trans_tagger_fann.fm b/fm/maca_trans_tagger_fann.fm
new file mode 100644
index 0000000000000000000000000000000000000000..3f5e2c2add768705ac78571e918e69d6c8ee3a24
--- /dev/null
+++ b/fm/maca_trans_tagger_fann.fm
@@ -0,0 +1,21 @@
+b0f
+b0len
+b0p1
+b0p2
+b0p3
+b0p4
+b0p5
+b0s1
+b0s2
+b0s3
+b0s4
+b0s5
+b0sgn
+b0U1
+b1f
+b1sgn
+bm1f
+bm1p
+bm2f
+bm2p
+bm3p
diff --git a/makefiles/maca_trans_tagger.makefile b/makefiles/maca_trans_tagger.makefile
index f32b03a31bdfa0918700a58123f0eb1c8b0f73ca..ce3d13ed870e396e6f90caf8048c50ebe626a741 100644
--- a/makefiles/maca_trans_tagger.makefile
+++ b/makefiles/maca_trans_tagger.makefile
@@ -2,14 +2,21 @@
## compile
##-----------------------------------------------------------------------
-compile: $(MODEL_FILENAME)
+compile: $(MODEL_FILENAME) $(FANN_TRAIN)
$(CFF_TRAIN): $(MCF_TRAIN)
maca_trans_tagger_mcf2cff -C $(MCD_FILENAME) --input $< --mode TRAIN --feat_model $(FEATURES_MODEL_FILENAME) --vocabs $(VOCABS_FILENAME) --cff $@ -s $(NUMBER_OF_SENTENCES) $(STREAM_MODE) -P $(FORM_POS_FILENAME)
+$(CFF_FANN_TRAIN): $(MCF_TRAIN)
+ maca_trans_tagger_mcf2cff -C $(MCD_FILENAME) --input $< --mode TRAIN --feat_model $(FEATURES_MODEL_FANN_FILENAME) --vocabs $(VOCABS_FANN_FILENAME) --cff $@ -s $(NUMBER_OF_SENTENCES) $(STREAM_MODE) -P $(FORM_POS_FILENAME)
+
+
$(CFF_CUTOFF_TRAIN): $(CFF_TRAIN)
cff_cutoff --input $< --vocabs $(VOCABS_FILENAME) --cutoff $(CFF_CUTOFF) > $@
+$(FANN_TRAIN): $(CFF_FANN_TRAIN)
+ cff2fann --vocabs $(VOCABS_FANN_FILENAME) --cff $< --feat_model $(FEATURES_MODEL_FANN_FILENAME) -C $(MCD_FILENAME) > $@
+
$(MODEL_FILENAME): $(CFF_CUTOFF_TRAIN)
#$(MODEL_FILENAME): $(CFF_TRAIN)
perceptron_train --cff $< --model $(MODEL_FILENAME) -n $(PERCEPTRON_ITERATIONS)
diff --git a/tools/conllu2mcf.c b/tools/conllu2mcf.c
new file mode 100644
index 0000000000000000000000000000000000000000..a2fc73b015c309270de065d82dcce484c5c0973b
--- /dev/null
+++ b/tools/conllu2mcf.c
@@ -0,0 +1,158 @@
+#include<stdio.h>
+#include<stdlib.h>
+#include<string.h>
+#include<strings.h>
+#include<math.h>
+#include<getopt.h>
+#include"conll_lib.h"
+#include"hash_str.h"
+
+#define NB_COL 7
+
+typedef struct options
+{
+ FILE * fd_parses; // parser output
+ int verbose_level;
+ int snum;
+ char *filename;
+ char columns[NB_COL];
+} options;
+
+/*---------------------------------------------------------------------------------*/
+
+options op;
+
+void print_options(options *op)
+{
+ fprintf(stderr, "file name = %s\n", op->filename);
+ fprintf(stderr, "verbose level = %d\n", op->verbose_level);
+ fprintf(stderr, "maximum number of sentences to process = %d\n", op->snum);
+}
+
+void reset_options(options * op)
+{
+ int i;
+ op->filename = NULL;
+ op->fd_parses = stdin;
+ op->verbose_level = 0;
+ op->snum = 100000000;
+ for(i=0; i < NB_COL; i++)
+ op->columns[i] = '0';
+}
+
+/*---------------------------------------------------------------------------------*/
+void print_help_message(char *program_name)
+{
+ fprintf(stderr, "%s usage: %s [options]\n", program_name, program_name);
+ fprintf(stderr, "OPTIONS :\n");
+ fprintf(stderr, " -f <file> : hypothesis conll file\n");
+ fprintf(stderr, " -n <int> : process n sentences (default is 100 000 000)\n");
+ fprintf(stderr, " -v 1|2|3 : verbosity level\n");
+ fprintf(stderr, " -h : print this message\n");
+
+ fprintf(stderr, " -1 : content of column 1 in the mcf file produced\n");
+ fprintf(stderr, " -2 : content of column 2 in the mcf file produced\n");
+ fprintf(stderr, " -3 : content of column 3 in the mcf file produced\n");
+ fprintf(stderr, " -4 : content of column 4 in the mcf file produced\n");
+ fprintf(stderr, " -5 : content of column 5 in the mcf file produced\n");
+ fprintf(stderr, " -6 : content of column 6 in the mcf file produced\n");
+ fprintf(stderr, " -7 : content of column 7 in the mcf file produced\n");
+ fprintf(stderr, " : values of options -1 to -7 must be one of\n");
+ fprintf(stderr, " : I for id\n");
+ fprintf(stderr, " : W for form\n");
+ fprintf(stderr, " : L for lemma\n");
+ fprintf(stderr, " : C for coarse part of speech\n");
+ fprintf(stderr, " : P for part of speech\n");
+ fprintf(stderr, " : F for features\n");
+ fprintf(stderr, " : H for head\n");
+ fprintf(stderr, " : D for deprel\n");
+
+}
+
+
+
+
+/*---------------------------------------------------------------------------------*/
+
+void parse_options(int argc, char *argv[], options * op)
+{
+ char c;
+
+ reset_options(op);
+ /*
+ if(argc ==1){
+ print_help_message(argv[0]);
+ exit(1);
+ }*/
+
+ while ((c = getopt (argc, argv, "hIWLCPFHDf:n:v:1:2:3:4:5:6:7:8:9:")) != -1)
+ switch (c)
+ {
+ case 'h':
+ print_help_message(argv[0]);
+ exit(0);
+ case 'f':
+ op->filename = strdup(optarg);
+ if((op->fd_parses = fopen(op->filename, "r")) == NULL){
+ fprintf(stderr, "cannot open hypothesis file %s : aborting\n", op->filename);
+ exit(1);
+ }
+ break;
+ case '1':
+ op->columns[0] = optarg[0];
+ break;
+ case '2':
+ op->columns[1] = optarg[0];
+ break;
+ case '3':
+ op->columns[2] = optarg[0];
+ break;
+ case '4':
+ op->columns[3] = optarg[0];
+ break;
+ case '5':
+ op->columns[4] = optarg[0];
+ break;
+ case '6':
+ op->columns[5] = optarg[0];
+ break;
+ case '7':
+ op->columns[6] = optarg[0];
+ break;
+ case 'n':
+ op->snum = atoi(optarg);
+ break;
+ case 'v':
+ op->verbose_level = atoi(optarg);
+ break;
+ }
+
+ /* if (op->fd_parses == NULL){
+ fprintf(stderr, "error : cannot open parse file: aborting\n");
+ exit(1);
+ }*/
+}
+
+/*---------------------------------------------------------------------------------*/
+/*---------------------------------------------------------------------------------*/
+int main(int argc, char *argv[])
+{
+ sentence *s = allocate_sentence();
+ int snum = 0;
+ int res;
+ parse_options(argc, argv, &op);
+
+ print_options(&op);
+
+
+ for(res = load_sentence(op.fd_parses, s); res && (snum < op.snum); res = load_sentence(op.fd_parses, s)){
+ s->num = snum;
+ snum++;
+ compute_relative_index_of_heads(s);
+ print_sentence_mcf3(s, op.columns, NB_COL);
+ }
+ if(op.filename)
+ fclose(op.fd_parses);
+ free_sentence(s);
+ return 0;
+}
diff --git a/tools/eval_mcf.pl b/tools/eval_mcf.pl
new file mode 100755
index 0000000000000000000000000000000000000000..f8fb57be800915309d779c63b392f97aea15d6d9
--- /dev/null
+++ b/tools/eval_mcf.pl
@@ -0,0 +1,324 @@
+#!/usr/bin/perl
+
+$arg = shift;
+while($arg){
+ if($arg eq "-g"){$ref = shift;}
+ elsif($arg eq "-s"){$hyp = shift;}
+ elsif($arg eq "-G"){$ref_mcd = shift;}
+ elsif($arg eq "-S"){$hyp_mcd = shift;}
+ elsif($arg eq "-tac"){$TAGGING_ACCURACY_PER_CATEGORY = 1;}
+ elsif($arg eq "-tcm"){$TAGGING_CONFUSION_MATRIX = 1;}
+ elsif($arg eq "-tec"){$TAGGING_ERRORS_PER_CATEGORY = 1;}
+ elsif($arg eq "-paf"){$PARSING_ACCURACY_PER_FUNCTION = 1;}
+ elsif($arg eq "-lcm"){$LABELING_CONFUSION_MATRIX = 1;}
+ elsif($arg eq "-acm"){$ATTACHEMENT_CONFUSION_MATRIX = 1;}
+ elsif($arg eq "-all"){
+ $TAGGING_ACCURACY_PER_CATEGORY = 1;
+ $TAGGING_CONFUSION_MATRIX = 1;
+ $TAGGING_ERRORS_PER_CATEGORY = 1;
+ $PARSING_ACCURACY_PER_FUNCTION = 1;
+ $LABELING_CONFUSION_MATRIX = 1;
+ $ATTACHEMENT_CONFUSION_MATRIX = 1;
+ }
+ elsif($arg eq "-h"){
+ print "usage eval07.pl OPTIONS -g <reference file> -s <system output>\n";
+ print "OPTIONS :\n";
+ print "\t-tac tagging accuracy per category\n";
+ print "\t-tcm tagging confusion matrix\n";
+ print "\t-tec tagging errors per category\n";
+ print "\t-paf parsing accuracy per function\n";
+ print "\t-lcm labeling confusion matrix\n";
+ print "\t-acm attachment confusion matrix\n";
+ print "\t-all all options\n";
+ exit;
+}
+$arg = shift;
+}
+
+# determine the column in the reference file
+
+$ref_form_col = 0;
+$ref_pos_col = 1;
+$ref_lemma_col = 2;
+$ref_gov_col = 3;
+$ref_fct_col = 4;
+$ref_seg_col = 5;
+$ref_morph_col = 10;
+
+if($ref_mcd)
+{
+ for($i=0; $i<length $ref_mcd; $i++){
+ $car = substr($ref_mcd, $i, 1);
+# print "car $i = $car\n";
+ if($car eq 'W'){$ref_form_col = $i; next;}
+ if($car eq 'P'){$ref_pos_col = $i; next;}
+ if($car eq 'L'){$ref_lemma_col = $i; next;}
+ if($car eq 'G'){$ref_gov_col = $i; next;}
+ if($car eq 'F'){$ref_fct_col = $i; next;}
+ if($car eq 'S'){$ref_seg_col = $i; next;}
+ if($car eq 'M'){$ref_morph_col = $i; next;}
+ }
+}
+
+# determine the column in the hypothesis file
+
+$hyp_form_col = 0;
+$hyp_pos_col = 1;
+$hyp_lemma_col = 2;
+$hyp_gov_col = 3;
+$hyp_fct_col = 4;
+$hyp_seg_col = 5;
+$hyp_morph_col = 10;
+
+if($hyp_mcd)
+{
+ for($i=0; $i<length $hyp_mcd; $i++){
+ $car = substr($hyp_mcd, $i, 1);
+# print "car $i = $car\n";
+ if($car eq 'W'){$hyp_form_col = $i; next;}
+ if($car eq 'P'){$hyp_pos_col = $i; next;}
+ if($car eq 'L'){$hyp_lemma_col = $i; next;}
+ if($car eq 'G'){$hyp_gov_col = $i; next;}
+ if($car eq 'F'){$hyp_fct_col = $i; next;}
+ if($car eq 'S'){$hyp_seg_col = $i; next;}
+ if($car eq 'M'){$hyp_morph_col = $i; next;}
+ }
+}
+
+
+open REF, $ref or die "cannot open file $ref";
+open HYP, $hyp or die "cannot open file $hyp";
+
+
+
+
+my $line_nb;
+my $word_nb;
+my $correct_pos_nb;
+my $correct_gov_nb;
+my $correct_gov_fct_nb;
+
+sub is_punctuation_ptb{
+ my $pos = shift(@_);
+
+ if($pos eq "``"){return 1;}
+ if($pos eq ","){return 1;}
+ if($pos eq ":"){return 1;}
+ if($pos eq "."){return 1;}
+ if($pos eq "''"){return 1;}
+ if($pos eq "-LRB-"){return 1;}
+ if($pos eq "-RRB-"){return 1;}
+ return 0;
+}
+sub is_punctuation_ftb{
+ my $pos = shift(@_);
+
+ if($pos eq "PCT"){return 1;}
+ if($pos eq "PONCT"){return 1;}
+ if($pos eq "ponctw"){return 1;}
+ if($pos eq "poncts"){return 1;}
+ return 0;
+}
+sub is_punctuation_ud{
+ my $pos = shift(@_);
+
+ if($pos eq "PUNCT"){return 1;}
+ return 0;
+}
+
+
+while(<REF>){
+ chop;
+ $line_nb++;
+# ($ref_form, $ref_pos, $ref_lemma, $ref_gov, $ref_fct, $ref_seg) = split /\t/;
+ @ref_array = split /\t/;
+ $column_nb = -1;
+ foreach $elt (@ref_array){
+ $column_nb++;
+ if($column_nb == $ref_form_col){$ref_form = $elt; next;}
+ if($column_nb == $ref_pos_col){$ref_pos = $elt; next;}
+ if($column_nb == $ref_lemma_col){$ref_lemma = $elt; next;}
+ if($column_nb == $ref_gov_col){$ref_gov = $elt; next;}
+ if($column_nb == $ref_fct_col){$ref_fct = $elt; next;}
+ if($column_nb == $ref_seg_col){$ref_seg = $elt; next;}
+ if($column_nb == $ref_morph_col){$ref_morph = $elt; next;}
+ }
+ $_ = <HYP>;
+ chop $_;
+# print;
+# ($hyp_form, $hyp_pos, $hyp_lemma, $hyp_gov, $hyp_fct, $hyp_seg) = split /\t/;
+
+ @hyp_array = split /\t/;
+ $column_nb = -1;
+ foreach $elt (@hyp_array){
+ $column_nb++;
+ if($column_nb == $hyp_form_col){$hyp_form = $elt; next;}
+ if($column_nb == $hyp_pos_col){$hyp_pos = $elt; next;}
+ if($column_nb == $hyp_lemma_col){$hyp_lemma = $elt; next;}
+ if($column_nb == $hyp_gov_col){$hyp_gov = $elt; next;}
+ if($column_nb == $hyp_fct_col){$hyp_fct = $elt; next;}
+ if($column_nb == $hyp_seg_col){$hyp_seg = $elt; next;}
+ if($column_nb == $hyp_morph_col){$hyp_morph = $elt; next;}
+ }
+
+
+
+# print "ref = $hyp_seg\n";
+
+ if($ref_seg){ $nb_ref_seg++;}
+ if($hyp_seg){ $nb_hyp_seg++;}
+
+ if(($ref_seg) && ($hyp_seg)){ $nb_hyp_ref_seg++;}
+
+# if(($ref_index) && (!is_punctuation_ftb($ref_pos)) && (!is_punctuation_ptb($ref_pos))){
+ if((!is_punctuation_ftb($ref_pos)) && (!is_punctuation_ptb($ref_pos))){
+# if((!is_punctuation_ftb($ref_pos)) && (!is_punctuation_ptb($ref_pos)) && (!is_punctuation_ud($ref_pos))){
+ if($ref_form ne $hyp_form){die "mismatch line $line_nb\n";}
+ $word_nb++;
+ $pos_nb{$ref_pos}++;
+ $fct_nb{$ref_fct}++;
+
+ if($ref_pos eq $hyp_pos){
+ $correct_pos_total_nb++;
+ $correct_pos_nb{$ref_pos}++;
+ }
+ else{
+ $false_pos_form{$ref_pos}{$ref_form}++;
+ $pos_confusion_matrix{$ref_pos}{$hyp_pos}++;
+
+
+# print "$ref_form $ref_pos $hyp_pos\n";
+# print "$ref_pos $hyp_pos\n";
+ }
+
+ if($ref_lemma eq $hyp_lemma){
+ $correct_lemma_total_nb++;
+ }
+ else{
+# print "$ref_form \t $ref_lemma \t $hyp_lemma\n";
+ }
+ $ref_dist = $ref_gov - $ref_index;
+ $hyp_dist = $hyp_gov - $hyp_index;
+# if($ref_gov eq $hyp_gov){
+ if($ref_dist eq $hyp_dist){
+ $correct_gov_nb++;
+ $correct_gov_total_nb++;
+ if($ref_fct eq $hyp_fct){
+ $correct_gov_fct_total_nb++;
+ $correct_gov_fct_nb{$ref_fct}++;
+ }
+ else{
+ $labeling_confusion_matrix{$ref_fct}{$hyp_fct}++;
+ }
+ }
+ else{
+ $attachement_confusion_matrix{$ref_fct}{$hyp_fct}++;
+ }
+
+ }
+
+ $ref_index = "";
+}
+
+
+close REF;
+close HYP;
+
+
+my $pos_acc = $correct_pos_total_nb / $word_nb * 100;
+my $lemma_acc = $correct_lemma_total_nb / $word_nb * 100;
+my $las = $correct_gov_fct_total_nb / $word_nb * 100;
+my $uas = $correct_gov_total_nb / $word_nb * 100 ;
+
+
+
+my $seg_recall = $nb_hyp_ref_seg / $nb_ref_seg;
+my $seg_precision = $nb_hyp_ref_seg / ($nb_hyp_seg + 1);
+
+
+printf(stderr "pos acc = %.2f lemma acc = %.2f uas = %.2f las = %.2f seg recall = %.2f seg precision = %.2f size = %d\n", $pos_acc, $lemma_acc, $uas, $las, $seg_recall, $seg_precision, $word_nb);
+printf(stdout "%s\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%d\n", $hyp, $pos_acc, $lemma_acc, $uas, $las, $seg_recall, $seg_precision, $word_nb);
+
+
+
+if($TAGGING_ACCURACY_PER_CATEGORY){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ printf "TAGGING ACCURACY PER CATEGORY\n";
+ printf "CAT\tFREQ\tACC\tIMPACT\n";
+ foreach $pos (keys %correct_pos_nb){
+ $acc = $correct_pos_nb{$pos} / $pos_nb{$pos};
+ $freq = $pos_nb{$pos} / $word_nb;
+ if($word_nb == $correct_pos_total_nb){
+ $impact = 0;
+ }
+ else{
+ $impact = ($pos_nb{$pos} - $correct_pos_nb{$pos}) / ($word_nb - $correct_pos_total_nb);
+ }
+ printf("%s\t%6.2f\t%6.2f\t%6.2f\n", $pos, $freq*100, $acc*100, $impact*100);
+ }
+}
+
+if($TAGGING_CONFUSION_MATRIX){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ print "TAGGING CONFUSION MATRIX\n";
+ foreach $ref_pos (keys %pos_confusion_matrix){
+ $pos_error_nb = $pos_nb{$ref_pos} - $correct_pos_nb{$ref_pos};
+ print "$ref_pos ($pos_error_nb) :";
+ foreach $hyp_pos (keys %{$pos_confusion_matrix{$ref_pos}}){
+ print "\t$hyp_pos ($pos_confusion_matrix{$ref_pos}{$hyp_pos})";
+ }
+ print "\n";
+ }
+}
+
+
+if($TAGGING_ERRORS_PER_CATEGORY){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ print "TAGGING ERRORS PER CATEGORY\n";
+ foreach $pos (keys %false_pos_form){
+ print "\n$pos\n";
+ foreach $form (keys %{$false_pos_form{$pos}}){
+ print "\t$form $false_pos_form{$pos}{$form}\n";
+ }
+ }
+}
+
+if($PARSING_ACCURACY_PER_FUNCTION){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ printf "LABELED ATTACHMENT SCORE PER LABEL\n";
+ printf "LABEL FREQ ACC IMPACT\n";
+ foreach $fct (keys %correct_gov_fct_nb){
+ $acc = $correct_gov_fct_nb{$fct} / $fct_nb{$fct};
+ $freq = $fct_nb{$fct}/$word_nb;
+ $impact = ($fct_nb{$fct} - $correct_gov_fct_nb{$fct}) / ($word_nb - $correct_gov_fct_total_nb++);
+ printf("%-10s%6.2f\t%6.2f\t%6.2f\n", $fct, $freq*100, $acc*100, $impact*100);
+ }
+}
+
+if($ATTACHEMENT_CONFUSION_MATRIX){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ printf "ATTACHEMENT CONFUSION MATRIX\n";
+ foreach $ref_fct (keys %attachement_confusion_matrix){
+ $attachement_error_nb = $fct_nb{$ref_fct} - $correct_gov_fct_nb{$ref_fct};
+ print "$ref_fct ($attachement_error_nb) :";
+ foreach $hyp_fct (keys %{$attachement_confusion_matrix{$ref_fct}}){
+ print "\t$hyp_fct ($attachement_confusion_matrix{$ref_fct}{$hyp_fct})";
+ }
+ print "\n";
+ }
+
+}
+
+if($LABELING_CONFUSION_MATRIX){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ printf "LABELING CONFUSION MATRIX\n";
+ foreach $ref_fct (keys %labeling_confusion_matrix){
+ $fct_error_nb = $fct_nb{$ref_fct} - $correct_gov_fct_nb{$ref_fct};
+ print "$ref_fct ($fct_error_nb) :";
+ foreach $hyp_fct (keys %{$labeling_confusion_matrix{$ref_fct}}){
+ print "\t$hyp_fct ($labeling_confusion_matrix{$ref_fct}{$hyp_fct})";
+ }
+ print "\n";
+ }
+
+}
diff --git a/tools/eval_wpmlgfs.pl b/tools/eval_wpmlgfs.pl
new file mode 100755
index 0000000000000000000000000000000000000000..a6d10975698e9b8a58291e867e283d755a4e16e0
--- /dev/null
+++ b/tools/eval_wpmlgfs.pl
@@ -0,0 +1,234 @@
+#!/usr/bin/perl
+
+$arg = shift;
+while($arg){
+ if($arg eq "-g"){$ref = shift;}
+ elsif($arg eq "-s"){$hyp = shift;}
+ elsif($arg eq "-tac"){$TAGGING_ACCURACY_PER_CATEGORY = 1;}
+ elsif($arg eq "-tcm"){$TAGGING_CONFUSION_MATRIX = 1;}
+ elsif($arg eq "-tec"){$TAGGING_ERRORS_PER_CATEGORY = 1;}
+ elsif($arg eq "-paf"){$PARSING_ACCURACY_PER_FUNCTION = 1;}
+ elsif($arg eq "-lcm"){$LABELING_CONFUSION_MATRIX = 1;}
+ elsif($arg eq "-acm"){$ATTACHEMENT_CONFUSION_MATRIX = 1;}
+ elsif($arg eq "-all"){
+ $TAGGING_ACCURACY_PER_CATEGORY = 1;
+ $TAGGING_CONFUSION_MATRIX = 1;
+ $TAGGING_ERRORS_PER_CATEGORY = 1;
+ $PARSING_ACCURACY_PER_FUNCTION = 1;
+ $LABELING_CONFUSION_MATRIX = 1;
+ $ATTACHEMENT_CONFUSION_MATRIX = 1;
+ }
+ elsif($arg eq "-h"){
+ print "usage eval07.pl OPTIONS -g <gold file> -s <system output>\n";
+ print "OPTIONS :\n";
+ print "\t-tac tagging accuracy per category\n";
+ print "\t-tcm tagging confusion matrix\n";
+ print "\t-tec tagging errors per category\n";
+ print "\t-paf parsing accuracy per function\n";
+ print "\t-lcm labeling confusion matrix\n";
+ print "\t-acm attachment confusion matrix\n";
+ print "\t-all all options\n";
+ exit;
+}
+$arg = shift;
+}
+
+open REF, $ref or die "cannot open file $ref";
+open HYP, $hyp or die "cannot open file $hyp";
+
+
+my $line_nb;
+my $word_nb;
+my $correct_pos_nb;
+my $correct_gov_nb;
+my $correct_gov_fct_nb;
+
+sub is_punctuation_ptb{
+ my $pos = shift(@_);
+
+ if($pos eq "``"){return 1;}
+ if($pos eq ","){return 1;}
+ if($pos eq ":"){return 1;}
+ if($pos eq "."){return 1;}
+ if($pos eq "''"){return 1;}
+ if($pos eq "-LRB-"){return 1;}
+ if($pos eq "-RRB-"){return 1;}
+ return 0;
+}
+sub is_punctuation_ftb{
+ my $pos = shift(@_);
+
+ if($pos eq "PCT"){return 1;}
+ if($pos eq "PONCT"){return 1;}
+ if($pos eq "ponctw"){return 1;}
+ if($pos eq "poncts"){return 1;}
+ return 0;
+}
+
+
+while(<REF>){
+ chop;
+ $line_nb++;
+ ($ref_form, $ref_pos, $ref_morpho, $ref_lemma, $ref_gov, $ref_fct, $ref_seg) = split /\t/;
+ $_ = <HYP>;
+ chop $_;
+# print;
+ ($hyp_form, $hyp_pos, $ref_morpho, $hyp_lemma, $hyp_gov, $hyp_fct, $hyp_seg) = split /\t/;
+# print "ref = $hyp_seg\n";
+
+ if($ref_seg){ $nb_ref_seg++;}
+ if($hyp_seg){ $nb_hyp_seg++;}
+
+ if(($ref_seg) && ($hyp_seg)){ $nb_hyp_ref_seg++;}
+
+# if(($ref_index) && (!is_punctuation_ftb($ref_pos)) && (!is_punctuation_ptb($ref_pos))){
+ if((!is_punctuation_ftb($ref_pos)) && (!is_punctuation_ptb($ref_pos))){
+ if($ref_form ne $hyp_form){die "mismatch line $line_nb\n";}
+ $word_nb++;
+ $pos_nb{$ref_pos}++;
+ $fct_nb{$ref_fct}++;
+
+ if($ref_pos eq $hyp_pos){
+ $correct_pos_total_nb++;
+ $correct_pos_nb{$ref_pos}++;
+ }
+ else{
+ $false_pos_form{$ref_pos}{$ref_form}++;
+ $pos_confusion_matrix{$ref_pos}{$hyp_pos}++;
+
+
+# print "$ref_form $ref_pos $hyp_pos\n";
+# print "$ref_pos $hyp_pos\n";
+ }
+
+ if($ref_lemma eq $hyp_lemma){
+ $correct_lemma_total_nb++;
+ }
+ else{
+# print "$ref_form \t $ref_lemma \t $hyp_lemma\n";
+ }
+ $ref_dist = $ref_gov - $ref_index;
+ $hyp_dist = $hyp_gov - $hyp_index;
+# if($ref_gov eq $hyp_gov){
+ if($ref_dist eq $hyp_dist){
+ $correct_gov_nb++;
+ $correct_gov_total_nb++;
+ if($ref_fct eq $hyp_fct){
+ $correct_gov_fct_total_nb++;
+ $correct_gov_fct_nb{$ref_fct}++;
+ }
+ else{
+ $labeling_confusion_matrix{$ref_fct}{$hyp_fct}++;
+ }
+ }
+ else{
+ $attachement_confusion_matrix{$ref_fct}{$hyp_fct}++;
+ }
+
+ }
+
+ $ref_index = "";
+}
+
+
+close REF;
+close HYP;
+
+
+my $pos_acc = $correct_pos_total_nb / $word_nb * 100;
+my $lemma_acc = $correct_lemma_total_nb / $word_nb * 100;
+my $las = $correct_gov_fct_total_nb / $word_nb * 100;
+my $uas = $correct_gov_total_nb / $word_nb * 100 ;
+
+
+
+my $seg_recall = $nb_hyp_ref_seg / $nb_ref_seg;
+my $seg_precision = $nb_hyp_ref_seg / ($nb_hyp_seg + 1);
+
+
+printf(stderr "pos acc = %.2f lemma acc = %.2f uas = %.2f las = %.2f seg recall = %.2f seg precision = %.2f size = %d\n", $pos_acc, $lemma_acc, $uas, $las, $seg_recall, $seg_precision, $word_nb);
+printf(stdout "%s\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%.2f\t%d\n", $hyp, $pos_acc, $lemma_acc, $uas, $las, $seg_recall, $seg_precision, $word_nb);
+
+
+
+if($TAGGING_ACCURACY_PER_CATEGORY){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ printf "TAGGING ACCURACY PER CATEGORY\n";
+ printf "CAT\tFREQ\tACC\tIMPACT\n";
+ foreach $pos (keys %correct_pos_nb){
+ $acc = $correct_pos_nb{$pos} / $pos_nb{$pos};
+ $freq = $pos_nb{$pos} / $word_nb;
+ if($word_nb == $correct_pos_total_nb){
+ $impact = 0;
+ }
+ else{
+ $impact = ($pos_nb{$pos} - $correct_pos_nb{$pos}) / ($word_nb - $correct_pos_total_nb);
+ }
+ printf("%s\t%6.2f\t%6.2f\t%6.2f\n", $pos, $freq*100, $acc*100, $impact*100);
+ }
+}
+
+if($TAGGING_CONFUSION_MATRIX){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ print "TAGGING CONFUSION MATRIX\n";
+ foreach $ref_pos (keys %pos_confusion_matrix){
+ $pos_error_nb = $pos_nb{$ref_pos} - $correct_pos_nb{$ref_pos};
+ print "$ref_pos ($pos_error_nb) :";
+ foreach $hyp_pos (keys %{$pos_confusion_matrix{$ref_pos}}){
+ print "\t$hyp_pos ($pos_confusion_matrix{$ref_pos}{$hyp_pos})";
+ }
+ print "\n";
+ }
+}
+
+
+if($TAGGING_ERRORS_PER_CATEGORY){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ print "TAGGING ERRORS PER CATEGORY\n";
+ foreach $pos (keys %false_pos_form){
+ print "\n$pos\n";
+ foreach $form (keys %{$false_pos_form{$pos}}){
+ print "\t$form $false_pos_form{$pos}{$form}\n";
+ }
+ }
+}
+
+if($PARSING_ACCURACY_PER_FUNCTION){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ printf "LABELED ATTACHMENT SCORE PER LABEL\n";
+ printf "LABEL FREQ ACC IMPACT\n";
+ foreach $fct (keys %correct_gov_fct_nb){
+ $acc = $correct_gov_fct_nb{$fct} / $fct_nb{$fct};
+ $freq = $fct_nb{$fct}/$word_nb;
+ $impact = ($fct_nb{$fct} - $correct_gov_fct_nb{$fct}) / ($word_nb - $correct_gov_fct_total_nb++);
+ printf("%-10s%6.2f\t%6.2f\t%6.2f\n", $fct, $freq*100, $acc*100, $impact*100);
+ }
+}
+
+if($ATTACHEMENT_CONFUSION_MATRIX){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ printf "ATTACHEMENT CONFUSION MATRIX\n";
+ foreach $ref_fct (keys %attachement_confusion_matrix){
+ $attachement_error_nb = $fct_nb{$ref_fct} - $correct_gov_fct_nb{$ref_fct};
+ print "$ref_fct ($attachement_error_nb) :";
+ foreach $hyp_fct (keys %{$attachement_confusion_matrix{$ref_fct}}){
+ print "\t$hyp_fct ($attachement_confusion_matrix{$ref_fct}{$hyp_fct})";
+ }
+ print "\n";
+ }
+
+}
+
+if($LABELING_CONFUSION_MATRIX){
+ print "\n\n--------------------------------------------------------------------------------------\n";
+ printf "LABELING CONFUSION MATRIX\n";
+ foreach $ref_fct (keys %labeling_confusion_matrix){
+ $fct_error_nb = $fct_nb{$ref_fct} - $correct_gov_fct_nb{$ref_fct};
+ print "$ref_fct ($fct_error_nb) :";
+ foreach $hyp_fct (keys %{$labeling_confusion_matrix{$ref_fct}}){
+ print "\t$hyp_fct ($labeling_confusion_matrix{$ref_fct}{$hyp_fct})";
+ }
+ print "\n";
+ }
+
+}
diff --git a/tools/fplm2fP_ud.pl b/tools/fplm2fP_ud.pl
new file mode 100755
index 0000000000000000000000000000000000000000..0758ed5ef5224f8f105aa1edb7dde3c1c34372e3
--- /dev/null
+++ b/tools/fplm2fP_ud.pl
@@ -0,0 +1,58 @@
+#!/usr/bin/perl
+
+
+$postag{"ADJ"} = 1;
+$postag{"ADP"} = 1;
+$postag{"ADV"} = 1;
+$postag{"AUX"} = 1;
+$postag{"CCONJ"} = 1;
+$postag{"DET"} = 1;
+$postag{"INTJ"} = 1;
+$postag{"NOUN"} = 1;
+$postag{"NUM"} = 1;
+$postag{"PART"} = 1;
+$postag{"PRON"} = 1;
+$postag{"PROPN"} = 1;
+$postag{"PUNCT"} = 1;
+$postag{"SCONJ"} = 1;
+$postag{"SYM"} = 1;
+$postag{"VERB"} = 1;
+$postag{"X"} = 1;
+
+while(<>){
+ ($form, $pos, $lemma, $morpho) = split /\t/;
+ if($postag{$pos}){
+ $h_form2pos{$form}{$pos} = 1;
+ $h_pos{$pos} += 1;
+ }
+}
+
+$nbelem = keys %h_form2pos;
+print "$nbelem\n";
+
+$nbelem = keys %h_pos;
+print "$nbelem\n";
+$first = 1;
+foreach $pos (keys %h_pos){
+ if($first){
+ $first = 0;
+ }
+ else{
+ print "\t";
+ }
+ print $pos;
+}
+print "\n";
+
+foreach $form (keys %h_form2pos){
+ print "$form\t";
+ foreach $pos (keys %h_pos){
+ if($h_form2pos{$form}{$pos}){
+ print "1";
+ }
+ else{
+ print "0";
+ }
+ }
+ print "\n";
+}