diff --git a/maca_common/include/mcd.h b/maca_common/include/mcd.h
index 1fd242cb23c8dee976a0e1fcbe29022e0e9f6b28..5c6747eba43a0cff072a031c0e412ace373d26bb 100644
--- a/maca_common/include/mcd.h
+++ b/maca_common/include/mcd.h
@@ -10,7 +10,7 @@
#define MCD_WF_NB 36
-#define MCD_WF_INDEX 0
+#define MCD_WF_ID 0
#define MCD_WF_FORM 1
#define MCD_WF_LEMMA 2
#define MCD_WF_CPOS 3
@@ -53,7 +53,7 @@
#define mcd_get_dico_label(m) (m)->dico_array[MCD_WF_LABEL]
-#define mcd_get_index_col(m) (m)->wf2col[MCD_WF_INDEX]
+#define mcd_get_index_col(m) (m)->wf2col[MCD_WF_ID]
#define mcd_get_form_col(m) (m)->wf2col[MCD_WF_FORM]
#define mcd_get_lemma_col(m) (m)->wf2col[MCD_WF_LEMMA]
#define mcd_get_cpos_col(m) (m)->wf2col[MCD_WF_CPOS]
diff --git a/maca_common/include/word.h b/maca_common/include/word.h
index 7606d4cf31bd94d38a7d087a8def69db4650fb0d..14d3c05a0461313c47f163a434f58a6c6f0fe28c 100644
--- a/maca_common/include/word.h
+++ b/maca_common/include/word.h
@@ -3,13 +3,13 @@
#include "mcd.h"
-#define word_get_index(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_INDEX]
+#define word_get_id(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_ID]
#define word_get_form(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_FORM]
#define word_get_lemma(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_LEMMA]
#define word_get_cpos(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_CPOS]
#define word_get_pos(w) ((w) == NULL) ? -1 : ((w)->wf_array[MCD_WF_POS])
#define word_get_feats(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_FEATS]
-#define word_get_gov(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_GOV]
+#define word_get_gov(w) ((w) == NULL) ? 0 : (w)->wf_array[MCD_WF_GOV]
#define word_get_label(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_LABEL]
#define word_get_stag(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_STAG]
#define word_get_sent_seg(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_SENT_SEG]
@@ -39,11 +39,12 @@
#define word_get_X(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_X]
#define word_get_Y(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_Y]
#define word_get_Z(w) ((w) == NULL) ? -1 : (w)->wf_array[MCD_WF_Z]
+#define word_get_input(w) ((w) == NULL) ? NULL : (w)->input
#define word_get_signature(w) ((w) == NULL) ? -1 : (w)->signature
#define word_get_U1(w) ((w) == NULL) ? -1 : (w)->U1
-#define word_get_relative_index(w) ((w) == NULL) ? -1 : (w)->relative_index
+#define word_get_index(w) ((w) == NULL) ? -1 : (w)->index
-#define word_set_index(w, val) (w)->wf_array[MCD_WF_INDEX] = (val)
+#define word_set_id(w, val) (w)->wf_array[MCD_WF_ID] = (val)
#define word_set_form(w, val) (w)->wf_array[MCD_WF_FORM] = (val)
#define word_set_lemma(w, val) (w)->wf_array[MCD_WF_LEMMA] = (val)
#define word_set_cpos(w, val) (w)->wf_array[MCD_WF_CPOS] = (val)
@@ -80,7 +81,7 @@
#define word_set_Y(w, val) (w)->wf_array[MCD_WF_Y] = (val)
#define word_set_Z(w, val) (w)->wf_array[MCD_WF_Z] = (val)
#define word_set_signature(w, val) (w)->signature = (val)
-#define word_set_relative_index(w, val) (w)->relative_index = (val)
+#define word_set_index(w, val) (w)->index = (val)
typedef struct _word {
int wf_array[MCD_WF_NB]; /* array containing the codes corresponding to the different word features */
@@ -89,7 +90,7 @@ typedef struct _word {
int signature; /* pos tags that this form can have (represented as a boolean string) */
int label;
char *form;
- int relative_index;
+ int index;
} word;
word *word_new(char *input);
@@ -104,6 +105,6 @@ word *word_read(FILE *f, mcd *mcd_struct);
word *word_parse_buffer(char *buffer, mcd *mcd_struct);
int word_is_eos(word *w, mcd *mcd_struct);
-int word_get_gov_relative_index(word *w);
+int word_get_gov_index(word *w);
#endif
diff --git a/maca_common/src/form2pos.c b/maca_common/src/form2pos.c
index ab86d127727f7197b2c6ea93469f74bd4003fb88..b2efb19294a1af2416128925693e2b6bc2e01a86 100644
--- a/maca_common/src/form2pos.c
+++ b/maca_common/src/form2pos.c
@@ -65,9 +65,9 @@ form2pos *form2pos_read(char *filename)
int form2pos_get_signature(form2pos *f2p, char *form)
{
- /* if(form == NULL)
+ if(form == NULL)
return -1;
- else*/
+ else
return hash_get_val(f2p->h_form2signature, form);
}
diff --git a/maca_common/src/mcd.c b/maca_common/src/mcd.c
index 64be8859569b5af09667538c9a46239ad8d7511c..9a2dc6d94d2afefe9b1b70389c6bf71bb844479d 100644
--- a/maca_common/src/mcd.c
+++ b/maca_common/src/mcd.c
@@ -231,11 +231,11 @@ mcd *mcd_read(char *mcd_filename, int verbose)
mcd *mcd_build_conll07(void)
{
mcd *m = mcd_new(8);
- m->wf[0]=MCD_WF_INDEX;
+ m->wf[0]=MCD_WF_ID;
m->wf_str[0]=strdup("INDEX");
m->representation[0]= MCD_REPRESENTATION_INT;
m->filename[0] = strdup("_");
- m->wf2col[MCD_WF_INDEX] = 0;
+ m->wf2col[MCD_WF_ID] = 0;
m->wf[1]=MCD_WF_FORM;
m->wf_str[1]=strdup("FORM");
@@ -332,11 +332,11 @@ mcd *mcd_build_ifpls(void)
{
mcd *m = mcd_new(6);
- m->wf[0]=MCD_WF_INDEX;
+ m->wf[0]=MCD_WF_ID;
m->wf_str[0]=strdup("INDEX");
m->representation[0]= MCD_REPRESENTATION_INT;
m->filename[0] = strdup("_");
- m->wf2col[MCD_WF_INDEX] = 0;
+ m->wf2col[MCD_WF_ID] = 0;
m->wf[1]=MCD_WF_FORM;
m->wf_str[1]=strdup("FORM");
@@ -389,7 +389,7 @@ dico_vec *mcd_build_dico_vec(mcd *mcd_struct)
int mcd_wf_code(char *wf)
{
- if(!strcmp(wf, "INDEX")) return MCD_WF_INDEX;
+ if(!strcmp(wf, "INDEX")) return MCD_WF_ID;
if(!strcmp(wf, "FORM")) return MCD_WF_FORM;
if(!strcmp(wf, "LEMMA")) return MCD_WF_LEMMA;
if(!strcmp(wf, "CPOS")) return MCD_WF_CPOS;
diff --git a/maca_common/src/sentence.c b/maca_common/src/sentence.c
index aaa5e1b587d77cbc97e9f65b23e6c75fe25e85af..a97486b02c565d33b0b87073e005d3ca0834c71e 100644
--- a/maca_common/src/sentence.c
+++ b/maca_common/src/sentence.c
@@ -48,7 +48,7 @@ void sentence_add_word(sentence *s, word *w)
s->length++;
s->words = (word **)realloc(s->words, s->length * sizeof(word *));
s->words[s->length -1] = w;
- word_set_relative_index(w, s->length -1);
+ word_set_index(w, s->length -1);
}
void sentence_free(sentence *s)
diff --git a/maca_common/src/word.c b/maca_common/src/word.c
index be7d2f8d26f9763fb82cc89df9fd9412a79d89f6..1a668e4421a96b40af11657254c764c5c894447d 100644
--- a/maca_common/src/word.c
+++ b/maca_common/src/word.c
@@ -20,7 +20,7 @@ word *word_new(char *input)
w->wf_array[MCD_WF_GOV] = 0;
w->form = NULL;
- w->relative_index = -1;
+ w->index = -1;
w->signature = -1;
return w;
}
@@ -102,8 +102,8 @@ word *word_create_dummy(mcd *mcd_struct)
word *w = word_new(NULL);
/* int type; */
- w->wf_array[MCD_WF_INDEX] = 0;
- w->relative_index = 0;
+ w->wf_array[MCD_WF_ID] = 0;
+ w->index = 0;
/* for(type = 1; type < MCD_WF_NB; type++)
w->wf_array[type] = -1;*/
/* if(mcd_struct->wf2col[type] != -1)
@@ -120,8 +120,8 @@ void word_print2(FILE *f, word *w)
printf("form = %d\t", word_get_form(w));
printf("lemma = %d\t", word_get_lemma(w));
printf("pos = %d\t", word_get_pos(w));
- printf("index = %d\t", word_get_index(w));
- printf("rel index = %d\n", word_get_relative_index(w));
+ printf("index = %d\t", word_get_id(w));
+ printf("rel index = %d\n", word_get_index(w));
}
void word_print(FILE *f, word *w)
@@ -140,16 +140,10 @@ int word_is_eos(word *w, mcd *mcd_struct)
return word_get_sent_seg(w);
}
-int word_get_gov_relative_index(word *w)
+int word_get_gov_index(word *w)
{
int index;
if(word_get_gov(w) == 0) return -1;
-
- /* printf("in word_get_gov_rel_index(%d)\n", word_get_relative_index(w));
- printf("gov = %d\n", word_get_gov(w)); */
-
- index = (word_get_relative_index(w)) + (word_get_gov(w));
- /* printf("index = %d\n", index); */
- return index;
- /* return word_get_relative_index(w) + word_get_gov(w); */
+ index = (word_get_index(w)) + (word_get_gov(w));
+ return index;
}
diff --git a/maca_common/src/word_buffer.c b/maca_common/src/word_buffer.c
index aabe1876898902cf68621439b6b14c3ef0173a31..a22ae9edcee304521f76dce0ef04ac52e7785a54 100644
--- a/maca_common/src/word_buffer.c
+++ b/maca_common/src/word_buffer.c
@@ -49,19 +49,19 @@ void word_buffer_print_compact(FILE *f, word_buffer *wb)
{
word *w;
w = word_buffer_bm3(wb);
- if(w){ fprintf(f, "%d:%s ", word_get_relative_index(w), w->form);}
+ if(w){ fprintf(f, "%d:%s ", word_get_index(w), w->form);}
w = word_buffer_bm2(wb);
- if(w){ fprintf(f, "%d:%s ", word_get_relative_index(w), w->form);}
+ if(w){ fprintf(f, "%d:%s ", word_get_index(w), w->form);}
w = word_buffer_bm1(wb);
- if(w){ fprintf(f, "%d:%s ", word_get_relative_index(w), w->form);}
+ if(w){ fprintf(f, "%d:%s ", word_get_index(w), w->form);}
w = word_buffer_b0(wb);
- if(w){ fprintf(f, "[%d:%s] ", word_get_relative_index(w), w->form);}
+ if(w){ fprintf(f, "[%d:%s] ", word_get_index(w), w->form);}
w = word_buffer_b1(wb);
- if(w){ fprintf(f, "%d:%s ", word_get_relative_index(w), w->form);}
+ if(w){ fprintf(f, "%d:%s ", word_get_index(w), w->form);}
w = word_buffer_b2(wb);
- if(w){ fprintf(f, "%d:%s ", word_get_relative_index(w), w->form);}
+ if(w){ fprintf(f, "%d:%s ", word_get_index(w), w->form);}
w = word_buffer_b3(wb);
- if(w){ fprintf(f, "%d:%s ", word_get_relative_index(w), w->form);}
+ if(w){ fprintf(f, "%d:%s ", word_get_index(w), w->form);}
fprintf(f, "\n");
}
@@ -83,7 +83,7 @@ int word_buffer_add(word_buffer *wb, word *w)
wb->array = (word **)realloc(wb->array, wb->size * sizeof(word *));
}
wb->array[wb->nbelem] = w;
- word_set_relative_index(w, wb->nbelem);
+ word_set_index(w, wb->nbelem);
wb->nbelem++;
return wb->nbelem - 1;
@@ -116,6 +116,7 @@ int word_buffer_read_next_word(word_buffer *wb)
word *w = NULL;
w = word_read(wb->input_file, wb->mcd_struct);
+ /* if((w) && (w->input)) printf("## %s\n", w->input); */
if(w == NULL) return -1;
word_buffer_add(wb, w);
return wb->nbelem - 1;
@@ -141,6 +142,8 @@ int word_buffer_move_left(word_buffer *wb)
int word_buffer_end(word_buffer *wb)
{
+
+ /* printf("in word_buffer_end current index = %d nb elem = %d\n", wb->current_index, wb->nbelem); */
return (wb->current_index >= wb->nbelem)? 1 : 0;
}
@@ -165,7 +168,7 @@ int word_buffer_read_sentence(word_buffer *wb)
/* fprintf(stderr, "%s", buffer); */
if((buffer[0] == '\n') || (buffer[0] == ' ') || (buffer[0] == '\t')) break; /* end of the sentence */
w = word_parse_buffer(buffer, word_buffer_get_mcd(wb));
- word_set_relative_index(w, index);
+ word_set_index(w, index);
index++;
word_buffer_add(wb, w);
if(word_is_eos(w, word_buffer_get_mcd(wb))) break;
diff --git a/maca_trans_parser/src/config.c b/maca_trans_parser/src/config.c
index 8e809a8b0b6b6ac1d85a232f97b6abb8a6245b42..68e15dd464e277074117c0489c45ca56c34b379c 100644
--- a/maca_trans_parser/src/config.c
+++ b/maca_trans_parser/src/config.c
@@ -30,7 +30,7 @@ config *config_new(FILE *f, mcd *mcd_struct, int lookahead)
w = word_read(c->f, c->mcd_struct);
if(w == NULL) return NULL;
- word_set_relative_index(w, c->current_index);
+ word_set_index(w, c->current_index);
c->current_index++;
queue_add(c->bf, w);
return w;
@@ -48,7 +48,7 @@ void config_free(config *c)
int config_is_terminal(config *c)
{
- return word_buffer_is_last(c->bf);
+ return word_buffer_end(c->bf);
}
config *config_initial(FILE *f, mcd *mcd_struct, int lookahead)
@@ -102,12 +102,12 @@ void config_add_mvt(config *c, int mvt)
void config_print(FILE *f, config *c)
{
- word *b0 = NULL;
- word *s0 = NULL;
+ /* word *b0 = NULL; */
+ /* word *s0 = NULL; */
if(c){
if(!stack_is_empty(c->st))
- s0 = stack_elt_n(c->st, 0);
- b0 = word_buffer_b0(c->bf);
+ /* s0 = stack_elt_n(c->st, 0); */
+ /* b0 = word_buffer_b0(c->bf); */
/* if(s0) { printf("s0 = "); word_print2(stdout, s0);} */
/* if(b0) { printf("b0 = "); word_print2(stdout, b0);} */
diff --git a/maca_trans_parser/src/depset.c b/maca_trans_parser/src/depset.c
index 3ce6b4776925c1e2320b651adff6193ff7361b06..3298f7e419bbd05f78377d1608263db3a4f499dd 100644
--- a/maca_trans_parser/src/depset.c
+++ b/maca_trans_parser/src/depset.c
@@ -44,10 +44,10 @@ void depset_add(depset *d, word *gov, int label, word *dep)
int new_length;
if(gov == NULL || dep == NULL) return;
- word *max = (word_get_relative_index(gov) > word_get_relative_index(dep)) ? gov : dep;
+ word *max = (word_get_index(gov) > word_get_index(dep)) ? gov : dep;
- if(word_get_relative_index(max) >= d->length){
- new_length = word_get_relative_index(max) + 1;
+ if(word_get_index(max) >= d->length){
+ new_length = word_get_index(max) + 1;
d->array = (dependency *)realloc(d->array, new_length * sizeof(dependency));
for(i=d->length; i < new_length; i++){
d->array[i].gov = NULL;
@@ -56,9 +56,9 @@ void depset_add(depset *d, word *gov, int label, word *dep)
}
d->length = new_length;
}
- d->array[word_get_relative_index(dep)].gov = gov;
- d->array[word_get_relative_index(dep)].dep = dep;
- d->array[word_get_relative_index(dep)].label = label;
+ d->array[word_get_index(dep)].gov = gov;
+ d->array[word_get_index(dep)].dep = dep;
+ d->array[word_get_index(dep)].label = label;
}
void depset_print(FILE *f, depset *d)
@@ -66,7 +66,7 @@ void depset_print(FILE *f, depset *d)
int i;
for(i=0; i < d->length; i++){
if((d->array[i].gov) && (d->array[i].dep))
- fprintf(f, "(%d, %d, %d) ", word_get_relative_index(d->array[i].dep), d->array[i].label, word_get_relative_index(d->array[i].gov));
+ fprintf(f, "(%d, %d, %d) ", word_get_index(d->array[i].dep), d->array[i].label, word_get_index(d->array[i].gov));
}
fprintf(f, "\n");
}
@@ -78,7 +78,7 @@ void depset_print2(FILE *f, depset *d, dico *dico_labels)
char *label;
for(i=1; i < d->length; i++){
if((d->array[i].gov) && (d->array[i].dep)){
- distance = word_get_relative_index(d->array[i].gov) - word_get_relative_index(d->array[i].dep);
+ distance = word_get_index(d->array[i].gov) - word_get_index(d->array[i].dep);
/* fprintf(f, "%s\t%d\t%s\n", d->array[i].dep->input, distance, dico_int2string(dico_labels, d->array[i].label)); */
label = dico_int2string(dico_labels, d->array[i].label);
@@ -103,10 +103,10 @@ void depset_print3(FILE *f, depset *d, dico *dico_labels)
for(i=1; i < d->length; i++){
if((d->array[i].gov) && (d->array[i].dep)){
if(d->array[i].label == root_code)
- fprintf(f, "%d\t%s\t%d\t%s\n", word_get_relative_index(d->array[i].dep), d->array[i].dep->input, 0, dico_int2string(dico_labels, d->array[i].label));
+ fprintf(f, "%d\t%s\t%d\t%s\n", word_get_index(d->array[i].dep), d->array[i].dep->input, 0, dico_int2string(dico_labels, d->array[i].label));
else{
- distance = word_get_relative_index(d->array[i].gov) - word_get_relative_index(d->array[i].dep);
- fprintf(f, "%d\t%s\t%d\t%s\n", word_get_relative_index(d->array[i].dep), d->array[i].dep->input, distance, dico_int2string(dico_labels, d->array[i].label));
+ distance = word_get_index(d->array[i].gov) - word_get_index(d->array[i].dep);
+ fprintf(f, "%d\t%s\t%d\t%s\n", word_get_index(d->array[i].dep), d->array[i].dep->input, distance, dico_int2string(dico_labels, d->array[i].label));
}
}
}
@@ -129,9 +129,9 @@ void depset_print_new_index(FILE *f, depset *d, dico *dico_labels)
for(i=1; i < d->length; i++){
if((d->array[i].gov) && (d->array[i].dep)){
- /* fprintf(f, "%d\t", word_get_relative_index(d->array[i].dep)); */
- fprintf(f, "%d\t", word_get_relative_index(d->array[i].dep));
- fprintf(f, "%s\t%d\t%s\n", skip_index(d->array[i].dep->input), word_get_relative_index(d->array[i].gov), dico_int2string(dico_labels, d->array[i].label));
+ /* fprintf(f, "%d\t", word_get_index(d->array[i].dep)); */
+ fprintf(f, "%d\t", word_get_index(d->array[i].dep));
+ fprintf(f, "%s\t%d\t%s\n", skip_index(d->array[i].dep->input), word_get_index(d->array[i].gov), dico_int2string(dico_labels, d->array[i].label));
}
}
fprintf(f, "\n");
@@ -148,8 +148,8 @@ int depset_compare(depset *d1, depset *d2)
if(d1->length != d2->length){ fprintf(stdout, "fail\n"); return 0;}
for(i=0; i < d1->length; i++){
for(j=0; j < d2->length; j++){
- if((word_get_relative_index(d1->array[i].gov) == word_get_relative_index(d2->array[j].gov))
- && (word_get_relative_index(d1->array[i].dep) == word_get_relative_index(d2->array[j].dep))
+ if((word_get_index(d1->array[i].gov) == word_get_index(d2->array[j].gov))
+ && (word_get_index(d1->array[i].dep) == word_get_index(d2->array[j].dep))
&& (d1->array[i].label == d2->array[j].label)) break;
}
if(j == d2->length){
diff --git a/maca_trans_parser/src/feat_fct.c b/maca_trans_parser/src/feat_fct.c
index c37a01b0f43e84679ecbcee31c49a041f0f089d0..5740caadc35a99d1b09714354b3179343d3d4fc6 100644
--- a/maca_trans_parser/src/feat_fct.c
+++ b/maca_trans_parser/src/feat_fct.c
@@ -8,6 +8,9 @@
/* word features */
/* words in the stack */
+int s0g(config *c) {return word_get_gov(stack_s0(config_get_stack(c)));}
+int s0sf(config *c) {return word_get_label(stack_s0(config_get_stack(c)));}
+
int s0f(config *c) {return word_get_form(stack_s0(config_get_stack(c)));}
int s0l(config *c) {return word_get_lemma(stack_s0(config_get_stack(c)));}
int s0c(config *c) {return word_get_cpos(stack_s0(config_get_stack(c)));}
@@ -44,6 +47,9 @@ int s0Z(config *c) {return word_get_Z(stack_s0(config_get_stack(c)));}
int s0U1(config *c) {return word_get_U1(stack_s0(config_get_stack(c)));}
int s0sgn(config *c) {return word_get_signature(stack_s0(config_get_stack(c)));}
+int s1g(config *c) {return word_get_gov(stack_s1(config_get_stack(c)));}
+int s1sf(config *c) {return word_get_label(stack_s1(config_get_stack(c)));}
+
int s1f(config *c) {return word_get_form(stack_s1(config_get_stack(c)));}
int s1l(config *c) {return word_get_lemma(stack_s1(config_get_stack(c)));}
int s1c(config *c) {return word_get_cpos(stack_s1(config_get_stack(c)));}
@@ -400,8 +406,8 @@ int ldep_s0r(config *c){
int i;
if(top){
- if(word_get_relative_index(top) >= c->ds->length) return -1;
- for(i=word_get_relative_index(top); i > 0; i--)
+ if(word_get_index(top) >= c->ds->length) return -1;
+ for(i=word_get_index(top); i > 0; i--)
if(c->ds->array[i].gov == top)
return i;
}
@@ -422,7 +428,7 @@ int rdep_s0r(config *c){
int i;
if(top)
- for(i=word_get_relative_index(top); i < c->ds->length; i++)
+ for(i=word_get_index(top); i < c->ds->length; i++)
if(c->ds->array[i].gov == top)
return i;
return -1;
@@ -442,8 +448,8 @@ int ldep_b0r(config *c){
int i;
if(top){
- if(word_get_relative_index(top) >= c->ds->length) return -1;
- for(i=word_get_relative_index(top); i > 0; i--)
+ if(word_get_index(top) >= c->ds->length) return -1;
+ for(i=word_get_index(top); i > 0; i--)
if(c->ds->array[i].gov == top)
return i;
}
@@ -468,7 +474,7 @@ int rdep_b0r(config *c){
int i;
if(top)
- for(i=word_get_relative_index(top); i < c->ds->length; i++)
+ for(i=word_get_index(top); i < c->ds->length; i++)
if(c->ds->array[i].gov == top)
return i;
return -1;
@@ -545,7 +551,7 @@ int dist_s0_b0(config *c){
if(stack_is_empty(c->st) || word_buffer_is_empty(c->bf))
return 0;
- dist = word_get_relative_index(word_buffer_b0(c->bf)) - word_get_relative_index(stack_top(c->st));
+ dist = word_get_index(word_buffer_b0(c->bf)) - word_get_index(stack_top(c->st));
return (abs(dist) > 6)? 6 : dist;
}
diff --git a/maca_trans_parser/src/feat_fct.h b/maca_trans_parser/src/feat_fct.h
index 84f8b2adacab1ec71540ef6fde58490b794373e3..6f91a3afc8d1e91aae837918ec2fee505f3dbc80 100644
--- a/maca_trans_parser/src/feat_fct.h
+++ b/maca_trans_parser/src/feat_fct.h
@@ -6,6 +6,9 @@
typedef int (*feat_fct) (config *c);
/* word features */
+int s0g(config *c);
+int s0sf(config *c);
+
int s0f(config *c);
int s0l(config *c);
int s0c(config *c);
@@ -44,6 +47,10 @@ int s0sgn(config *c);
int s0r(config *c);
+int s1g(config *c);
+int s1sf(config *c);
+
+
int s1f(config *c);
int s1l(config *c);
int s1c(config *c);
diff --git a/maca_trans_parser/src/feat_lib.c b/maca_trans_parser/src/feat_lib.c
index 1567bd72904f9042a2d129df18f396109926e8c7..6a2293f5327804028f34d66605af322ffaadeb14 100644
--- a/maca_trans_parser/src/feat_lib.c
+++ b/maca_trans_parser/src/feat_lib.c
@@ -33,6 +33,9 @@ feat_lib *feat_lib_build(void)
{
feat_lib *fl = feat_lib_new();
+ feat_lib_add(fl, FEAT_TYPE_LABEL, (char *)"s0sf", s0sf);
+ feat_lib_add(fl, FEAT_TYPE_INT , (char *)"s0g", s0g);
+
feat_lib_add(fl, FEAT_TYPE_FORM, (char *)"s0f", s0f);
feat_lib_add(fl, FEAT_TYPE_LEMMA, (char *)"s0l", s0l);
feat_lib_add(fl, FEAT_TYPE_CPOS, (char *)"s0c", s0c);
@@ -69,6 +72,9 @@ feat_lib *feat_lib_build(void)
feat_lib_add(fl, FEAT_TYPE_INT_3, (char *)"s0U1", s0U1);
feat_lib_add(fl, FEAT_TYPE_INT, (char *)"s0sgn", s0sgn);
+ feat_lib_add(fl, FEAT_TYPE_FORM, (char *)"s1g", s1g);
+ feat_lib_add(fl, FEAT_TYPE_FORM, (char *)"s1sf", s1sf);
+
feat_lib_add(fl, FEAT_TYPE_FORM, (char *)"s1f", s1f);
feat_lib_add(fl, FEAT_TYPE_LEMMA, (char *)"s1l", s1l);
feat_lib_add(fl, FEAT_TYPE_CPOS, (char *)"s1c", s1c);
diff --git a/maca_trans_parser/src/maca_trans_parser_mcf2cff.c b/maca_trans_parser/src/maca_trans_parser_mcf2cff.c
index 645307c5596c17747d6d2bd7574f1cdd9279a4a4..77f2ab64664ecae7eabdfd722fc30f3b4773a1d4 100644
--- a/maca_trans_parser/src/maca_trans_parser_mcf2cff.c
+++ b/maca_trans_parser/src/maca_trans_parser_mcf2cff.c
@@ -54,38 +54,39 @@ void generate_training_file_stream(FILE *output_file, context *ctx)
int eos_label = dico_string2int(ctx->dico_labels, "eos");
word_buffer *ref = word_buffer_load_mcf(ctx->input_filename, ctx->mcd_struct);
FILE *mcf_file = myfopen(ctx->input_filename, "r");
-
+ int start_sentence_index = 1;
+
/* create an mcd that corresponds to ctx->mcd_struct, but without gov and label */
+ /* the idea is to ignore syntax in the mcf file that will be read */
+ /* it is ugly !!! */
+
mcd *mcd_struct_hyp = mcd_copy(ctx->mcd_struct);
mcd_remove_wf_column(mcd_struct_hyp, MCD_WF_GOV);
mcd_remove_wf_column(mcd_struct_hyp, MCD_WF_LABEL);
c = config_initial(mcf_file, mcd_struct_hyp, 5);
-
+
while(!word_buffer_end(ref)){
- /* printf("************ REF ************\n");
+ /*printf("************ REF ************\n");
word_buffer_print(stdout, ref);
- printf("*****************************\n");*/
+ printf("*****************************\n");*/
- printf("*****************************\n");
- config_print(stdout,c);
+ /* printf("*****************************\n"); */
config2feat_vec_cff(ctx->features_model, c, ctx->d_perceptron_features, fv, ctx->mode);
/* feat_vec_print(stdout, fv); */
- mvt_code = oracle_parser_arc_eager(c, ref);
+ mvt_code = oracle_parser_arc_eager(c, ref, start_sentence_index);
mvt_type = movement_type(mvt_code);
mvt_label = movement_label(mvt_code);
-
- /* printf("mvt code = %d\n", mvt_code);
- printf("mvt type = %d\n", mvt_type); */
- movement_print(stdout, mvt_code, ctx->dico_labels);
+
+ /* config_print(stdout,c); */
+ /* movement_print(stdout, mvt_code, ctx->dico_labels); */
fprintf(output_file, "%d", mvt_code);
feat_vec_print(output_file, fv);
-
if(mvt_type == MVT_LEFT){
movement_left_arc(c, mvt_label, 0);
continue;
@@ -94,35 +95,21 @@ void generate_training_file_stream(FILE *output_file, context *ctx)
if(mvt_type == MVT_RIGHT){
movement_right_arc(c, mvt_label, 0);
word_buffer_move_right(ref);
-
if((mvt_label == eos_label)){ /* sentence is complete */
- printf("sentence complete\n");
-
- while(movement_reduce(c,0)){
- printf("reduce\n");
- printf("*****************************\n");
- config_print(stdout,c);
-
- }
-
- /* pop eos from stack */
- /* stack_pop(config_get_stack(c)); */
- /* pop root from stack */
- /* stack_pop(config_get_stack(c)); */
-
- config_print(stdout,c);
+ sentence_nb++;
+ start_sentence_index = word_get_index(word_buffer_b0(config_get_buffer(c))) - 1;
+ /* printf("%d\n", start_sentence_index); */
- printf("ref current index = %d\n", word_buffer_get_current_index(ref));
+ /* printf("*****************************\n"); */
+ /* config_print(stdout,c); */
if(word_buffer_is_last(ref)){
- printf("it is the end\n");
+ /* printf("it is the end\n"); */
break;
}
-
- /* change index of dummy word */
- word_set_relative_index(stack_top(config_get_stack(c)), word_get_relative_index(word_buffer_b0(config_get_buffer(c))) - 1);
- }
+ }
continue;
}
+
if(mvt_type == MVT_REDUCE){
movement_reduce(c, 0);
continue;
@@ -133,61 +120,7 @@ void generate_training_file_stream(FILE *output_file, context *ctx)
continue;
}
}
-}
-
-
-#if 0
-void generate_training_file_buffer(FILE *output_file, context *ctx)
-{
- config *c;
- int mvt_code;
- char mvt_type;
- int mvt_label;
- feat_vec *fv = feat_vec_new(feature_types_nb);
- sentence *ref = NULL;
- int sentence_nb = 0;
- FILE *conll_file = myfopen(ctx->input_filename, "r");
- FILE *conll_file_ref = myfopen(ctx->input_filename, "r");
-
- c = config_initial(conll_file, ctx->mcd_struct, 0);
-
- while((ref = sentence_read(conll_file_ref, ctx->mcd_struct)) && (sentence_nb < ctx->sent_nb)){
- /* sentence_print(stdout, ref, NULL); */
- word_buffer_read_sentence(c->bf);
- while(!config_is_terminal(c)){
- /* config_print(stdout,c); */
-
- config2feat_vec_cff(ctx->features_model, c, ctx->d_perceptron_features, fv, ctx->mode);
-
- mvt_code = oracle_parser(c, ref);
-
- mvt_type = movement_type(mvt_code);
- mvt_label = movement_label(mvt_code);
-
- /* printf("mvt type = %d mvt label = %d\n", mvt_type, mvt_label); */
-
- fprintf(output_file, "%d", mvt_code);
- feat_vec_print(output_file, fv);
-
- if(mvt_type == MVT_LEFT){
- movement_left_arc(c, mvt_label, 0);
- continue;
- }
- if(mvt_type == MVT_RIGHT){
- movement_right_arc(c, mvt_label, 0);
- continue;
- }
- if(mvt_type == MVT_SHIFT){
- movement_shift(c, 0, 0);
- continue;
- }
- }
- config_free(c);
- c = config_initial(conll_file, ctx->mcd_struct, 0);
- sentence_nb++;
- }
}
-#endif
int main(int argc, char *argv[])
{
@@ -238,16 +171,9 @@ int main(int argc, char *argv[])
output_file = myfopen(ctx->cff_filename, "w");
else
output_file = stdout;
-
-
- generate_training_file_stream(output_file, ctx);
-
- /*
- if(ctx->stream_mode)
- generate_training_file_stream(output_file, ctx);
- else
- generate_training_file_buffer(output_file, ctx);
- */
+
+ generate_training_file_stream(output_file, ctx);
+
if(ctx->mode == TRAIN_MODE){
/* dico_print(ctx->perceptron_features_filename, ctx->d_perceptron_features); */
dico_vec_print(ctx->vocabs_filename, ctx->vocabs);
diff --git a/maca_trans_parser/src/maca_trans_tagger_mcf2cff.c b/maca_trans_parser/src/maca_trans_tagger_mcf2cff.c
index 3dadf7bbb07deb0a9d498b45b18bebf959b40552..6936c6399c713da664876b45ec181e39ee043bb2 100644
--- a/maca_trans_parser/src/maca_trans_tagger_mcf2cff.c
+++ b/maca_trans_parser/src/maca_trans_tagger_mcf2cff.c
@@ -87,7 +87,8 @@ void generate_training_file_stream(FILE *output_file, context *ctx)
postag = oracle_tagger(c, NULL);
fprintf(output_file, "%d", postag);
feat_vec_print(output_file, fv);
- movement_tagger(c, postag, 0, 1);
+ int res = movement_tagger(c, postag, 0, 1);
+ if(res == 0) break;
}
}
diff --git a/maca_trans_parser/src/movement_parser.c b/maca_trans_parser/src/movement_parser.c
index 254cab39d47f915545df5cb42b183986191b2d7a..d53556822e141778d4a7908aa5e4b40e2c0b95ab 100644
--- a/maca_trans_parser/src/movement_parser.c
+++ b/maca_trans_parser/src/movement_parser.c
@@ -37,10 +37,10 @@ int movement_left_arc(config *c, int label, float score)
{
if(stack_is_empty(c->st)) return 0;
if(word_buffer_is_empty(c->bf)) return 0;
- if(word_get_relative_index(stack_top(c->st)) == 0) return 0; /* the dummy word cannot be a dependent */
+ if(word_get_index(stack_top(c->st)) == 0) return 0; /* the dummy word cannot be a dependent */
/* create a new dependency */
- word_set_gov(stack_top(c->st), word_get_relative_index(word_buffer_b0(c->bf)));
+ word_set_gov(stack_top(c->st), word_get_index(word_buffer_b0(c->bf)));
word_set_label(stack_top(c->st), label);
/* depset_add(c->ds, word_buffer_b0(c->bf), label, stack_top(c->st)); */
@@ -57,7 +57,7 @@ int movement_right_arc(config *c, int label, float score)
/* create a new dependency */
- word_set_gov(word_buffer_b0(c->bf), word_get_relative_index(stack_top(c->st)));
+ word_set_gov(word_buffer_b0(c->bf), word_get_index(stack_top(c->st)));
word_set_label(word_buffer_b0(c->bf), label);
/* depset_add(c->ds, stack_top(c->st), label, word_buffer_b0(c->bf)); */
@@ -96,7 +96,7 @@ config *movement_left_arc_dup(config *c, int label, float score, feat_vec *fv)
config *copy = NULL;
if(stack_is_empty(c->st)) return NULL;
if(word_buffer_is_empty(c->bf)) return NULL;
- if(word_get_relative_index(stack_top(c->st)) == 0) return NULL;
+ if(word_get_index(stack_top(c->st)) == 0) return NULL;
copy = config_copy(c);
depset_add(copy->ds, word_buffer_b0(copy->bf), label, stack_top(copy->st));
diff --git a/maca_trans_parser/src/movement_parser_arc_eager.c b/maca_trans_parser/src/movement_parser_arc_eager.c
index 56beba1d06b79bd3b91f19a8cbfa3125b630e9b1..0481cb0296e7ba7ef39d400d1d5e22a6eabe41e6 100644
--- a/maca_trans_parser/src/movement_parser_arc_eager.c
+++ b/maca_trans_parser/src/movement_parser_arc_eager.c
@@ -35,29 +35,22 @@ int movement_label(int mvt)
int movement_left_arc(config *c, int label, float score)
{
-
- if(stack_height(c->st) < 2) return 0; /* the dummy word cannot be a dependent */
- if(word_buffer_is_empty(c->bf)) return 0;
-
- /* word on top of the stack should not have a governor */
- /* printf("index word top of stack = %d\n", word_get_relative_index(stack_top(c->st))); */
- /* printf("word_get_gov_relative_index(stack_top(c->st)) = %d\n", word_get_gov_relative_index(stack_top(c->st))); */
-
- if(word_get_gov_relative_index(stack_top(c->st)) != -1) return 0;
-
- word *gov = word_buffer_b0(c->bf);
- word *dep = stack_top(c->st);
- int dist = (word_get_relative_index(gov)) - (word_get_relative_index(dep));
+ if(stack_height(config_get_stack(c)) < 2) return 0; /* the dummy word cannot be a dependent */
+ if(word_buffer_is_empty(config_get_buffer(c))) return 0;
+ /* word on top of the stack should not have a governor */
+ if(word_get_gov(stack_top(config_get_stack(c))) != 0) return 0;
- printf("create left arc %d <- %d dist = %d\n", word_get_relative_index(dep), word_get_relative_index(gov), dist);
+ word *gov = word_buffer_b0(config_get_buffer(c));
+ word *dep = stack_top(config_get_stack(c));
+ int dist = (word_get_index(gov)) - (word_get_index(dep));
+ /* printf("create left arc %d <- %d dist = %d\n", word_get_index(dep), word_get_index(gov), dist); */
/* create a new dependency */
word_set_gov(dep, dist);
word_set_label(dep, label);
- /* depset_add(c->ds, word_buffer_b0(c->bf), label, stack_top(c->st)); */
- stack_pop(c->st);
+ stack_pop(config_get_stack(c));
config_add_mvt(c, movement_left_code(label));
return 1;
}
@@ -65,45 +58,43 @@ int movement_left_arc(config *c, int label, float score)
int movement_right_arc(config *c, int label, float score)
{
/* printf("RA "); */
- if(stack_is_empty(c->st)) return 0;
- if(word_buffer_is_empty(c->bf)) return 0;
+ if(stack_is_empty(config_get_stack(c))) return 0;
+ if(word_buffer_is_empty(config_get_buffer(c))) return 0;
- word *gov = stack_top(c->st);
- word *dep = word_buffer_b0(c->bf);
- int dist = (word_get_relative_index(gov)) - (word_get_relative_index(dep));
+ word *gov = stack_top(config_get_stack(c));
+ word *dep = word_buffer_b0(config_get_buffer(c));
+ int dist = (word_get_index(gov)) - (word_get_index(dep));
- printf("create right arc %d -> %d dist = %d\n", word_get_relative_index(gov), word_get_relative_index(dep), dist);
+ /* printf("create right arc %d -> %d dist = %d\n", word_get_index(gov), word_get_index(dep), dist); */
/* create a new dependency */
word_set_gov(dep, dist);
word_set_label(dep, label);
- stack_push(c->st, word_buffer_b0(c->bf));
- word_buffer_move_right(c->bf);
+ stack_push(config_get_stack(c), word_buffer_b0(config_get_buffer(c)));
+ word_buffer_move_right(config_get_buffer(c));
config_add_mvt(c, movement_right_code(label));
return 1;
}
-
int movement_shift(config *c, int stream, float score)
{
- if(word_buffer_is_empty(c->bf)) return 0;
- /* printf("SH\n"); */
-
- stack_push(c->st, word_buffer_b0(c->bf));
- word_buffer_move_right(c->bf);
-
+ if(word_buffer_is_empty(config_get_buffer(c))) return 0;
+ stack_push(config_get_stack(c), word_buffer_b0(config_get_buffer(c)));
+ word_buffer_move_right(config_get_buffer(c));
config_add_mvt(c, MVT_SHIFT);
return 1;
}
int movement_reduce(config *c, float score)
{
- if(stack_is_empty(c->st)) return 0;
- if(word_get_gov(stack_top(c->st)) == 0) return 0; /* word on top of stack does not have a governor */
- stack_pop(c->st);
+ if(stack_is_empty(config_get_stack(c))) return 0;
+ /* word on top of stack must have a governor */
+ if(word_get_gov(stack_top(config_get_stack(c))) == 0) return 0;
+ stack_pop(config_get_stack(c));
+ config_add_mvt(c, MVT_REDUCE);
return 1;
}
diff --git a/maca_trans_parser/src/movement_tagger.c b/maca_trans_parser/src/movement_tagger.c
index 3ac695a1e63c5336b9745560c04e985fb5001f5d..5d33788a4b0fe791cdda53e07d30cd22c71d1fc6 100644
--- a/maca_trans_parser/src/movement_tagger.c
+++ b/maca_trans_parser/src/movement_tagger.c
@@ -6,7 +6,7 @@
int movement_tagger(config *c, int postag, float score, int stream)
{
- if(word_buffer_is_empty(c->bf)) return 0;
+ if(word_buffer_is_last(c->bf)) return 0;
word_set_pos(word_buffer_b0(c->bf), postag);
word_buffer_move_right(c->bf);
diff --git a/maca_trans_parser/src/oracle_parser.c b/maca_trans_parser/src/oracle_parser.c
index 418f422bb4500ab6bf642300fe9b6c4d809f1180..6f422d9352474b073beae348254e801dd3654a56 100644
--- a/maca_trans_parser/src/oracle_parser.c
+++ b/maca_trans_parser/src/oracle_parser.c
@@ -9,11 +9,11 @@ int check_all_dependents_of_word_in_ref_are_in_hyp(config *c, sentence *ref, int
int dep;
for(dep=0; dep < ref->length; dep++){
- if(word_get_gov_relative_index(ref->words[dep]) == word_index){ /* found a dependent of word in ref */
+ if(word_get_gov_index(ref->words[dep]) == word_index){ /* found a dependent of word in ref */
/* look for a dependency in hyp such that its dependent is dep */
if((dep >= c->ds->length)
|| (c->ds->array[dep].gov == NULL)
- || (word_get_relative_index(c->ds->array[dep].gov) != word_index)
+ || (word_get_index(c->ds->array[dep].gov) != word_index)
|| (c->ds->array[dep].label != word_get_label(ref->words[dep])))
return 0;
}
@@ -29,22 +29,22 @@ int oracle_parser(config *c, sentence *ref)
if(!stack_is_empty(c->st) && !word_buffer_is_empty(c->bf)){
s0 = stack_top(c->st);
- s0_index = word_get_relative_index(s0);
+ s0_index = word_get_index(s0);
b0 = word_buffer_b0(c->bf);
- b0_index = word_get_relative_index(b0);
+ b0_index = word_get_index(b0);
/* printf("s0 = %d b0 = %d\n", s0_index, b0_index); */
- /*printf("dans ref gov de %d = %d\n", s0_index, word_get_gov_relative_index(ref->words[s0_index]));
- printf("dans ref gov de %d = %d\n", b0_index, word_get_gov_relative_index(ref->words[b0_index])); */
+ /*printf("dans ref gov de %d = %d\n", s0_index, word_get_gov_index(ref->words[s0_index]));
+ printf("dans ref gov de %d = %d\n", b0_index, word_get_gov_index(ref->words[b0_index])); */
/* LEFT ARC b0 is the governor and s0 the dependent */
- if(word_get_gov_relative_index(ref->words[s0_index]) == b0_index)
+ if(word_get_gov_index(ref->words[s0_index]) == b0_index)
return movement_left_code(word_get_label(ref->words[s0_index]));
/* RIGHT ARC s0 is the governor and b0 the dependent */
- if((word_get_gov_relative_index(ref->words[b0_index]) == s0_index)
+ if((word_get_gov_index(ref->words[b0_index]) == s0_index)
&& check_all_dependents_of_word_in_ref_are_in_hyp(c, ref, b0_index)){
return movement_right_code(word_get_label(ref->words[b0_index]));
}
diff --git a/maca_trans_parser/src/oracle_parser_arc_eager.c b/maca_trans_parser/src/oracle_parser_arc_eager.c
index a24442bf116d7df7ed426a1f45ae4a091ae565d6..082130ad311cbdda948874d56d89545170689b8e 100644
--- a/maca_trans_parser/src/oracle_parser_arc_eager.c
+++ b/maca_trans_parser/src/oracle_parser_arc_eager.c
@@ -4,25 +4,25 @@
#include"word_buffer.h"
#include"movement_parser_arc_eager.h"
-int check_all_dependents_of_word_in_ref_are_in_hyp(config *c, word_buffer *ref, int word_index)
+int check_all_dependents_of_word_in_ref_are_in_hyp(config *c, word_buffer *ref, int word_index, int start_sentence_index)
{
int dep;
int gov_ref;
int gov_hyp;
-
- for(dep=1; dep < ref->nbelem; dep++){
- gov_ref = word_get_gov_relative_index(word_buffer_get_word_n(ref, dep));
+ int max = ((start_sentence_index + 500) > ref->nbelem)? ref->nbelem : (start_sentence_index + 500);
+ for(dep=start_sentence_index; dep < max; dep++){
+ gov_ref = word_get_gov_index(word_buffer_get_word_n(ref, dep));
if(gov_ref == word_index){ /* found a dependent of word in ref */
/* look for a dependency in hyp such that its dependent is dep */
- printf("found a dep of word %d in ref, it is %d\n", word_index, dep);
- gov_hyp = word_get_gov_relative_index(word_buffer_get_word_n(config_get_buffer(c), dep));
+ /* printf("found a dep of word %d in ref, it is %d\n", word_index, dep); */
+ gov_hyp = word_get_gov_index(word_buffer_get_word_n(config_get_buffer(c), dep));
- printf("gov of %d in hyp is %d\n", dep,gov_hyp);
+ /* printf("gov of %d in hyp is %d\n", dep,gov_hyp); */
if(gov_hyp != gov_ref) return 0;
/*
if((dep >= c->ds->length)
|| (c->ds->array[dep].gov == NULL)
- || (word_get_relative_index(c->ds->array[dep].gov) != word_index)
+ || (word_get_index(c->ds->array[dep].gov) != word_index)
|| (c->ds->array[dep].label != word_get_label(ref->words[dep])))
return 0;*/
}
@@ -30,21 +30,21 @@ int check_all_dependents_of_word_in_ref_are_in_hyp(config *c, word_buffer *ref,
return 1;
}
-int oracle_parser_arc_eager(config *c, word_buffer *ref)
+int oracle_parser_arc_eager(config *c, word_buffer *ref, int start_sentence_index)
{
word *s0; /* word on top of stack */
word *b0; /* next word in the bufer */
int s0_index, b0_index;
int s0_gov_index, b0_gov_index;
- if(!stack_is_empty(c->st) && !word_buffer_is_empty(c->bf)){
- s0 = stack_top(c->st);
- s0_index = word_get_relative_index(s0);
- s0_gov_index = word_get_gov_relative_index(word_buffer_get_word_n(ref, s0_index));
+ if(!stack_is_empty(config_get_stack(c)) && !word_buffer_is_empty(config_get_buffer(c))){
+ s0 = stack_top(config_get_stack(c));
+ s0_index = word_get_index(s0);
+ s0_gov_index = word_get_gov_index(word_buffer_get_word_n(ref, s0_index));
- b0 = word_buffer_b0(c->bf);
- b0_index = word_get_relative_index(b0);
- b0_gov_index = word_get_gov_relative_index(word_buffer_get_word_n(ref, b0_index));
+ b0 = word_buffer_b0(config_get_buffer(c));
+ b0_index = word_get_index(b0);
+ b0_gov_index = word_get_gov_index(word_buffer_get_word_n(ref, b0_index));
/* printf("s0_index = %d b0_index = %d\n", s0_index, b0_index);
printf("dans ref gov de s0 (%d) = %d\n", s0_index, s0_gov_index);
@@ -52,22 +52,17 @@ int oracle_parser_arc_eager(config *c, word_buffer *ref)
/* LEFT ARC b0 is the governor and s0 the dependent */
if(s0_gov_index == b0_index){
-
- /* printf("oracle says left\n"); */
return movement_left_code(word_get_label(word_buffer_get_word_n(ref, s0_index)));
}
/* RIGHT ARC s0 is the governor and b0 the dependent */
if(b0_gov_index == s0_index){
- /* printf("oracle says right\n"); */
return movement_right_code(word_get_label(word_buffer_get_word_n(ref, b0_index)));
}
/* REDUCE */
-
- printf("all dep in ref are in hyp = %d\n", check_all_dependents_of_word_in_ref_are_in_hyp(c, ref, s0_index));
- if((stack_height(c->st) > 2)
- && check_all_dependents_of_word_in_ref_are_in_hyp(c, ref, s0_index)
- && (word_get_gov(stack_top(c->st)) != 0)) /* word on top of the stack has a goveror */
+ if((stack_height(config_get_stack(c)) > 2)
+ && check_all_dependents_of_word_in_ref_are_in_hyp(c, ref, s0_index, start_sentence_index)
+ && (word_get_gov(stack_top(config_get_stack(c))) != 0)) /* word on top of the stack has a goveror */
{
return MVT_REDUCE;
}
@@ -75,4 +70,5 @@ int oracle_parser_arc_eager(config *c, word_buffer *ref)
/* SHIFT */
return MVT_SHIFT;
}
+ return -1;
}
diff --git a/maca_trans_parser/src/oracle_parser_arc_eager.h b/maca_trans_parser/src/oracle_parser_arc_eager.h
index 1aad136a6348a0f5aae413f5e2142a6454628a45..4f7568293d6636c10cbbc417adf6c1ed12b90042 100644
--- a/maca_trans_parser/src/oracle_parser_arc_eager.h
+++ b/maca_trans_parser/src/oracle_parser_arc_eager.h
@@ -6,6 +6,6 @@
#include"word_buffer.h"
-int oracle_parser_arc_eager(config *c, word_buffer *ref);
+int oracle_parser_arc_eager(config *c, word_buffer *ref, int start_sentence_index);
#endif
diff --git a/maca_trans_parser/src/queue.c b/maca_trans_parser/src/queue.c
index 599739a6f1fe4132289e1a479c8b593dfc042e08..4e487004e5f2e21caa9725a04cd1c5630442fd20 100644
--- a/maca_trans_parser/src/queue.c
+++ b/maca_trans_parser/src/queue.c
@@ -7,7 +7,7 @@ int queue_renumber_words(queue *bf)
int i;
int index = 1;
for(i=0; i < bf->nbelem; i++){
- word_set_relative_index(queue_elt_n(bf, i), index++);
+ word_set_index(queue_elt_n(bf, i), index++);
}
return index;
}
@@ -23,7 +23,7 @@ int queue_read_sentence(queue *bf, FILE *f, mcd *mcd_struct)
/* fprintf(stderr, "%s", buffer); */
if((buffer[0] == '\n') || (buffer[0] == ' ') || (buffer[0] == '\t')) break; /* end of the sentence */
w = word_parse_buffer(buffer, mcd_struct);
- word_set_relative_index(w, index);
+ word_set_index(w, index);
index++;
queue_add(bf, w);
if(word_is_eos(w, mcd_struct)) break;
@@ -43,12 +43,12 @@ void queue_print(FILE *f, queue *q)
fprintf(f, "(");
if(q->tail >= q->head)
for(i=q->head; i < q->tail; i++)
- fprintf(f, "%d ", word_get_relative_index(q->array[i]));
+ fprintf(f, "%d ", word_get_index(q->array[i]));
else{
for(i=q->head; i < q->size; i++)
- fprintf(f, "%d ", word_get_relative_index(q->array[i]));
+ fprintf(f, "%d ", word_get_index(q->array[i]));
for(i=0; i < q->tail; i++)
- fprintf(f, "%d ", word_get_relative_index(q->array[i]));
+ fprintf(f, "%d ", word_get_index(q->array[i]));
}
fprintf(f, ")\n");
}
diff --git a/maca_trans_parser/src/simple_decoder_parser.c b/maca_trans_parser/src/simple_decoder_parser.c
index b71f307ea6afb0851b0d8b2980d06abeb9984ff5..81aea380973f88da091b72888d9a52c6a9b250e5 100644
--- a/maca_trans_parser/src/simple_decoder_parser.c
+++ b/maca_trans_parser/src/simple_decoder_parser.c
@@ -10,50 +10,6 @@
#include"feature_table.h"
#include"dico.h"
-#if 0
-void simple_decoder_buffer(context *ctx, FILE *f, feature_table *ft, int root_label)
-{
- int mvt_code;
- int mvt_type;
- int mvt_label;
- float max;
- feat_vec *fv = feat_vec_new(feature_types_nb);
- config *c = config_initial(f, ctx->mcd_struct, 0);
-
- /* read a sentence and put it in the buffer */
- while(word_buffer_read_sentence(c->bf) > 1){
- while(!config_is_terminal(c)){
- config2feat_vec_cff(ctx->features_model, c, ctx->d_perceptron_features, fv, LOOKUP_MODE);
- mvt_code = feature_table_argmax(fv, ft, &max);
- mvt_type = movement_type(mvt_code);
- mvt_label = movement_label(mvt_code);
-
- if(mvt_type == MVT_LEFT)
- if(movement_left_arc(c, mvt_label, max))
- continue;
-
- if(mvt_type == MVT_RIGHT)
- if(movement_right_arc(c, mvt_label, max))
- continue;
-
- movement_shift(c, 0, max);
- }
-
- /* config_print(stdout, c); */
-
- config_connect_subtrees(c, root_label);
- depset_print2(stdout, c->ds, ctx->dico_labels);
-
-
- /* config_free(c); */
- c = config_initial(f, ctx->mcd_struct, 0);
- }
-
- feat_vec_free(fv);
-}
-
-#endif
-
void simple_decoder_stream(context *ctx, FILE *f, feature_table *ft, int root_label)
{
int mvt_code;
@@ -62,74 +18,64 @@ void simple_decoder_stream(context *ctx, FILE *f, feature_table *ft, int root_la
float max;
feat_vec *fv = feat_vec_new(feature_types_nb);
config *c = NULL;
-
+ word *dep;
+
c = config_initial(f, ctx->mcd_struct, 5);
- while(!config_is_terminal(c)){
- /* config_print(stdout, c); */
+ while(1){
config2feat_vec_cff(ctx->features_model, c, ctx->d_perceptron_features, fv, LOOKUP_MODE);
- /* feat_vec_print(stdout, fv); */
mvt_code = feature_table_argmax(fv, ft, &max);
mvt_type = movement_type(mvt_code);
mvt_label = movement_label(mvt_code);
-
- /* printf("code predicted = %d\n", mvt_code); */
- /* movement_print(stdout, mvt_code, ctx->dico_labels); */
- /* sentence is complete */
- if((stack_height(c->st)==1) && (mvt_type == MVT_RIGHT) && (mvt_label == root_label)){
- /* if((mvt_type == MVT_RIGHT) && (mvt_label == root_label)){ */
- /* if(mvt_label == root_label){ */
- /* printf("sentence complete\n"); */
- /*config_print(stdout, c); */
-
- /* create the root arc */
- movement_right_arc(c, mvt_label, 0);
-
- /* shift dummy word in stack */
- stack_pop(config_get_stack(c));
-
- /* config_print(stdout, c); */
-
- /* config_connect_subtrees(c, root_label); */
- /* depset_print_new_index(stdout, c->ds, ctx->dico_labels); */
-
- depset_print2(stdout, c->ds, ctx->dico_labels);
-
- /* pop the dummy word */
- stack_pop(c->st);
- /* remplace it with a fresh one */
- stack_push(c->st, word_create_dummy(ctx->mcd_struct));
-
- /* empty depset */
- depset_free(c->ds);
- c->ds = depset_new();
-
- /* renumber the words that are left in the buffer */
- /* c->current_index = queue_renumber_words(c->bf); */
- continue;
- }
-
- if(mvt_type == MVT_LEFT)
- if(movement_left_arc(c, mvt_label, max))
+ /* config_print(stdout, c); */
+ /* movement_print(stdout, mvt_code, ctx->dico_labels); */
+
+ if(mvt_type == MVT_LEFT){
+ dep = stack_s0(config_get_stack(c));
+ if(movement_left_arc(c, mvt_label, max)){
+ /* printf("%d\t", word_get_index(dep));
+ printf("%s\t", word_get_input(dep));
+ printf("%d\t", word_get_gov(dep));
+ printf("%s\n", dico_int2string(ctx->dico_labels, word_get_label(dep)));*/
continue;
+ }
+ }
- if(mvt_type == MVT_RIGHT)
- if(movement_right_arc(c, mvt_label, max))
+ if(mvt_type == MVT_RIGHT){
+ dep = word_buffer_b0(config_get_buffer(c));
+ if(movement_right_arc(c, mvt_label, max)){
+ /* printf("%d\t", word_get_index(dep));
+ printf("%s\t", word_get_input(dep));
+ printf("%d\t", word_get_gov(dep));
+ printf("%s\n", dico_int2string(ctx->dico_labels, word_get_label(dep)));*/
continue;
+ }
+ }
if(mvt_type == MVT_REDUCE)
if(movement_reduce(c, max))
continue;
movement_shift(c, 1, max);
+
+ if(word_buffer_is_last(config_get_buffer(c))) break;
}
-
- /* config_print(stdout, c); */
-
- /* config_connect_subtrees(c, root_label); */
-
- depset_print2(stdout, c->ds, ctx->dico_labels);
+ for(int i=1; i < config_get_buffer(c)->nbelem; i++){
+ dep = word_buffer_get_word_n(config_get_buffer(c), i);
+ printf("%s\t", word_get_input(dep));
+ printf("%d\t", word_get_gov(dep));
+ /* printf("label = %d\n", word_get_label(dep)); */
+ char *label = (word_get_label(dep) == -1)? NULL : dico_int2string(ctx->dico_labels, word_get_label(dep));
+ if(label != NULL)
+ printf("%s\t", label) ;
+ else
+ printf("_\t");
+ if((label != NULL) && !strcmp(label, "eos"))
+ printf("1\n");
+ else
+ printf("0\n");
+ }
/* config_free(c); */
feat_vec_free(fv);
@@ -145,16 +91,10 @@ void simple_decoder(context *ctx)
root_label = dico_string2int(ctx->dico_labels, ctx->root_label);
if(root_label == -1) root_label = 0;
- simple_decoder_stream(ctx, f, ft, root_label);
- /*
- if(ctx->stream_mode)
- simple_decoder_stream(ctx, f, ft, root_label);
- else
- simple_decoder_buffer(ctx, f, ft, root_label);
- */
+ simple_decoder_stream(ctx, f, ft, root_label);
+
feature_table_free(ft);
if(ctx->input_filename)
fclose(f);
}
-
diff --git a/maca_trans_parser/src/simple_decoder_tagger.c b/maca_trans_parser/src/simple_decoder_tagger.c
index a145fee8ef889546f71fb55375ecf88bf6d265d8..9e3341c362a39892648f55cbf32b665856cdff48 100644
--- a/maca_trans_parser/src/simple_decoder_tagger.c
+++ b/maca_trans_parser/src/simple_decoder_tagger.c
@@ -85,9 +85,10 @@ void simple_decoder_stream(context *ctx)
float max;
word *w;
dico *dico_pos = dico_vec_get_dico(ctx->vocabs, (char *)"POS");
-
- c = config_initial_no_dummy_word(f, ctx->mcd_struct, 5);
- while(!config_is_terminal(c)){
+ int res;
+ c = config_initial_no_dummy_word(f, ctx->mcd_struct, 5);
+
+ while(1){
if(ctx->f2p)
add_signature_to_words_in_word_buffer(c->bf, ctx->f2p);
/* config_print(stdout, c); */
@@ -100,10 +101,11 @@ void simple_decoder_stream(context *ctx)
w = word_buffer_b0(c->bf);
printf("%s\t%s\n", w->input, dico_int2string(dico_pos, postag));
- if(postag != -1)
- movement_tagger(c, postag, max, 1);
-
-
+ res = movement_tagger(c, postag, max, 1);
+
+ /* printf(" current index = %d nb elem = %d\n", c->bf->current_index, c->bf->nbelem); */
+
+ if(res == 0) break;
}
/* config_print(stdout, c); */
diff --git a/maca_trans_parser/src/stack.c b/maca_trans_parser/src/stack.c
index 24ed623226c1272f39cdfb60efb623bd351b0b0c..cbd419e9c496e6e6cc6c89b3616d0f21ea6a161d 100644
--- a/maca_trans_parser/src/stack.c
+++ b/maca_trans_parser/src/stack.c
@@ -93,7 +93,7 @@ void stack_print(FILE *buffer, stack *s)
if(s){
fprintf(buffer, "[");
for(i=0; i < stack_height(s); i++)
- fprintf(buffer, " %d", word_get_relative_index(s->array[i]));
+ fprintf(buffer, " %d", word_get_index(s->array[i]));
fprintf(buffer, "]");
}
}