modified oracle functions

maca_common/include/word.h

@@ -96,6 +96,13 @@ typedef struct _word {
 #define word_set_signature(w, val) ((w)->signature = (val))
 #define word_set_index(w, val)     ((w)->index = (val))
 
+int word_sprint_lemma(char *s, word *w);
+int word_sprint_form(char *s, word *w);
+
+
+
+
+
 word *word_new(char *input);
 word *word_create_dummy(mcd *mcd_struct);
 
@@ -110,6 +117,6 @@ word *word_parse_buffer(char *buffer, mcd *mcd_struct);
 int word_is_eos(word *w, mcd *mcd_struct);
 int word_get_gov_index(word *w);
 void word_print_col_n(FILE *f, word *w, int n);
-void word_sprint_col_n(char *s, word *w, int n);
+int word_sprint_col_n(char *s, word *w, int n);
 
 #endif

maca_common/src/word.c

@@ -166,13 +166,15 @@ void word_print_col_n(FILE *f, word *w, int n)
   }
 }
 
-void word_sprint_col_n(char *s, word *w, int n)
+int word_sprint_col_n(char *s, word *w, int n)
 {
   int i;
   int col = 0;
   int j = 0;
+  s[0] = '\0';
+  if(n == -1) return 0;
   char *buffer = w->input;
-  if(buffer == NULL) return;
+  if(buffer == NULL) return 0;
   int l= strlen(buffer);
   for(i=0; i < l; i++){
     if(buffer[i] == '\t') {
@@ -183,4 +185,5 @@ void word_sprint_col_n(char *s, word *w, int n)
       s[j++] = buffer[i];
   }
   s[j] = '\0';
+  return 1;
 }

maca_trans_parser/CMakeLists.txt

@@ -3,6 +3,8 @@ set(SOURCES src/context.c
   src/feat_fct.c
   src/feat_types.c
   src/oracle_parser_arc_eager.c
+  src/oracle_tagparser_arc_eager.c
+  src/oracle_tagger.c
   src/simple_decoder_parser_arc_eager.c
   src/simple_decoder_tagger.c
   src/feat_lib.c
@@ -20,14 +22,13 @@ set(SOURCES src/context.c
   src/classifier.c
   src/simple_decoder_tagparser_arc_eager.c
 
+
+#  src/simple_decoder_parser.c
+#  src/oracle_parser.c
 #  src/global_feat_vec.c
 #  src/movement_parser_arc_eager.c
 #  src/movement_tagparser_arc_eager.c
 #  src/movement_tagger.c
-#  src/oracle_parser.c
-#  src/oracle_tagparser_arc_eager.c
-#  src/oracle_tagger.c
-#  src/simple_decoder_parser.c
 #  src/simple_decoder_tagparser_arc_eager.c
 #  src/simple_decoder_forrest.c
 #  src/simple_decoder_tagger_bt.c

maca_trans_parser/src/maca_trans_tagger_mcf2cff.c

@@ -12,10 +12,10 @@
 #include"word_emb.h"
 #include"config2feat_vec.h"
 
-int oracle_tagger(config *c)
+/*int oracle_tagger(config *c)
 {
   return word_get_pos(word_buffer_b0(config_get_buffer(c)));
-}
+  }*/
 
 #if 1
 void add_signature_to_words_in_word_buffer(word_buffer *bf, form2pos *f2p)
@@ -96,7 +96,6 @@ void generate_training_file(FILE *output_file, context *ctx)
   int mvt_code;
   feat_vec *fv = ctx->classif->fv;
   FILE *mcf_file = myfopen(ctx->input_filename, "r");
-  int postag;
   int word_nb = 0;
   /* dico *dico_pos = dico_vec_get_dico(ctx->vocabs, (char *)"POS"); */
   
@@ -109,11 +108,10 @@ void generate_training_file(FILE *output_file, context *ctx)
 
     if((++word_nb % 1000) == 0) fprintf(stderr, "\rword %d", word_nb);
 
-    postag = oracle_tagger(c);
     config2feat_vec_cff(ctx->classif->fm, c, ctx->classif->d_features, fv, ctx->mode);
     /* word_print(stdout, word_buffer_b0(config_get_buffer(c))); */
     
-    mvt_code = mvt_tagset_get_code(classifier_get_output_tagset(ctx->classif), MVT_POS, postag);
+    mvt_code = oracle_tagger(c, classifier_get_output_tagset(ctx->classif));
     fprintf(output_file, "%d", mvt_code);
     feat_vec_print(output_file, fv);
     

maca_trans_parser/src/oracle_parser_arc_eager.c

@@ -2,7 +2,6 @@
 #include<stdlib.h>
 #include<string.h>
 #include"word_buffer.h"
-/* #include"movement_parser_arc_eager.h" */
 #include"mvt_tagset.h"
 #include"config.h"
 #include"dico.h"
@@ -69,7 +68,6 @@ int oracle_parser_arc_eager(config *c, word_buffer *ref, int root_label, mvt_tag
        ){
       
       return mvt_tagset_get_code(tagset, MVT_ROOT, 0);
-      /* return MVT_PARSER_ROOT; */
     }
     
     /* word on the top of the stack is an end of sentence marker */
@@ -77,31 +75,25 @@ int oracle_parser_arc_eager(config *c, word_buffer *ref, int root_label, mvt_tag
        && (word_get_sent_seg(word_buffer_get_word_n(config_get_buffer(c), s0_index)) != 1)){
 	
 	return mvt_tagset_get_code(tagset, MVT_EOS, 0);
-	/* return MVT_PARSER_EOS; */
     }
     
     /* LEFT ARC  b0 is the governor and s0 the dependent */
     if(s0_gov_index == b0_index){
       return mvt_tagset_get_code(tagset, MVT_LEFT, word_get_label(word_buffer_get_word_n(ref, s0_index)));
-      /* return movement_parser_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){
       return mvt_tagset_get_code(tagset, MVT_RIGHT, word_get_label(word_buffer_get_word_n(ref, b0_index)));
-      /* return movement_parser_right_code(word_get_label(word_buffer_get_word_n(ref, b0_index)));  */
     }
     /* REDUCE */
     if((stack_nbelem(config_get_stack(c)) > 1) 
        && check_all_dependents_of_word_in_ref_are_in_hyp(c, ref, s0_index)     /* word on top must have all its dependents */
        && (word_get_gov(stack_top(config_get_stack(c))) != WORD_INVALID_GOV)){ /* word on top of the stack has a governor */
       return mvt_tagset_get_code(tagset, MVT_REDUCE, 0);
-      /* return MVT_PARSER_REDUCE; */
     }
   }
 
   /* SHIFT */ 
   return mvt_tagset_get_code(tagset, MVT_SHIFT, 0);
- /* return MVT_PARSER_SHIFT; */
-
 }

maca_trans_parser/src/oracle_tagger.c

 #include"oracle_tagger.h"
 
-int oracle_tagger(config *c)
+int oracle_tagger(config *c, mvt_tagset *tagset)
 {
-    return word_get_pos(word_buffer_b0(config_get_buffer(c)));
+  int postag = word_get_pos(word_buffer_b0(config_get_buffer(c)));
+  int mvt_code = mvt_tagset_get_code(tagset, MVT_POS, postag);
+  return mvt_code;
 }

maca_trans_parser/src/oracle_tagger.h

@@ -5,6 +5,11 @@
 #include<stdlib.h>
 #include"config.h"
 
-int oracle_tagger(config *c);
+
+#include"mvt_tagset.h"
+
+
+int oracle_tagger(config *c, mvt_tagset *tagset);
+
 
 #endif

maca_trans_parser/src/oracle_tagparser_arc_eager.c

 #include<stdio.h>
 #include<stdlib.h>
 #include<string.h>
-#include"word_buffer.h"
-#include"movement_tagparser_arc_eager.h"
+#include"config.h"
+#include"tm.h"
+#include"oracle_tagger.h"
+#include"oracle_parser_arc_eager.h"
 
-int check_all_dependents_of_word_in_ref_are_in_hyp(config *c, word_buffer *ref, int word_index)
+int oracle_tagparser_arc_eager(config *c, word_buffer *ref, tm *machine, int root_label)
 {
-  int dep;
-  int gov_ref;
-  int gov_hyp;
-  int sentence_change;  
-
-#if 0
-  for(dep = word_index - 1; (dep >= 0) && (word_get_sent_seg(word_buffer_get_word_n(ref, dep)) == 0); dep--){
-    gov_ref = word_get_gov_index(word_buffer_get_word_n(ref, dep)); 
-    if(gov_ref  == word_index){ /* dep is a dependent of word in ref */
-      /* check that dep has the same governor in hyp */
-      gov_hyp = word_get_gov_index(word_buffer_get_word_n(config_get_buffer(c), dep));
-      if(gov_hyp != gov_ref) return 0;
-    }
-  }
-
-  for(dep = word_index + 1; ((dep < word_buffer_get_nbelem(ref)) && (word_get_sent_seg(word_buffer_get_word_n(ref, dep)) == 0)); dep++){
-    gov_ref = word_get_gov_index(word_buffer_get_word_n(ref, dep)); 
-    if(gov_ref  == word_index){ /* dep is a dependent of word in ref */
-      /* check that dep has the same governor in hyp */
-      gov_hyp = word_get_gov_index(word_buffer_get_word_n(config_get_buffer(c), dep));
-      if(gov_hyp != gov_ref) return 0;
-    }
-  }
-#endif
-
-#if 1
-  for(dep = word_index - 1; (dep >= 0) && (word_get_sent_seg(word_buffer_get_word_n(ref, dep)) == 0); dep--){
-    gov_ref = word_get_gov_index(word_buffer_get_word_n(ref, dep)); 
-    if(gov_ref  == word_index){ /* dep is a dependent of word in ref */
-      /* check that dep has the same governor in hyp */
-      gov_hyp = word_get_gov_index(word_buffer_get_word_n(config_get_buffer(c), dep));
-      if(gov_hyp != gov_ref) return 0;
-    }
-  }
-
-  sentence_change = 0;
-  for(dep = word_index + 1; (dep < word_buffer_get_nbelem(ref)) && (sentence_change == 0); dep++){
-    if(word_get_sent_seg(word_buffer_get_word_n(ref, dep)) == 1)
-      sentence_change = 1;
-    gov_ref = word_get_gov_index(word_buffer_get_word_n(ref, dep)); 
-    if(gov_ref  == word_index){ /* dep is a dependent of word in ref */
-      /* look for a dependency in hyp such that its dependent is dep */
-      gov_hyp = word_get_gov_index(word_buffer_get_word_n(config_get_buffer(c), dep));
-      if(gov_hyp != gov_ref) return 0;
-    }
-  }
-
-#endif
-  
-  return 1;
-}
-
-int oracle_tagparser_arc_eager(config *c, word_buffer *ref, int root_label)
-{
-  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;
-  int s0_label;
-  /* int s0_label_in_hyp; */
-
-  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));
-
-  
-  /* give a pos to b0  if it does not have one */
-  if(word_get_pos(b0) == -1){
-    /* word_set_pos(b0, word_get_pos(word_buffer_get_word_n(ref, b0_index))); */
-    /* return movement_tagparser_postag(word_get_pos(b0)); */
-
-    return movement_tagparser_postag_code(word_get_pos(word_buffer_get_word_n(ref, b0_index)));
-  }
-  
-  
-  /* if(!stack_is_empty(config_get_stack(c)) && !word_buffer_is_empty(config_get_buffer(c))){ */
-  if(!stack_is_empty(config_get_stack(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));
-    s0_label = word_get_label(word_buffer_get_word_n(ref, s0_index));
-    /* s0_label_in_hyp = word_get_label(word_buffer_get_word_n(config_get_buffer(c), s0_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);
-	  printf("dans ref gov de b0 (%d) = %d\n", b0_index, b0_gov_index);*/
-    
-    
-    
-    
-    /* s0 is the root of the sentence */
-    if((s0_label == root_label)
-       && check_all_dependents_of_word_in_ref_are_in_hyp(c, ref, s0_index)
-       ){
-      return MVT_TAGPARSER_ROOT;
-    }
-
-    /* word on the top of the stack is an end of sentence marker */
-      if((word_get_sent_seg(word_buffer_get_word_n(ref, s0_index)) == 1)
-       && (word_get_sent_seg(word_buffer_get_word_n(config_get_buffer(c), s0_index)) != 1)){
-      return MVT_TAGPARSER_EOS;
-      }
-
-    /* LEFT ARC  b0 is the governor and s0 the dependent */
-    if(s0_gov_index == b0_index){
-      return movement_tagparser_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){
-      return movement_tagparser_right_code(word_get_label(word_buffer_get_word_n(ref, b0_index))); 
-    }
-    /* REDUCE */
-    if((stack_nbelem(config_get_stack(c)) > 1) 
-       && check_all_dependents_of_word_in_ref_are_in_hyp(c, ref, s0_index)     /* word on top must have all its dependents */
-       && (word_get_gov(stack_top(config_get_stack(c))) != WORD_INVALID_GOV)){ /* word on top of the stack has a governor */
-      return MVT_TAGPARSER_REDUCE;
-    }
-  }
-
-  /* SHIFT */
-  return MVT_TAGPARSER_SHIFT;
-
+  char *state_name = tm_get_state_name(machine, c->current_state_nb);
+  tm_state *state = machine->state_array[c->current_state_nb];
+  classifier *classif = state->classif;
+  mvt_tagset *output_tagset = classif->output_tagset;
+  
+  if(state_name == NULL) return -1;
+
+  if(!strcmp(state_name, "TAGGER"))
+    return oracle_tagger(c, output_tagset);
+  if(!strcmp(state_name, "PARSER"))
+    return oracle_parser_arc_eager(c, ref, root_label, output_tagset);
+  return -1;
 }

maca_trans_parser/src/tm.c

@@ -60,6 +60,11 @@ tm_state *tm_get_state(tm *machine, char *state_name)
   return machine->state_array[state_number];
 }
 
+char *tm_get_state_name(tm *machine, int state_nb)
+{
+  return dico_int2string(machine->d_states, state_nb);
+}
+
 
 void tm_add_transition(tm *machine, char *origin_state_name, char *destination_state_name, int mvt_code)
 {

maca_trans_parser/src/tm.h

@@ -41,6 +41,8 @@ tm            *tm_load(char *filename, char *absolute_path, int verbose);
 void           tm_print(FILE *f, tm *machine);
 void           tm_link_to_classifier_vector(tm *machine, classifier_vec *classif_vec);
 int            tm_delta(tm *machine, int state_nb, int symbol);
+char          *tm_get_state_name(tm *machine, int state_nb);
+
 /* tm_state      *tm_delta(tm *machine, tm_state *state, int symbol); */
 
 #endif