modified maca_trans_lemmatizer so that it works when no morphological rules...

modified maca_trans_lemmatizer so that it works when no morphological rules are supplied, in which case it only uses the exception lexicon

maca_trans_parser/src/maca_trans_lemmatizer.c

@@ -119,7 +119,7 @@ int main(int argc, char *argv[])
   int l_rule_code;
   char *l_rule;
   float max;
-
+  feature_table *ft = NULL;
   
   maca_lemmatizer_check_options(ctx);
   maca_lemmatizer_set_linguistic_resources_filenames(ctx);
@@ -131,8 +131,17 @@ int main(int argc, char *argv[])
   ctx->vocabs = dico_vec_read(ctx->vocabs_filename, ctx->hash_ratio);
   mcd_link_to_dico(ctx->mcd_struct, ctx->vocabs, ctx->verbose);
 
+  if(d_l_rules->nbelem){
     ctx->d_perceptron_features = dico_vec_get_dico(ctx->vocabs, (char *)"d_perceptron_features");
-  feature_table *ft = feature_table_load(ctx->perc_model_filename, ctx->verbose);
+    ft = feature_table_load(ctx->perc_model_filename, ctx->verbose);
+  }
+  else{
+    if(ctx->verbose)
+      fprintf(stderr, "no morphological rules loaded\n");
+    ctx->d_perceptron_features = NULL;
+    ft = NULL;
+  }
+
   c = config_new(f, ctx->mcd_struct, 5); 
   
   while(!config_is_terminal(c)){
@@ -152,6 +161,10 @@ int main(int argc, char *argv[])
 	print_word(b0, ctx->mcd_struct, lemma_from_fplm);
       }
     // if lemma is not found in exception file, predict an l_rule 
+      else{
+	if(ft == NULL){ /* no rule model just print the form as a lemma */
+	  print_word(b0, ctx->mcd_struct, form);
+	}
 	else{
 	  config2feat_vec_cff(ctx->features_model, c, ctx->d_perceptron_features, fv, LOOKUP_MODE);
 	  //	feat_vec_print_string(fv, ctx->d_perceptron_features);
@@ -182,17 +195,17 @@ int main(int argc, char *argv[])
 	  if(i == 10){
 	    print_word(b0, ctx->mcd_struct, form);
 	  }
-	
 	  free(vcode_array);
 	}
       }
+    }
     word_buffer_move_right(c->bf);
   }
   config_free(c); 
   if (ctx->input_filename) fclose(f);
   context_free(ctx);
   fplm_free(exceptions);
-  feature_table_free(ft);
+  if(ft) feature_table_free(ft);
   return 0;
 }
 

maca_trans_parser/src/maca_trans_lemmatizer_mcf2cff.c

@@ -171,15 +171,17 @@ int main(int argc, char *argv[])
   fplm_struct *exceptions;
   ctx = context_read_options(argc, argv);
 
-
-
-
-
   //  decode_lemmatizer_set_linguistic_resources_filenames(ctx);
   maca_trans_lemmatizer_mcf2cff_check_options(ctx);
   exceptions = fplm_load_file(ctx->fplm_filename, ctx->verbose);
   d_l_rules = dico_read(ctx->l_rules_filename, 0.5);
 
+  if(d_l_rules->nbelem == 0){
+    /* do not produce cff file when the rule file is empty */
+    /*    exit(1);*/
+  }
+
+  
   ctx->features_model = feat_model_read(ctx->features_model_filename, feat_lib_build(), ctx->verbose);
 
   if(ctx->mode == TRAIN_MODE){
@@ -203,22 +205,25 @@ int main(int argc, char *argv[])
   /* add the feature dictionnary to the dico vector */
   dico_vec_add(ctx->vocabs, ctx->d_perceptron_features);
   
+
   /* open output file */
   if(ctx->cff_filename)
     output_file = myfopen(ctx->cff_filename, "w");
   else
     output_file = stdout;
   
+  if(d_l_rules->nbelem)
     generate_training_file(output_file, ctx, d_l_rules, exceptions);
   
+  
+  if(ctx->cff_filename)
+    fclose(output_file);
+  
   if(ctx->mode == TRAIN_MODE){
     /* dico_print(ctx->perceptron_features_filename, ctx->d_perceptron_features); */
     dico_vec_print(ctx->vocabs_filename, ctx->vocabs);
-    
   }
 
-  if(ctx->cff_filename)
-    fclose(output_file);
   context_free(ctx);
   return 0;
 }

perceptron/exec/perceptron_train.c

@@ -35,12 +35,12 @@ int main(int argc, char *argv[])
   train_check_options(ctx);
 
   look_for_number_of_features_and_classes(ctx->cff_filename, &nb_feat, &nb_class);
-
+  if(nb_class > 1){
     ft = feature_table_new(nb_feat, nb_class);
     fprintf(stderr, "table allocated (%d x %d)\n", nb_feat, nb_class); 
     perceptron_avg(ctx->cff_filename, ft, ctx->iteration_nb);
     feature_table_dump(ctx->perc_model_filename, ft);
-  
+  }
   perceptron_context_free(ctx);
 
   return 0;

perceptron/lib/src/cf_file.c

@@ -56,8 +56,11 @@ void look_for_number_of_features_and_classes(char *filename, int *max_feat, int
   FILE *f = fopen(filename, "r");
   char *token;
   int nb;
+
   *max_feat = 0;
   *max_class = 0;
+  if(f == NULL)
+    return;
   while(fgets(buffer, 10000, f)){
     buffer[strlen(buffer) - 1] = '\0';
     token = strtok(buffer, "\t");