ENH: Added scripts to create the data for reproducing the SemEval 201…

…4 paper.

scripts_srl/train_test_semantic_parser.sh

@@ -41,7 +41,7 @@ then
     allow_unseen_predicates=false # This should be irrelevant.
     use_predicate_senses=false #true
     formalism=$7 #pcedt #pas #dm
-    subfolder=sdp/${formalism}
+    subfolder=semeval2014_data/${formalism}
 else
     allow_self_loops=true
     allow_root_predicate=false
@@ -84,15 +84,10 @@ if [ "$language" == "english" ]
 then
     if [ "$file_format" == "sdp" ]
     then
-        file_train=${path_data}/${formalism}_augmented_train+dev0.sdp
-        #file_train=${path_data}/${formalism}_augmented_train+dev0+dev.sdp
-        files_test[0]=${path_data}/${formalism}_augmented_dev.sdp
-        files_test[1]=${path_data}/${formalism}_augmented_test.sdp
-        #files_test[1]=${path_data}/${formalism}_augmented_test_blind.sdp
-        #files_test[1]=${path_data}/${formalism}_augmented_train+dev0.sdp
-        #files_test[0]=${path_data}/${formalism}_augmented_dev0.sdp
-        #files_test[1]=${path_data}/${formalism}_augmented_dev.sdp
-        #files_test[2]=${path_data}/${formalism}_augmented_train.sdp
+        file_train=${path_data}/${language}_${formalism}_augmented_train.sdp
+        #file_train=${path_data}/${language}_${formalism}_augmented_train+dev.sdp
+        files_test[0]=${path_data}/${language}_${formalism}_augmented_dev.sdp
+        files_test[1]=${path_data}/${language}_${formalism}_augmented_test.sdp
     else
         file_train=${path_data}/${language}_train.conll2008
         files_test[0]=${path_data}/${language}_test.conll2008

scripts_srl/train_test_submission_closed.sh

@@ -0,0 +1,24 @@
+model_type=af+as+cs+gp+cp+ccp
+open=false
+
+formalism=dm
+C=0.01
+cost_fp=0.3
+cost_fn=0.7
+echo "${C} ${cost_fp} ${cost_fn} ${formalism} ${model_type} ${open}"
+./train_test_semantic_parser.sh english ${C} ${cost_fp} ${cost_fn} ${model_type} ${open} ${formalism} >& out_open-${open}_deterministic_${formalism}_${C}_${cost_fp}_${cost_fn}_${model_type}
+
+formalism=pas
+C=0.01
+cost_fp=0.4
+cost_fn=0.6
+echo "${C} ${cost_fp} ${cost_fn} ${formalism} ${model_type} ${open}"
+./train_test_semantic_parser.sh english ${C} ${cost_fp} ${cost_fn} ${model_type} ${open} ${formalism} >& out_open-${open}_deterministic_${formalism}_${C}_${cost_fp}_${cost_fn}_${model_type}
+
+formalism=pcedt
+C=0.01
+cost_fp=0.3
+cost_fn=0.7
+echo "${C} ${cost_fp} ${cost_fn} ${formalism} ${model_type} ${open}"
+./train_test_semantic_parser.sh english ${C} ${cost_fp} ${cost_fn} ${model_type} ${open} ${formalism} >& out_open-${open}_deterministic_${formalism}_${C}_${cost_fp}_${cost_fn}_${model_type}
+

scripts_srl/train_test_submission_open.sh

@@ -0,0 +1,24 @@
+model_type=af+as+cs+cp+ccp #af+as+cs+gp+cp+ccp
+open=true #false
+
+formalism=dm
+C=0.01
+cost_fp=0.4
+cost_fn=0.6
+echo "${C} ${cost_fp} ${cost_fn} ${formalism} ${model_type} ${open}"
+./train_test_semantic_parser.sh english ${C} ${cost_fp} ${cost_fn} ${model_type} ${open} ${formalism} >& out_open-${open}_deterministic_${formalism}_${C}_${cost_fp}_${cost_fn}_${model_type}
+
+formalism=pas
+C=0.01
+cost_fp=0.4
+cost_fn=0.6
+echo "${C} ${cost_fp} ${cost_fn} ${formalism} ${model_type} ${open}"
+./train_test_semantic_parser.sh english ${C} ${cost_fp} ${cost_fn} ${model_type} ${open} ${formalism} >& out_open-${open}_deterministic_${formalism}_${C}_${cost_fp}_${cost_fn}_${model_type}
+
+formalism=pcedt
+C=0.01
+cost_fp=0.4
+cost_fn=0.6
+echo "${C} ${cost_fp} ${cost_fn} ${formalism} ${model_type} ${open}"
+./train_test_semantic_parser.sh english ${C} ${cost_fp} ${cost_fn} ${model_type} ${open} ${formalism} >& out_open-${open}_deterministic_${formalism}_${C}_${cost_fp}_${cost_fn}_${model_type}
+

semeval2014_data/scripts/augment_with_companion_data.py

@@ -0,0 +1,47 @@
+import sys
+
+keep_document_names = True
+
+filepath = sys.argv[1]
+filepath_companion = sys.argv[2]
+f = open(filepath)
+f_companion = open(filepath_companion)
+for line, line_companion in zip(f, f_companion):
+    line = line.rstrip('\n')
+    line_companion = line_companion.rstrip('\n')
+    if line.startswith('#') and line.split('\t')[0] != '#':
+        if keep_document_names: print line
+    elif line == '':
+        print line
+    else:
+        fields = line.split("\t")
+        fields_companion = line_companion.split("\t")
+        word_index = fields[0]
+        word = fields[1]
+        lemma = fields[2]
+        pos = fields[3]
+        if len(fields) == 4:
+            top = '-'
+            pred = '-'
+        else:
+            top = fields[4]
+            pred = fields[5]
+        args = fields[6:]
+        predicted_pos = fields_companion[0]
+        head = fields_companion[1]
+        deprel = fields_companion[2]
+
+        fields_output = [word_index,
+                         word,
+                         lemma,
+                         pos, #predicted_pos
+                         head,
+                         deprel,
+                         top,
+                         pred]
+        fields_output.extend(args);
+        line_output = '\t'.join(fields_output)
+        print line_output
+
+f.close()
+f_companion.close()