diff --git a/util/treebanking-scripts/report_stats.py b/util/treebanking-scripts/report_stats.py
index a6b666e..b5e6073 100644
--- a/util/treebanking-scripts/report_stats.py
+++ b/util/treebanking-scripts/report_stats.py
@@ -7,10 +7,14 @@ def report_stats(treebanks_path):
     all_sentences = []
     all_accepted = []
     all_rejected = []
+    all_overgenerated = []
+    all_illformed = []
     for i, tsuite in enumerate(sorted(glob.iglob(treebanks_path + '/**'))):
         sentences = []
         accepted = []
         rejected = []
+        overgenerated = []
+        illformed = []
         ts = itsdb.TestSuite(tsuite)
         items = list(ts.processed_items())
         #print("{} sentences in corpus {} including possible sentences with no parse.".format(len(items), ts.path.stem))
@@ -19,17 +23,29 @@ def report_stats(treebanks_path):
             sentences.append(response['i-input'])
             # In a thinned parsed forest, results will be empty if the item was not accepted as correct in treebanking.
             if len(response['results']) > 0:
-                accepted.append(response['i-input'])
-                all_accepted.append(response['i-input'])
+                if response['i-wf'] == 1:
+                    accepted.append(response['i-input'])
+                    all_accepted.append(response['i-input'])
+                else:
+                    overgenerated.append(response['i-input'])
+                    all_overgenerated.append(response['i-input'])
+                    illformed.append(response['i-input'])
                 #deriv = response.result(0).derivation()
             else:
                 #print('Rejected: {}'.format(response['i-input']))
+                if response['i-wf'] == 0:
+                    illformed.append(response['i-input'])
+                    all_illformed.append(response['i-input'])
                 rejected.append(response['i-input'])
                 all_rejected.append(response['i-input'])
-        acc = len(accepted)/len(sentences)
-        print('Corpus {} accuracy {} out of {} ({:.4f})'.format(ts.path.stem, len(accepted), len(sentences), acc))
-    acc = len(all_accepted) / len(all_sentences)
-    print('Total accuracy: {} out of {} ({:.4f})'.format(len(all_accepted), len(all_sentences), acc))
+        acc = len(accepted)/(len(sentences) - len(illformed))
+        overgen = len(overgenerated)/len(illformed) if len(illformed) > 0 else 0
+        print('Corpus {} accuracy {} out of {} ({:.4f})'.format(ts.path.stem, len(accepted), len(sentences)-len(illformed), acc))
+        #print('Corpus {} overgeneration {} out of {} ({:.4f})'.format(ts.path.stem, len(overgenerated), len(illformed), overgen))
+    acc = len(all_accepted) / (len(all_sentences) - len(all_illformed))
+    overgen = len(all_overgenerated) / len(all_illformed) if len(all_illformed) > 0 else 0
+    print('Total accuracy: {} out of {} ({:.4f})'.format(len(all_accepted), len(all_sentences)-len(all_illformed), acc))
+    #print('Total overgeneration: {} out of {} ({:.4f})'.format(len(all_overgenerated), len(all_illformed), overgen))
 
 
 if __name__ == '__main__':