Create toBenchIEformat.py

evaluation_data/benchIE/toBenchIEformat.py

@@ -0,0 +1,32 @@
+sent2id = dict()
+with open('sample300_en.txt', 'r') as f:
+    i = 1
+    for line in f.readlines():
+        sent2id[line.replace('\n', '')] = i
+        i += 1
+
+with open('output_extractions.txt', 'r') as fin, open('compactIE_1.2_new.txt', 'w') as fout:
+    for line in fin.readlines():
+        sentence, extraction, score = line.split('\t')
+        sentId = sent2id[sentence]
+        try:
+            arg1 = extraction[extraction.index('<arg1>') + 6:extraction.index('</arg1>')]
+            arg1 = arg1.strip()
+        except:
+            print("subject error!", extraction)
+            arg1 = ""
+        try:
+            rel = extraction[extraction.index('<rel>') + 5:extraction.index('</rel>')]
+            rel = rel.strip()
+        except:
+            print("predicate error!", extraction)
+            rel = ""
+        try:
+            arg2 = extraction[extraction.index('<arg2>') + 6:extraction.index('</arg2>')]
+            arg2 = arg2.strip()
+            if arg2 == "":
+                continue
+        except:
+            print("object error!", extraction)
+            arg2 = ""
+        print("{}\t{}\t{}\t{}".format(sentId, arg1, rel, arg2), file=fout)