Create str_utils.py

models/Extracting Entities and Events as a Single Task Using a Transition-Based Neural Model/str_utils.py

@@ -0,0 +1,38 @@
+import re
+
+def normalize_sent(sent):
+    sent = str(sent).replace('``', '"')
+    sent = str(sent).replace("''", '"')
+    sent = str(sent).replace('-LRB-', '(')
+    sent = str(sent).replace('-RRB-', ')')
+    sent = str(sent).replace('-LSB-', '(')
+    sent = str(sent).replace('-RSB-', ')')
+    return sent
+
+def collapse_role_type(role_type):
+    '''
+    collapse role types from 36 to 28 following Bishan Yang 2016
+    we also have to handle types like 'Beneficiary#Recipient'
+    :param role_type:
+    :return:
+    '''
+    if role_type.startswith('Time-'):
+        return 'Time'
+    idx = role_type.find('#')
+    if idx != -1:
+        role_type = role_type[:idx]
+
+    return role_type
+
+def normalize_tok(tok, lower=False, normalize_digits=False):
+
+    if lower:
+        tok = tok.lower()
+    if normalize_digits:
+        tok = re.sub(r"\d", "0", tok)
+        tok = re.sub(r"^(\d+[,])*(\d+)$", "0", tok)
+    return tok
+
+def capitalize_first_char(sent):
+    sent = str(sent[0]).upper() + sent[1:]
+    return  sent