From 3e1224b84ef926e3077a28512cf6fd688e7803e6 Mon Sep 17 00:00:00 2001
From: Alice BRENON <alice.brenon@ens-lyon.fr>
Date: Sat, 10 Dec 2022 10:06:24 +0100
Subject: [PATCH] Don't forget to escape XML characters in output
---
scripts/stanza-annotator.py | 9 +++++----
1 file changed, 5 insertions(+), 4 deletions(-)
diff --git a/scripts/stanza-annotator.py b/scripts/stanza-annotator.py
index e5e6052..2d341b9 100755
--- a/scripts/stanza-annotator.py
+++ b/scripts/stanza-annotator.py
@@ -4,6 +4,7 @@ import os
import os.path
import stanza
import sys
+from xml.sax.saxutils import escape, quoteattr
class Annotator:
def __init__(self, source, target):
@@ -45,10 +46,10 @@ class Annotator:
print('\t</s>', file=target)
def encode_token(self, token, target):
- form = token.text
- lemma = '+'.join(map(lambda w: w.lemma, token.words))
- upos = '+'.join(map(lambda w: w.upos, token.words))
- print(f'\t\t<w lemma="{lemma}" pos="{upos}">{form}</w>', file=target)
+ form = escape(token.text)
+ lemma = quoteattr('+'.join(map(lambda w: w.lemma, token.words)))
+ upos = quoteattr('+'.join(map(lambda w: w.upos, token.words)))
+ print(f'\t\t<w lemma={lemma} pos={upos}>{form}</w>', file=target)
if __name__ == '__main__':
Annotator(sys.argv[1], sys.argv[2]).run()
--
GitLab