Replaced emails by [email]

mailcom/parse.py

@@ -182,6 +182,16 @@ def pseudonymize_numbers(self, sentence):
 
         return "".join(new_list)
 
+    def pseudonymize_email_addresses(self, sentence):
+        split = sentence.split(" ")
+        new_list = []
+        for word in split:
+            if "@" in word:
+                new_list.append("[email]")
+            else:
+                new_list.append(word)
+        return " ".join(new_list)
+
     def concatenate(self, sentences):
         return " ".join(sentences)
 
@@ -193,6 +203,7 @@ def pseudonymize(self, text: str):
             ner = self.get_ner(sent)
             ps_sent = " ".join(self.pseudonymize_ne(ner, sent)) if ner else sent
             ps_sent = self.pseudonymize_numbers(ps_sent)
+            ps_sent = self.pseudonymize_email_addresses(ps_sent)
             pseudonymized_sentences.append(ps_sent)
         return self.concatenate(pseudonymized_sentences)
 

mailcom/test/test_parse.py

@@ -204,3 +204,21 @@ def test_pseudonymize_no_entities(get_default_fr):
     text = "Ceci est une phrase simple sans entités nommées ni chiffres."
     pseudonymized_text = get_default_fr.pseudonymize(text)
     assert pseudonymized_text == text
+
+
+def test_pseudonymize_email_addresses(get_default_fr):
+    sentence = "My email is [email protected]."
+    pseudonymized_sentence = get_default_fr.pseudonymize_email_addresses(sentence)
+    assert pseudonymized_sentence == "My email is [email]"
+
+    sentence = "Contact us at [email protected] or [email protected]."
+    pseudonymized_sentence = get_default_fr.pseudonymize_email_addresses(sentence)
+    assert pseudonymized_sentence == "Contact us at [email] or [email]"
+
+    sentence = "No email addresses here!"
+    pseudonymized_sentence = get_default_fr.pseudonymize_email_addresses(sentence)
+    assert pseudonymized_sentence == "No email addresses here!"
+
+    sentence = ""
+    pseudonymized_sentence = get_default_fr.pseudonymize_email_addresses(sentence)
+    assert pseudonymized_sentence == ""