Merge branch 'main' into 2024-release-candidate

Author: jmccrae

NEW_SYNSETS.md

@@ -72,6 +72,11 @@ substituted, e.g., "a happy life"/"a felicitous outcome". This does not mean
 that they can be substituted in every sense, e.g., "happy to help" but not
 *"felicitous to help".
 
+For verbs, two verbs are distinct if they differ in the subject, but not if the 
+direct object or any other argument can be substituted with a quantifier such as
+"something" or "someone". For example, "to eat" and "to eat something" are not 
+distinct senses. However, "to change" and "to change something" are distinct senses
+as the subject has a different semantic role.
 
 ## Well-defined
 

scripts/from_yaml.py

@@ -22,10 +22,20 @@ def map_sense_key(sk):
     """
     if "%" in sk:
         e = sk.split("%")
-        return ("oewn-" + e[0].replace("'","-ap-").replace("/","-sl-").replace("!","-ex-").replace(",","-cm-").replace(":","-cn-").replace("+","-pl-") +
-            "__" + e[1].replace("_","-sp-").replace(":","."))
+        if len(e) > 2:
+            lemma = "%".join(e[:-1])
+            info = e[-1]
+        else:
+            lemma = e[0]
+            info = e[1]
+        return ("oewn-" + lemma.replace("'","-ap-").replace("/","-sl-").
+                replace("!","-ex-").replace(",","-cm-")
+                .replace(":","-cn-").replace("+","-pl-") +
+            "__" + info.replace("_","-sp-").replace(":","."))
     else:
-        return "oewn-" + sk.replace("%", "__").replace("'","-ap-").replace("/","-sl-").replace("!","-ex-").replace(",","-cm-").replace(":","-cn-").replace("+","-pl-")
+        return ("oewn-" + sk.replace("%", "__").replace("'","-ap-").
+            replace("/","-sl-").replace("!","-ex-").
+            replace(",","-cm-").replace(":","-cn-").replace("+","-pl-"))
 
 def unmap_sense_key(sk):
     """
@@ -52,15 +62,6 @@ def make_pos(y, pos):
     else:
         return pos
 
-
-def make_sense_id(y, lemma, pos):
-    """
-    Create a sense ID from a YAML entry
-    """
-    return "oewn-%s-%s-%s" % (
-        escape_lemma(lemma), make_pos(y, pos), y["synset"][:-2])
-
-
 def sense_from_yaml(y, lemma, pos, n):
     """
     Create a Sense object from the YAML data

scripts/sense_keys.py

@@ -124,7 +124,6 @@ def get_sense_key(wn, e, s):
     lemma = (e.lemma.written_form
         .replace(" ", "_")
         .replace("&apos", "'")
-        .replace(":", "-cn-")
         .lower())
     ss_type = ss_types[ss.part_of_speech]
     lex_filenum = lex_filenums[ss.lex_name]
@@ -149,12 +148,12 @@ def unmap_sense_key(sk, KEY_PREFIX_LEN=5):
         oewn_key = e[0][KEY_PREFIX_LEN:]
         r = "__".join(e[1:])
         return (oewn_key.replace("-ap-", "'").replace("-sl-", "/").replace("-ex-", "!")
-                .replace("-cm-",",").replace("-cl-",":").replace("-pl-","+") +
+                .replace("-cm-",",").replace("-cn-",":").replace("-pl-","+") +
             "%" + r.replace(".", ":").replace("-sp-","_"))
     else: 
         return (sk[KEY_PREFIX_LEN:].replace("__", "%").replace("-ap-", "'")
                 .replace("-sl-", "/").replace("-ex-", "!").replace("-cm-",",")
-                .replace("-cl-",":").replace("-pl-","+"))
+                .replace("-cn-",":").replace("-pl-","+"))
 
 
 

scripts/validate.py

@@ -277,6 +277,7 @@ def main():
                     errors += 1
 
     instances = set()
+    ilis = set()
 
     for synset in wn.synsets:
         if synset.id[-1:] != synset.part_of_speech.value:
@@ -374,6 +375,12 @@ def main():
                     (synset.id, item[1], item[0]))
                 errors += 1
 
+        if synset.ili != "in" and synset.ili in ilis:
+            print(f"ERROR: ILI {synset.ili} is duplicated")
+            errors += 1
+        else:
+            ilis.add(synset.ili)
+
     for synset in wn.synsets:
         for sr in synset.synset_relations:
             if sr.rel_type == SynsetRelType.HYPERNYM:

scripts/wordnet.py

@@ -102,7 +102,7 @@ def change_sense_id(self, sense, new_id):
         self.sense2synset[new_id] = sense.synset
         self.id2sense[new_id] = sense
 
-    def to_xml(self, xml_file, part=True):
+    def to_xml(self, xml_file, part=False):
         xml_file.write("""<?xml version="1.0" encoding="UTF-8"?>\n""")
         if part:
             xml_file.write(
@@ -241,8 +241,6 @@ def to_xml(self, xml_file, comments):
             n_str = " adjposition=\"%s\"" % self.adjposition
         else:
             n_str = ""
-        if self.n >= 0:
-            n_str = "%s n=\"%d\"" % (n_str, self.n)
         if self.sense_key:
             sk_str = " dc:identifier=\"%s\"" % escape_xml_lit(self.sense_key)
         else:
@@ -820,6 +818,8 @@ def elc(c):
             return '-ex-'
         elif c == '+':
             return '-pl-'
+        elif c == '%':
+            return '-pc-'
         elif xml_id_char_re.match(c):
             return c
         raise ValueError(f'Illegal character {c}')