X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/ae8694a6a6862d7cb66cd91102d2802c9899e7cf..57598a048e9124b905572ed8dc4fa9465b5d38a6:/src/nominatim_api/search/icu_tokenizer.py?ds=sidebyside

diff --git a/src/nominatim_api/search/icu_tokenizer.py b/src/nominatim_api/search/icu_tokenizer.py
index fa14531a..c18dd8be 100644
--- a/src/nominatim_api/search/icu_tokenizer.py
+++ b/src/nominatim_api/search/icu_tokenizer.py
@@ -202,7 +202,12 @@ class ICUQueryAnalyzer(AbstractQueryAnalyzer):
             standardized form search will work with. All information removed
             at this stage is inevitably lost.
         """
-        return cast(str, self.normalizer.transliterate(text))
+        norm = cast(str, self.normalizer.transliterate(text))
+        numspaces = norm.count(' ')
+        if numspaces > 4 and len(norm) <= (numspaces + 1) * 3:
+            return ''
+
+        return norm
 
     def split_query(self, query: qmod.QueryStruct) -> Tuple[QueryParts, WordDict]:
         """ Transliterate the phrases and split them into tokens.