diff --git a/src/calibre/spell/dictionary.py b/src/calibre/spell/dictionary.py
index 402dc9667b..fe3579a480 100644
--- a/src/calibre/spell/dictionary.py
+++ b/src/calibre/spell/dictionary.py
@@ -20,8 +20,7 @@ from calibre.spell import parse_lang_code
 from calibre.utils.config import JSONConfig
 from calibre.utils.icu import capitalize
 from calibre.utils.localization import get_lang, get_system_locale
-from polyglot.builtins import filter, iteritems, itervalues, unicode_type
-
+from polyglot.builtins import filter, iteritems, itervalues, map, unicode_type
 
 Dictionary = namedtuple('Dictionary', 'primary_locale locales dicpath affpath builtin name id')
 LoadedDictionary = namedtuple('Dictionary', 'primary_locale locales obj builtin name id')
@@ -35,6 +34,11 @@ dprefs.defaults['user_dictionaries'] = [{'name':_('Default'), 'is_active':True,
 not_present = object()
 
 
+def normalize_word(word):
+    # remove soft hyphens
+    return unicode_type(word).replace('\u00ad', '')
+
+
 class UserDictionary(object):
 
     __slots__ = ('name', 'is_active', 'words')
@@ -227,14 +231,17 @@ class Dictionaries(object):
         return ans
 
     def ignore_word(self, word, locale):
+        word = normalize_word(word)
         self.ignored_words.add((word, locale.langcode))
         self.word_cache[(word, locale)] = True
 
     def unignore_word(self, word, locale):
+        word = normalize_word(word)
         self.ignored_words.discard((word, locale.langcode))
         self.word_cache.pop((word, locale), None)
 
     def is_word_ignored(self, word, locale):
+        word = normalize_word(word)
         return (word, locale.langcode) in self.ignored_words
 
     @property
@@ -268,12 +275,14 @@ class Dictionaries(object):
         for d in itervalues(self.dictionaries):
             if d and getattr(d.primary_locale, 'langcode', None) == langcode:
                 for word in words:
+                    word = normalize_word(word)
                     d.obj.add(word)
 
     def remove_user_words(self, words, langcode):
         for d in itervalues(self.dictionaries):
             if d and d.primary_locale.langcode == langcode:
                 for word in words:
+                    word = normalize_word(word)
                     d.obj.remove(word)
 
     def add_to_user_dictionary(self, name, word, locale):
@@ -282,9 +291,11 @@ class Dictionaries(object):
             raise ValueError('Cannot add to the dictionary named: %s as no such dictionary exists' % name)
         wl = len(ud.words)
         if isinstance(word, (set, frozenset)):
+            word = frozenset(map(normalize_word, word))
             ud.words |= word
             self.add_user_words(word, locale.langcode)
         else:
+            word = normalize_word(word)
             ud.words.add((word, locale.langcode))
             self.add_user_words((word,), locale.langcode)
         if len(ud.words) > wl:
@@ -297,6 +308,7 @@ class Dictionaries(object):
         return False
 
     def remove_from_user_dictionaries(self, word, locale):
+        word = normalize_word(word)
         key = (word, locale.langcode)
         changed = False
         for ud in self.active_user_dictionaries:
@@ -312,7 +324,7 @@ class Dictionaries(object):
     def remove_from_user_dictionary(self, name, words):
         changed = False
         removals = defaultdict(set)
-        keys = [(w, l.langcode) for w, l in words]
+        keys = [(normalize_word(w), l.langcode) for w, l in words]
         for d in self.all_user_dictionaries:
             if d.name == name:
                 for key in keys:
@@ -329,6 +341,7 @@ class Dictionaries(object):
         return changed
 
     def word_in_user_dictionary(self, word, locale):
+        word = normalize_word(word)
         key = (word, locale.langcode)
         for ud in self.active_user_dictionaries:
             if key in ud.words:
@@ -364,6 +377,7 @@ class Dictionaries(object):
         return changed
 
     def recognized(self, word, locale=None):
+        word = normalize_word(word)
         locale = locale or self.default_locale
         key = (word, locale)
         ans = self.word_cache.get(key, None)
@@ -392,6 +406,7 @@ class Dictionaries(object):
         return ans
 
     def suggestions(self, word, locale=None):
+        word = normalize_word(word)
         locale = locale or self.default_locale
         d = self.dictionary_for_locale(locale)
         has_unicode_hyphen = '\u2010' in word