In many languages (including Polish) having a list of all words with number of syllables for each of them is superfluous. There are a few rules that allow to correctly compute number of syllables for >99% of words, and they are as follows:

• In general, one vowel = one syllable.
• If "i" appears in front of any other vowel, it's not counted.
• Repetition of the same vowel is also a single syllable (e.g. "czeeeemuuuu", translated as "whyyyy").
• "au" and "eu" are the complicated ones (though, luckily, they are quite rare):
  • In most cases "u" is pronounced as "ł", so it's a single syllable (this is what I implemented, I omitted the other cases due to their difficulty).
  • However, if "au" or "eu" falls on the border between the root and a prefix/suffix, then these vowels are pronounced separately.
  • And there are a few exceptions (Wiktionary lists them as: nauka, nauczka, nauczanie, nauczenie, laurka, Zeus, Dzeus, Seul, neutron, neutralny -- not counting non-lemma forms) where, again, "au" or "eu" are pronounced as separate vowels.
• All other vowel sequences are pronounced separately (so back to point one).

I had to rewrite the code a bit to allow a different set of syllable rules to be used (i.e. a different regex). Moreover:

• Replaced sapply() with vapply() (the former simplifies uncontrollably).
• Had to set default dictionary (i.e. when language is not "en") to an empty named integer vector, so that the code doesn't try to subset NULL later on (which resulted in returning complete garbage).
• Covered Polish examples with tests, some are skipped due to that implementing all rules "au" and "eu" can't be done with just a regex and will require creating a dictionary with exceptions.
• Updated README to reflect implementing rules for Polish language.

Tried to stick to your code style, I believe I succeeded at that. Should I add myself as a contributor?
Hopefully this will make it easier to implement other languages too!