SentenceAnagram.scala

def sentenceAnagrams(sentence: Sentence): List[Sentence] = {

    def subSentence(occ: Occurrences): List[Sentence] = {
      if (occ.isEmpty) List(List())
      else
        for {
          x <- combinations(occ)
          y <- dictionaryByOccurrences(x)
          z <- subSentence(subtract(occ, x))
        } yield y :: z
    }

    subSentence(sentenceOccurrences(sentence))
  }

Pretty solution

I really hate these combination tasks

awesome! Just need to replace this line:
x <- combinations(occ) filter dictionaryByOccurrences.contains

I think that dictionaryByOccurrences.get(x) getOrElse (List()) at y, like in https://gist.github.com/valtih1978/5983b7620798beeda317, is better in place of filtering at x both neatness and performansewise.

If I would have had a million guesses I wouldn't have found that...