Learning a new language is a difficult endeavor, the main encountered problem being vocabulary acquisition. The learning process can be improved through visual representations of coherent contexts, best represented in serious games. The game described in this paper, Semantic Boggle, is a serious game that exercises vocabulary. It is based on the traditional word-guessing game, but it brings educational value by identifying semantically-related words. The words are found using the ReaderBench framework and are placed in the game grid using a greedy algorithm. The assigned score is computed as the semantic similarity value multiplied by the normalized length of the seed. Our preliminary validation consisted of 20 users, who emphasized its interest and playability.