Diacritic Restoration is a necessity in the processing of languages with Latinbased scripts that utilizes characters outside the basic Latin alphabet used by the English language. Yorùbá is one such language, marking an underdot (dotbelow) on three characters and tone marks on all seven vowels and two syllabic nasals. The problem of restoring underdotted characters has been fairly addressed using characters as linguistic units for restoration. However, the existing character-based approaches and word-based approach has not been able to sufficiently address the restoration of tone marks in Yorùbá. In this study, we address tone-mark restoration as a subset of diacritic restoration. We proposed using syllables derived from words as linguistic tokens for tone-mark restoration. In our experimental setup, we used Yorùbá text collected from various sources as data with a total word count of 250,336 words. These words, on syllabification, yielded 464,274 syllables. The syllables were divided into training and testing data in different proportions, ranging from 99% used for training and 1% used for testing to 70% used for training and 30% used for testing. The aim of evaluating the different proportions was to determine how the ratio of training-to-test data affected the variations that may occur in the result. We applied memory-based learning to train the models. We also set up a similar experiment using a character token to be able to compare the performance. The result showed that ,by using syllables, we were able to increase the wordlevel accuracy to 96.23% (an average of almost 15% over using characters). We also found that using 75% of the data for training and the remaining 25% for testing gives results with the least variation in a ten-fold cross validation test. Hybridizing this method that uses syllabless as processing linguistic units with other methods like lexicon lookup might likely lead to improvement over the current result.