Parts of Speech Tagging: A Setswana Relative

Abstract Setswana qualificatives consists of multiple words. This makes part of speech tagging a challenging task especially for the relative part of speech. The Setswana relative has a wide variety of structure in terms of number of words, form, tense, and negation. A few studies have looked at part of speech tagging for Setswana complex parts of speech including the relative. However, these studies did not explore in detail all the different forms of a relative. In this study, we investigate the different forms of a Setswana relative and convert them into a general pattern. The relative patterns are stored in a trie data structure which is used to detect relative’s parts of speech in a given Setswana sentence. Tests show that most of the relative forms are consistent giving a performance rate of 78% for the test data. The direct relative structure gives a higher performance rate as its structure is simpler and less ambiguous compared to the indirect relative structure.