EXPLORE PUBLICATIONS BY COUNTRIES

Parts of Speech Tagging, Setswana Relative, Rule based POS Tagging.

Setswana relatives have been shown to have a wide range of structures compared to other qualificatives. They can take negation, tense and can be recursively extended using other qualificatives, adverbs, noun phrases, and verb phrases. Studies have also shown that the structure of indirect relatives is more challenging as it is less regular compared to that of direct relatives. As a result, proposed Setswana Relatives taggers performed badly on indirect relatives. In this study, we propose the use of noun phrase and verb phrase constructs to represent the structure of indirect relatives at a high level. This approach shows that indirect relatives are also regular making them also amenable to the use of regular expressions for their identification in a sentence. This study investigates the extent to which noun phrases and verb phrases could be used to construct a regular structure for indirect relatives. We developed patterns for indirect relatives which we then implemented in Python NLTK regular expressions. The proposed tagger has a recall of 69% and precision of 62%. The tagger fails in some instances due to challenges in identifying its sub-components of noun phrases, verb phrases, and qualificatives.