Proverb representation using semantic technologies: a case study of Nigerian Yoruba proverbs

Abstract

Indigenous Knowledge (IK) is the unique, traditional, local knowledge existing within and developed around the specific condition of women and men indigenous to a particular geographic area. Forms of expressing IK include folklores, songs, stories, festivals, outfit and proverbs. Proverbs are grounded upon years of experience and close observation of life and natural phenomena. Some research posed proverbs to be the bedrock of civilization of a society. By implication, almost all societies have proverbs or a form of it. The Nigerian Yoruba society is endowed with enormous proverbs which serve pivotal roles. The significance of proverbs to society and in particular Yoruba society cannot be overemphasised. Thus, attempts have been made to archive them both digitally and in print. However, present users with simple lists of proverbs to browse. These forms of representation also pay no attention to proverbs’ applicability to daily challenges; they are hard-coded and unfriendly. Thus, reducing proverbs to passive text rather than active wisdom. Due to the proliferation of intelligent linguistic agents, they can serve as media of communicating proverbs. However, agents need more than text to attain a level of intelligence whereby they respond to queries (complaints) with relevant indigenous proverbs. Hence, Yoruba proverbs are yet to be represented and preserved with technologies that maximise their usefulness. Thus, this study aimed to design a semantic proverb representation that allows linguistic AI applications to apply indigenous proverbs to domain contexts (complaints). Employing research through design methodology, minimum metadata about proverbs required to explain, interpret and apply best fitting proverb(s) to a domain context from a pool of proverbs were identified by reviewing literature. Case study based data was elicited through interview. The data gathered corroborated literature standpoint that the meaning of a proverb is context-dependent. By implication, the context of application affects the meaning of proverbs. After thematic analysis, there are four main factors that determine proverbs’ relevance to contexts, namely proverb intrinsic-attributes, proverb user attributes, context attributes and audience attributes. Subsequent to these findings, a generic computer understandable representation of proverb was designed. Notwithstanding, it must be ascertained that this representation is robust enough for intelligent agents to use when responding to complaints or query. Hence, a simple complaint-response app and a chatbot were developed by active coding. The chatbot was tested by seven users with a v complaint each. The users passed comments and corrections were made. Subsequently, it was tested and 31 complaint-response pair were recorded. These responses were given to 20 evaluators to rate their suitability to complaints on a scale of 1 to 5. The average rating was 3.12 rounded off to 3.0 implying the advice is sensible, acceptable to the listener, useful to some extent but might be general. I ascribe this limitation (over generalisation) to the application layer where models are not accurate and not the representation itself. I conclude that the representation is adequate for the purpose it was designed, but its veracity and efficacy are dependent on the accuracy of the application layer.