Emotion classification and sentiment analysis for sustainable agricultural development: exploring available tools for analyzing African farmer interviews
The emerging 4th industrial revolution is having a profound effect on the direction of agrarian development. Big data technologies are becoming embedded within all walks of life, leading to both significant advancements in utility and to critical ethical concerns about the organization of the social world. Academic attention is growing into how such technologies can be employed for farmers; using enriched forms of data collection to account for contextually embedded factors in smallholder decision making. Further, in the context of ongoing COVID-19 restrictions, research is increasingly being conducted remotely. This removes a significant interpersonal dimension from studies, a particular concern for those which deal with sensitive data such as gender empowerment. In this paper we explore emotion classification and sentiment analysis of text and audio data of farmers’ interviews in eastern and southern Africa and their evaluation of a set of sustainable agricultural practices. With this relatively benign dataset, which is known not to include any instances of affective behavior beyond normal discussion of farming techniques, we attempt to test the viability of these tools and what steps are necessary to make them reliable and accessible to researchers. Findings indicate additional insight can be made to support qualitative study, in several cases demonstrating a convergence between traditional anthropological assessment and expected emotional reaction. There are also unexpected responses and unforeseen learning for the process of qualitative data collection and processing. For future research and interventions, however, a series of limitations and developments are identified for this methodology to mature.