Abstract
Religious literature is among the most widely read ones. The Bible is the normative corpus for Christians. However, it is not a trivial text to read, due to its high degree of literary and linguistic variability, making sermon construction a laborious activity. This work proposes a method, called SwarmaBle, for automating the construction of Biblical sermons through a combination of natural language processing and swarm intelligence. The study combines quantitative and qualitative assessments of SwarmaBle, exploring its sensitivity to two hyperparameters and presenting a discussion about expert-evaluated recommended passages. The results show a consistent performance of SwarmaBle, achieving an average performance above 70% in the quantitative analysis. Domain experts reported positive evaluations regarding SwarmaBle’s ability to recommend meaningful Biblical passages for the given topic.