Consistency of Contextual Embedding in Literary Texts

Despite the fact that computational analysis has opened a wider window for literature and its pertinent text-based studies, it is not well-proven whether the contextual embedding technique works better with seemingly complex literary texts. In this regard, this research has investigated the consistency of contextual embedding in a literary corpus. Based on a small corpus consisting of two novels of Ian Fleming and two short stories of Guy De Maupassant, this study employed the formidable capabilities of BERT to dissect the consistency of contextual embedding in the literary texts. By the combination of a fine-tuned BERT model and the extraction of contextual embeddings, this study performed five correlated comparative studies among two shorter and two longer literary texts. A series of comparative studies among the assigned literary texts uncovered the striking contrasts within each narrative. this ensured the comprehensive examination of the semantic relationships and thematic associations between words. This study confirmed that contextual embedding worked significantly better with literary texts and produced consistent outcomes while preserving the thematic accuracy of the narratives.