Question

Please translate the following to simple English:
Studies investigating the differences between speech and writing have been conducted by researchers from various fields for a variety of reasons. From an anthropological perspective, understanding such differences contribute to the study of cultural evo- lution and the role that writing and literacy play in human culture. Educators and psychologists have studied the differences in order to understand the cognitive factors affecting acquisition of both modalities, while an understanding of the lexical and grammatical differences of the two modalities has been the focus of linguists and language teach-
ers (Akinnaso 1982, Olson 1996, Hung 2017).
In this work, we study the differences between the spoken and written modalities with a different aim: to inform design choices in the development of spoken language applications for isiZulu, especially given its resource scarce context.
When developing voice-enabled computer applica- tions for a given language, it is important to have an understanding of the typical features of the spo- ken form of the language. Moreover, since corpora used for language modelling are often based on writ- ten text, it is useful to have an understanding of the differences between the spoken and written forms of the language. Features that are known to occur more frequently in spoken language could be con- sidered during development, whether by engineer- ing rules to deal with them appropriately or by en- suring that systems are trained on corpora that ex- hibit the desired features in a balanced way. This is especially important in a resource scarce context, where existing data may not perfectly fit the in- tended use case and where informed decisions must be made in order to utilise the data most effec- tively.
Research attempting to describe and quantify the differences between spoken and written language has been done for languages such as English, but not for isiZulu. In this paper, we present a quan- titative investigation into such differences by con- sidering the morphology of tokens in a transcribed spoken isiZulu corpus and a written isiZulu cor- pus. We use morpheme tags as a proxy for features that typically differ between spoken and written lan- guage, and calculate relative differences of the oc- currence of specific morpheme tags from analyses produced by ZulMorph (Pretorius & Bosch 2003), a state-of-the-art finite-state morphological analyser for isiZulu.