The Predictability of English Spelling Using a Part-of-Speech Tagging Method

Abstract

This article aims to reveal the reality of English spelling using a method typically employed for morphological analysis or part-of-speech (POS) tagging in natural language processing (NLP). A survey was conducted in which word likelihood and tag sequence probabilities were replaced with letter likelihood and letter sequence probabilities. From the results, this article concludes that it is challenging to predict English spelling from sounds by probability.