From Confusion to Clarity: The Power of Artificial Intelligence in Understanding Linguistic Ambiguities

Abstract

In this study I investigate how artificial intelligence (AI) identifies and analyzes two types of linguistic ambiguity: structural and semantic. I also seek to bridge the gap between theoretical linguistics and recent advancements in AI. I examine examples of linguistically ambiguous sentences using two AI applications, ChatGPT-4o and Google Gemini. Using a comparative analysis, I evaluate their comprehension and capabilities in resolving linguistic ambiguities. To enhance the evaluation, I use some prompts to elicit a second interpretation in cases where the applications fail to detect either structural or semantic ambiguities. The results reveal that ChatGPT-4o is more effective at identifying such ambiguities than is Google Gemini, which often struggles with ambiguous sentences. Semantic ambiguities, particularly those involving modals such as “must” and “ought to,” or scope issues with quantifiers such as “every,” pose greater challenges for both applications than do structural ambiguities. Despite these challenges, both ChatGPT-4o and Google Gemini demonstrate useful disambiguation strategies, including referencing the context of the utterance and paraphrasing ambiguous sentences to clarify their meaning. The study highlights the significant potential of AI in imitating human capabilities related to sentence comprehension and the resolution of linguistic ambiguities.