ChatGPT's Perception of Context and Speech Acts

Abstract

This study investigates the proficiency of the AI-powered language model ChatGPT in analyzing both context and speech acts. Implementing Dell Hymes' contextual framework for context analysis and John Searle's approach for speech act analysis, the research aims to uncover ChatGPT's potentials and limitations in these domains. ChatGPT excels in specific contextual elements such as Ends, Act Sequence, Norms, and Genres but faces challenges in analyzing the other elements. Other limitations include a tendency to provide lengthy responses, repetition of details, and inconsistency in analyses across different chats. In speech act analysis, ChatGPT shows improvement compared to contextual analysis, with focused assessments resulting in higher accuracy. Similar to context analysis, inconsistencies and recurring errors are evident in speech act identification. The study concludes that ChatGPT's performance, while not flawless, demonstrates a significant degree of accuracy.