<![CDATA[Mandarin is one of the most demanding languages for adult foreign learners to speak fluently, owing to its tonal system, syllable-based homophony, and the absence of alphabetic correspondence between sound and script. Conventional classroom instruction rarely supplies the volume of individualized, low-stakes oral practice that tonal accuracy and fluency require. Artificial intelligence (AI) chatbots built on natural language processing (NLP) offer a scalable alternative, simulating conversational partners that can listen, transcribe, evaluate, and respond to spoken language in real time. This article reports a systematic literature review synthesizing 25 empirical and conceptual studies, published between 2021 and 2026, on the use of NLP-based AI chatbots to improve second-language speaking skills, with particular attention to Mandarin/Chinese-as-a-foreign-language contexts. Following a PRISMA-informed search and screening procedure across Google Scholar, Scopus, ERIC, and major computer-assisted language learning journals, studies were thematically synthesized. The review finds consistent evidence that chatbot-mediated practice improves oral fluency, pronunciation, and willingness to communicate while reducing speaking anxiety, but that this evidence is concentrated in English-as-a-foreign-language contexts; direct evidence on Mandarin tone acquisition remains scarce. The review's novelty lies in explicitly separating general EFL findings from the small Mandarin-specific evidence base and mapping NLP subcomponents onto discrete Mandarin speaking sub-skills. Implications for chatbot design and future research are discussed.]]>
Copyrights © 2026
