Cited by 145Second Language Acquisition and LearningArtificial Intelligence in EducationEFL/ESL Teaching and Learning

The impact of chatbots based on large language models on second language vocabulary acquisition

Open access
Zhang, Zhihui, Huang, Xiaomeng·2024·Heliyon

Abstract

In recent years, the integration of artificial intelligence (AI) and machine learning (ML) into education, particularly for Personalized Language Learning (PLL), has garnered significant attention. This approach tailors interventions to address the unique challenges faced by individual learners. Large Language Models (LLMs), including Chatbots, have demonstrated a substantial potential in automating and enhancing educational tasks, effectively capturing the complexity and diversity of human language. In this study, 52 foreign language students were randomly divided into two groups: one with the assistance of a Chatbot based on LLMs and one without. Both groups learned the same series of target words over eight weeks. Post-treatment assessments, including systematic observation and quantitative tests assessing both receptive and productive vocabulary knowledge, were conducted immediately after the study and again two weeks later. The findings demonstrate that employing an AI Chatbot based on LLMs significantly aids students in acquiring both receptive and productive vocabulary knowledge during their second language learning journey. Notably, Chatbots contribute to the long-term retention of productive vocabulary and facilitate incidental vocabulary learning. This study offers valuable insights into the practical benefits of LLM-based tools in language learning, with a specific emphasis on vocabulary development. Chatbots utilizing LLMs emerge as effective language learning aids. It emphasizes the importance of educators understanding the potential of these technologies in L2 vocabulary instruction and encourages the adoption of strategic teaching methods incorporating such tools.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1016/j.heliyon.2024.e25370

Cites

45 of 48 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • ChatGPT is open accessible, has extensive linguistic capabilities, and can simulate human-like dialogue across topics, providing immersive learning experiences that foster deeper cognitive processing.

    “Among these tools, ChatGPT has garnered significant attention due to its open accessibility, extensive linguistic capabilities, and ability to simulate human-like dialogue across a wide range of topics, thereby providing immersive learning experiences that foster deeper cognitive processing (Mugableh, 2024;Zhang & Huang, 2024).”

    From Integrating ChatGPT for vocabulary learning and retention: A classroom-based study of Saudi EFL learners (2025) · cites this paper as Zhang & Huang, 2024

    The Introduction (page not provided) says AI chatbots mimic human conversation, and the Conclusion (page not provided) says LLM chatbots generate natural, fluent responses and provide an interactive language-learning environment. The paper supports the dialogue and linguistic-capability aspects at the level of LLM chatbots, but it does not establish ChatGPT’s open accessibility or show that immersive experiences foster deeper cognitive processing. (LLM, conf=0.84)

    Experimental — accuracy not yet measured
  • Studies on ChatGPT in language learning found that AI-powered chatbots facilitate contextualized vocabulary acquisition, exposing learners to authentic, real-world language use.

    “Similarly, studies on ChatGPT in language learning (e.g., Karataş et al., 2024;Zhang & Huang, 2024) found that AI-powered chatbots facilitate contextualized vocabulary acquisition, exposing learners to authentic, real-world language use.”

    From Integrating ChatGPT for vocabulary learning and retention: A classroom-based study of Saudi EFL learners (2025) · cites this paper as Zhang & Huang, 2024

    The Finding and discussion section reports improved vocabulary acquisition and incidental learning with LLM chatbots. The Literature review describes learning words through context, and the Methodology describes examples and conversation practice. The paper does not establish that learners were exposed to authentic, real-world language use; its described activities include chatbot conversations and mock interviews (page numbers unavailable). (LLM, conf=0.86)

    Experimental — accuracy not yet measured
  • The results align with prior research advocating GenAI chatbots in language learning, particularly for promoting interactive, adaptive engagement.

    “These results align with prior research advocating GenAI chatbots in language learning, particularly for promoting interactive, adaptive engagement (Kohnke et al., 2023;Yigci et al., 2024;Zhang & Huang, 2024).”
  • Zhang and Huang (2024) found that an LLM-powered chatbot led to gains in both receptive and productive vocabulary among Chinese high school students.

    “Zhang and Huang (2024) investigated an LLM-powered chatbot's role in vocabulary learning among Chinese high school students, reporting gains in both receptive and productive vocabulary.”

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 2

  1. [14]Levy M. Computer-assisted Language Learning: Context and Conceptualization 1997 Oxford University Press
  2. [30]Daniels-Koch O. Freedman R. The Expertise Problem: Learning from Specialized Feedback 2022 10.48550/arXiv.2211.06519 arXiv preprint arXiv:2211.0651 · DOI

Cited by

Sorted by citations

Authors
Zhang, Z., & Huang, X.
Affiliations
  • University of Southern California (US)
  • Alibaba Group (China) (CN)
Corresponding author
Zhang, Zhihui
Venue
Heliyon
Publisher
Elsevier BV
Published
2024-02
Volume
10
Issue
3
Pages
e25370
ISSN
2405-8440 (ISSN-L)
OpenAlex
W4391429431
Cited by
107
References
46
Citations by year
2026: 44 · 2025: 49 · 2024: 14
FWCI
83.49
Type
journal article
Language
en
Open access
Locations
  • Heliyon (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Topics
  • Social Sciences › Psychology › Developmental and Educational Psychology › Second Language Acquisition and Learning
  • Physical Sciences › Computer Science › Information Systems › Artificial Intelligence in Education
  • Social Sciences › Arts and Humanities › Language and Linguistics › EFL/ESL Teaching and Learning
Keywords
chatbots, large language models, second language vocabulary acquisition, personalized language learning, receptive vocabulary knowledge, productive vocabulary knowledge, incidental vocabulary learning, vocabulary retention, foreign language learners, AI-assisted language learning

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review