Article: Automatic recognition of English discourse functions based on meta-function theory and BERT semantic representation Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.16 pp.1 - 11 Abstract: This paper proposes an automated method for identifying functional discourse segments in English teaching, integrating meta-function theory with BERT-based semantic representation. The framework constructs a seven-category label system derived from systemic functional linguistics and enhances segment embeddings through a sliding-window context modelling mechanism. A label attention strategy is incorporated to improve multi-label classification, addressing challenges such as functional overlap and ambiguous boundaries. Evaluated on multi-level English classroom corpora covering junior high school to university contexts, the method demonstrates superior performance compared to baseline models, achieving higher accuracy and better adaptability across diverse teaching styles. It also maintains stronger structural coherence in discourse organisation. The results validate the approach's effectiveness in supporting automated discourse analysis for intelligent educational systems. Inderscience Publishers - linking academia, business and industry through research

Title: Automatic recognition of English discourse functions based on meta-function theory and BERT semantic representation

Authors: Yanjing Huo

Addresses: Shuozhou Normal College, No. 85, Changning East Street, Shuozhou, Shanxi, 036000, China

Abstract: This paper proposes an automated method for identifying functional discourse segments in English teaching, integrating meta-function theory with BERT-based semantic representation. The framework constructs a seven-category label system derived from systemic functional linguistics and enhances segment embeddings through a sliding-window context modelling mechanism. A label attention strategy is incorporated to improve multi-label classification, addressing challenges such as functional overlap and ambiguous boundaries. Evaluated on multi-level English classroom corpora covering junior high school to university contexts, the method demonstrates superior performance compared to baseline models, achieving higher accuracy and better adaptability across diverse teaching styles. It also maintains stronger structural coherence in discourse organisation. The results validate the approach's effectiveness in supporting automated discourse analysis for intelligent educational systems.

Keywords: teaching discourse function; metafunction theory; BERT; multi-label classification; intelligent language education.

DOI: 10.1504/IJRIS.2026.154380

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.16, pp.1 - 11

Received: 05 Dec 2025
Accepted: 19 Jan 2026
Published online: 25 Jun 2026 *

Title: Automatic recognition of English discourse functions based on meta-function theory and BERT semantic representation

Keep up-to-date