Article: Automatic identification and hierarchical analysis of pragmatic errors in teaching Chinese as a second language Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.20 pp.28 - 38 Abstract: Pragmatic errors in L2 Chinese are characterised by strong concealment and high communicative harmfulness, posing core challenges for automatic identification due to context dependence and fuzzy type boundaries. Grounded in speech act theory and face theory, this study constructs a three-layer classification system covering the illocutionary force layer, politeness strategy layer, and context appropriateness layer, and designs a dual-channel feature extraction method integrating local context encoding and discourse function labelling. A sequence classification model trained on hierarchically annotated corpora is verified against single-layer baseline models. The hierarchical model achieves significant improvements across all error types, with the largest gain at the context appropriateness layer (Macro-F1: 0.779, +9.4 pp). Learners from different L1 backgrounds exhibit systematic differences in hierarchical error distribution. These findings offer quantifiable bases for differentiated pragmatic teaching interventions and demonstrate the feasibility of embedding automatic pragmatic diagnosis into intelligent L2 Chinese instruction systems. Inderscience Publishers - linking academia, business and industry through research

Title: Automatic identification and hierarchical analysis of pragmatic errors in teaching Chinese as a second language

Authors: Hongbo Ma

Addresses: School of Chinese Language and Literature, Anyang Normal University, Anyang Henan, 455002, China; School of Chinese Language and Literature, Capital Normal University, Beijing, 100048, China

Abstract: Pragmatic errors in L2 Chinese are characterised by strong concealment and high communicative harmfulness, posing core challenges for automatic identification due to context dependence and fuzzy type boundaries. Grounded in speech act theory and face theory, this study constructs a three-layer classification system covering the illocutionary force layer, politeness strategy layer, and context appropriateness layer, and designs a dual-channel feature extraction method integrating local context encoding and discourse function labelling. A sequence classification model trained on hierarchically annotated corpora is verified against single-layer baseline models. The hierarchical model achieves significant improvements across all error types, with the largest gain at the context appropriateness layer (Macro-F1: 0.779, +9.4 pp). Learners from different L1 backgrounds exhibit systematic differences in hierarchical error distribution. These findings offer quantifiable bases for differentiated pragmatic teaching interventions and demonstrate the feasibility of embedding automatic pragmatic diagnosis into intelligent L2 Chinese instruction systems.

Keywords: pragmatic error; hierarchical classification; interlanguage pragmatics; natural language processing; speech act.

DOI: 10.1504/IJRIS.2026.155775

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.20, pp.28 - 38

Received: 07 May 2026
Accepted: 17 Jun 2026
Published online: 13 Aug 2026 *

Title: Automatic identification and hierarchical analysis of pragmatic errors in teaching Chinese as a second language

Keep up-to-date