Leveraging Lora Fine-Tuning and Knowledge Bases for Construction Identification

Liu Kaipeng* and Wu Ling

Received : August 04, 2026 | Published : August 24, 2026

Citation: Kaipeng, L, Ling, W. (2026). Leveraging Lora Fine-Tuning and Knowledge Bases for Construction Identification. Journal of Applied Sciences and Applications in Engineering, 2(2), 1–11.

License: This article is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0) , which permits use, sharing, adaptation, distribution, and reproduction in any medium or format, provided appropriate credit is given to the original author(s) and the source, a link to the Creative Commons licence is provided, and any changes made are indicated.

Abstract

This study investigates automatic identification of the English ditransitive construction by integrating LoRA fi ne-tuning of a large language model with a Retrieval-Augmented Generation (RAG) framework. A binary classification task was conducted on annotated data from the British National Corpus. Results show that a LoRA-fi ne-tuned Qwen3-8B model significantly outperformed both a native Qwen3-MAX model and a theory-only RAG system, achieving an accuracy of 0.936 and an F1 score of 0.874. Error analysis reveals that fi ne-tuning shifts the model’s judgment from surface-form matching towards semantically grounded understanding. The RAG system’s poor performance underscores the necessity of combining theoretical knowledge with instance-based learning. This work provides an effective, low-cost computational pathway for implementing constructionist theories and offers a practical tool for linguistic research and second language pedagogy.

Keywords: Construction Grammar; Ditransitive Construction; Large Language Models; LoRA Fine-tuning; Retrieval-Augmented Generation.