Article: Extraction system of BiLSTM-CRF joint transfer learning Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.37 pp.89 - 103 Abstract: This research presents an academic term extraction model using a BiLSTM-CRF architecture enhanced by transfer learning. To address poor domain adaptability and scarce labelled data, a BERT model pre-trained on a general corpus is fine-tuned on academic texts. This approach transfers broad linguistic knowledge and adapts to domain-specific characteristics. The BiLSTM captures long-distance contextual dependencies, while the CRF layer optimises sequence labelling. Experiments on ACM and IEEE datasets show the model achieves an 89.42% F1-score, significantly outperforming traditional CRF and BiLSTM-CRF baselines. In small-sample scenarios, transfer learning boosts F1 by 8.6%. The system effectively reduces domain dependence and labelled data requirements, providing an efficient tool for automating academic knowledge processing. Inderscience Publishers - linking academia, business and industry through research

Title: Extraction system of BiLSTM-CRF joint transfer learning

Authors: XiaoJiao Peng

Addresses: Southwest Jiaotong University Hope College, Chengdu, 610400, China

Abstract: This research presents an academic term extraction model using a BiLSTM-CRF architecture enhanced by transfer learning. To address poor domain adaptability and scarce labelled data, a BERT model pre-trained on a general corpus is fine-tuned on academic texts. This approach transfers broad linguistic knowledge and adapts to domain-specific characteristics. The BiLSTM captures long-distance contextual dependencies, while the CRF layer optimises sequence labelling. Experiments on ACM and IEEE datasets show the model achieves an 89.42% F1-score, significantly outperforming traditional CRF and BiLSTM-CRF baselines. In small-sample scenarios, transfer learning boosts F1 by 8.6%. The system effectively reduces domain dependence and labelled data requirements, providing an efficient tool for automating academic knowledge processing.

Keywords: term extraction; academic English; BiLSTM; conditional random field; CRF; transfer learning; sequence annotation; natural language processing; NLP.

DOI: 10.1504/IJICT.2026.153262

International Journal of Information and Communication Technology, 2026 Vol.27 No.37, pp.89 - 103

Received: 23 Oct 2025
Accepted: 16 Dec 2025
Published online: 29 Apr 2026 *

Title: Extraction system of BiLSTM-CRF joint transfer learning

Keep up-to-date