Transformer-Based Agricultural Term Extraction Using ALBERT

Abstract The effectiveness of ALBERT, a lightweight transformer-based model, for the job of extracting agricultural terms from textual corpora relevant to a given domain is examined in this paper. Although models such Agricultural BERT, SCIBERT, and RoBERTa have been used in previous methods, this work is the first to investigate ALBERT in this particular setting. By means of thorough testing, we show that ALBERT produces competitive results while utilizing a much less number of parameters. Our results demonstrate its applicability for situations that demand accuracy without sacrificing computational efficiency. The findings promote resource-efficient models for information extraction in low-resource and real-time contexts and establish ALBERT as a competitive option for scalable, domain-specific natural language processing jobs in agriculture.