Text classification of research articles in emerging fields poses significant challenges due to their complex boundaries, interdisciplinary nature, and rapid evolution. Traditional methods, which rely on manually curated search terms and keyword matching, often lack recall due to the inherent incompleteness of keyword lists. In response to this limitation, this study introduces a deep learning-based ensemble approach that addresses the challenges of article classification in dynamic research areas, using the field of artificial intelligence (AI) as a case study. Our approach included using decision tree, sciBERT and regular expression matching on different fields of the articles, and a support vector machine (SVM) to merge the results from different models. We evaluated the effectiveness of our method on a manually labeled dataset, finding that our combined approach captured around 97% of AI-related articles in the web of science (WoS) corpus with a precision of 0.92. This presents a 0.15 increase in F1-score compared with existing search term based approach. Following this, we performed an ablation study to prove that each component in the ensemble model contributes to the overall performance, and that sciBERT outperforms other pre-trained BERT models in this case.
This study presents an ensemble approach that addresses the challenges of identification and analysis of research articles in rapidly evolving fields, using the field of Artificial Intelligence (AI) as a case study. Our approach included using decision tree, sciBERT and regular expression matching on different fields of the articles, and a SVM to merge the results from different models. We evaluated the effectiveness of our method on a manually labeled dataset, finding that our combined approach captured around 97% of AI-related articles in the web of science (WoS) corpus with a precision of 0.92. This presents a 0.15 increase in F1 score compared with existing search term based approach. Following this, we analyzed the publication volume trends and common research themes.We found that compared with existing methods, our ensemble approach revealed an increased degree of interdisciplinarity, and was able to identify more articles in certain subfields like feature extraction and optimization. This study demonstrates the potential of our approach as a tool for the accurate identification of scholarly articles, which is also capable of providing insights into the volume and content of a research area.
The rapid expansion of scientific literature poses significant challenges for accurately mapping emerging interdisciplinary fields, such as Uncrewed Aerial Vehicle (UAV) logistics, characterized by evolving terminology and fluid boundaries. This study develops a hybrid text classification algorithm to automate the identification of relevant publications, surpassing limitations of conventional keyword-based searches and standalone pre-trained language models. Utilizing a manually curated dataset of 5,636 articles retrieved from Web of Science, the approach integrates semantic representations from a fine-tuned MPNet model, domain-specific curated lexicon cosine similarities, and discriminative n-gram counts within a Random Forest classifier. Validated through five-fold cross-validation and ablation studies, the framework achieves a positive-class F1-score of 0.9014, demonstrating superior precision (0.9250) and recall (0.8790) compared to baselines, including SciBERT (F1: 0.7989) and MPNet alone (F1: 0.8723). These results highlight the efficacy of multi-feature fusion in distinguishing relevant UAV logistics publications from irrelevant ones. The methodology offers a potentially generalizable solution for literature mapping in dynamic domains.