Citation Classification Using Natural Language Processing and Machine Learning Models

摘要: In this paper, we address the problem of identifying quality citation as important or unimportant to developments presented in research papers. We gather features represented by four state-of-the-art machine learning techniques and combined them with newly engineered, natural language-based features. Using a known dataset 465 citations, manually labeled experts, our approach out-performed using fine-tuned Random Forest Classifier 90.7% F1 score 97.7% precision. also employ Convolutional Neural Networks AdamW optimizer focal loss function - that converges quickly on small data achieve considerably significant results.