Constrastive Pre-training for Citation Intent Classification

Our project hypothesises that better citation context representation will lead to improved classification performance and provide deeper insights into ambiguous citations. We explore the use of contrastive pre-training and joint input-label text model to enhance context representation of citation texts.