Identifying Important Citation Through Cascading Models and NLP Techniques

In this paper, we present a new method for identifying meaningful citations within research articles. Our approach leverages a set of features, including the similarity between titles (citing and cited articles) and the similarity between the citation context and the abstract of the cited article. In addition to these features, we also incorporate those used in the previous work by Hassan et al. [1]. We integrate these features in two ways: first, by directly augmenting the features from the previous work, and second, by employing an ensemble model that cascades predictions of context-abstract similarity as supplementary.