Keyword-based Research Field Discovery with External Knowledge Aware Hierarchical Co-clustering

Helping researchers in understanding the current position and significance of their research field benefits both the individual and the development of science and technology. Clustering techniques are traditional approaches based on text mining of papers and their metadata, but suffer from inconsistent representations of words in metadata uniquely assigned by authors, resulting in low accuracy. To address this issue, we propose the application of HICCAM, a hierarchical co-clustering method that exploits auxiliary knowledge of clustered objects to discover research fields. Our first step involved constructing augmented matrices representing paper abstracts and research keywords from external domains. We then cooperatively and accurately clustered a relational matrix of these objects using these augmented matrices. To validate the effectiveness of our framework, we conducted a case study using conference papers published in the field of computer science and various auxiliary knowledge. The comparative analysis identified the domain of auxiliary knowledge that contributes to the research field detection, and the visualization results showed it effective for field discovery in terms of easy interpretation and scalability brought by the hierarchical algorithm.