Exploiting Paper Contents and Citation Links to Identify and Characterise Specialisations

A scientific domain consists of subfields that can be further refined into specialisations. Specialisations emerge, evolve and consolidate, as reflected in particular in literature development, along a contents-based dimension where important problems are stated and addressed, and along a communal dimension where researchers collaborate and compete to solve those problems. We propose a generic framework that aims at effectively identifying and characterising the main specialisations of the subfields of a scientific domain by leveraging both paper contents and citation links. More specifically, the latent knowledge structure of a domain is discovered and progressively refined along both the contents-based and communal dimensions. Qualitative and quantitative experimental results show that our method can identify fine-grained specialisation of subfields and characterise them with key attributes (keywords, key papers and key authors), providing insights that are beyond the resolution limit of non-specialised approaches. One of the direct benefits of this research is to fulfil the highly specialised information needs of a scholarly researcher and significantly facilitate literature exploration.