Improving co-authorship network structures by combining multiple data sources: evidence from Italian academic statisticians

The aim of the present contribution is to merge bibliographic data for members of a bounded scientific community in order to derive a complete unified archive, with top-international and nationally oriented production, as a new basis to carry out network analysis on a unified co-authorship network. A two-step procedure is used to deal with the identification of duplicate records and the author name disambiguation. Specifically, for the second step we strongly drew inspiration from a well-established unsupervised disambiguation method proposed in the literature following a network-based approach and requiring a restricted set of record attributes. Evidences from Italian academic statisticians were provided by merging data from three bibliographic archives. Non-negligible differences were observed in network results in the comparison of disambiguated and not disambiguated data sets, especially in network measures at individual level.