Seed-based information retrieval in networks of research publications: Evaluation of direct citations, bibliographic coupling, co-citations and PubMed related article score

In this contribution, we deal with seed-based information retrieval in networks of research publications. Using systematic reviews as a baseline, and publication data from the NIH Open Citation Collection, we compare the performance of the three citation-based approaches direct citation, co-citation, and bibliographic coupling with respect to recall and precision measures. In addition, we include the PubMed Related Article score as well as combined approaches in the comparison. We also provide a fairly comprehensive review of earlier research in which citation relations have been used for information retrieval purposes. The results show an advantage for co-citation over bibliographic coupling and direct citation. However, combining the three approaches outperforms the exclusive use of co-citation in the study. The results further indicate, in line with previous research, that combining citation-based approaches with textual approaches enhances the performance of seed-based information retrieval. The results from the study may guide approaches combining citation-based and textual approaches in their choice of citation similarity measures. We suggest that future research use more structured approaches to evaluate methods for seed-based retrieval of publications, including comparative approaches as well as the elaboration of common data sets and baselines for evaluation.