Incorporating Altmetrics to Support Selection and Assessment of Publications During Literature Analyses

Abstract

Background. The constantly increasing number of scientific publications poses challenges for researchers to monitor, select, and assess the publications relevant for their own research. Several guidelines for assessing publications manually during a literature analysis exist, with researchers proposing (semi-)automated techniques to facilitate such assessments. Aims. Still, research indicates that current techniques require further improvements to facilitate the analysis of large sets of publications. In this paper, we propose a semi-automatic technique with which we aim to improve in this direction by facilitating the selection and assessment of publications. Method. Our technique uses publicly available data of a publication, namely citation counts, article-level metrics, venue metrics, and altmetrics, to guide an analyst in assessing its relevance and impact. To evaluate the feasibility of our technique and the included metrics, we performed an experimental analysis to automatically assign ratings to the retrieved publications. Results. The results indicate that our technique can help an analyst in assessing publications, and reduce manual effort. Through our technique, we achieve an average accuracy of 53 % with a recall of 71 %. While precision (14 %) and F1-score (21 %) are—not surprisingly, due to the high number of irrelevant results returned by automatic searches in digital libraries—low, we see an improvement of these values for more recent reviews for which we could collect more complete data. However, some manual effort is still required for the final selection of papers. Conclusions. While it is not possible to achieve full automation for selecting and quality assessing publications, we can see that our metrics-based technique can be a helpful means to provide an initial rating for the analyst. Also, incorporating altmetrics seems to be a promising addition to rate comparably recent publications, helping researchers to further facilitate the execution of literature analyses.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 7 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.