On the Correction of "Old" Omitted Citations by Bibliometric Databases

Omitted citations – i.e., missing links between a cited paper and the corresponding citing papers – are the main consequence of several bibliometric-database errors. To reduce these errors, databases may undertake two actions: (i) improving the control of the (new) papers to be indexed, i.e., limiting the introduction of “new” dirty data, and (ii) detecting and correcting errors in the papers already indexed by the database, i.e., cleaning “old” dirty data. The latter action is probably more complicated, as it requires the application of suitable errordetection procedures to a huge amount of data. Based on an extensive sample of scientific papers in the Engineering-Manufacturing field, this study focuses on old dirty data in the Scopus and WoS databases. To this purpose, a recent automated algorithm for estimating the omitted-citation rate of databases is applied to the same sample of papers, but in three different-time sessions. A database’s ability to clean the old dirty data is evaluated considering the variations in the omitted-citation rate from session to session. The major outcomes of this study are that: (i) both databases slowly correct old omitted citations, and (ii) a small portion of initially corrected citations can surprisingly come off from databases over time. Conference Topic Data Accuracy and disambiguation Introduction An important branch of the bibliometric literature examines errors in bibliometric databases. Several studies show that the major consequence of database errors is represented by omitted citations, i.e., citations that should be ascribed to a certain (cited) paper but, for some reason, are lost (Moed, 2005; Buchanan, 2006; Jacsó, 2006, Li et al., 2010; Olensky, 2013). Franceschini et al. (2013) proposed an automated algorithm for estimating the omittedcitation rate of bibliometric databases. This algorithm requires the combined use of two or more bibliometric databases and is based upon the hypothesis that the mismatch between the citations occurring in one database and another one is evidence of possible errors/omissions. In a further study by Franceschini et al. (2014), this algorithm was applied to a relatively large set of publications, showing that, depending on the bibliometric database in use (Scopus or WoS), omitted citations are not distributed uniformly among publishers; e.g., regarding the publications in the Engineering-Manufacturing field, citations from papers published by Wiley-Blackwell are more likely to be omitted by Scopus, while those from papers published by ASME (American Society of Mechanical Engineers) are more likely to be omitted by WoS. A reason behind this result is that some editorial styles imposed by certain publishers can probably hamper the correct identification of the cited papers by some databases. The presence of database errors, as well as journal coverage or author disambiguation, is probably one of the major concerns of database administrators. In the authors’ opinion, database administrators may undertake two actions for reducing database errors: 1. Limiting the introduction of “new” dirty data in a database, i.e., errors concerning new papers to be indexed; 2. Cleaning “old” dirty data, i.e., errors concerning papers/journals already indexed by a database.