Abstract:

Several ranking systems for evaluation of world universities were introduced and gained popularity in the recent years, such as ARWU, Webometrics, Leiden ranking, U-multirank etc. University of Belgrade and other Serbian universities started to get positions on these lists after restructuring and joining Bologna process. Number of published papers became one of the most important indicators in the evaluation of higher education institutions. Organizations that perform rankings take into account only the papers associated with the institution on the basis of affiliation, included into particular index database. University of Belgrade, which consists of 31 schools (e.g. faculties), 11 institutes and a number of clinics, which are loosely coupled members of the University, despite the regulations that the authors must include the name of the University into the affiliation, still has a significant number of published papers whose authors did not respect this rule. Even if the affiliation is correct and complete, it still happens that standard queries to the index databases do not return data about particular scientific paper. Some errors occur due to multiple different English translations of the institution's name. It also happens that the input field is too short for the institution's full name.

Another category of problems is related to recognition of authors' identity by personal name. Serbian language uses two alphabets - Cyrillic and Latin. The Latin alphabet includes letters with diacritical marks: čćžš. The name can be misspelled due to typographical errors, but it can also mix Cyrillic and Latin letters, or diacritical marks can be missing. The middle name (or middle letter) is sometimes unknown; a hyphen can be inserted or missing between the two second names, etc.

Another group of problems includes recognizing a person's identity in affiliation, when the name changed significantly, based on the context or additional information (change of surname on marriage, added names, nicknames, etc.). Distinguishing between persons with the same name belong to the third group of problems. A specific problem is the problem of pairing the scientific paper, identified by DOI, with authors, identified by signatures, on the basis of affiliation, using the techniques of recognizing names, as well as joining them with the institution, group of authors, the scientific field and keywords. At the international level, there are attempts to uniquely identify researchers, in order to solve this problem in an exact manner such as ResearcherID and ORCID. This paper describes the problems and benefits of mandatory inclusion of our researchers in these systems of unified identification, as well as the possible role of these identifiers in the information systems of our educational and research institution.

Most of the problems are solvable by providing detailed instructions for authors about correct affiliation of their institutions, both in Serbian and English, as well as by introduction of additional software that would process data from the index base, provide statistical analysis and report observed discrepancies. We will also present some examples of the name recognition implementations applied to specific forms and databases during the process of parsing, as well as estimation of error rates in practical person's identification.

Keywords:

University ranking, affiliation, personal names, identity.