Doctoral candidate Jinseok Kim has been awarded a Eugene Garfield Doctoral Dissertation Fellowship by Beta Phi Mu, the International Library and Information Studies Honor Society. Up to six recipients are selected each year for this prestigious award, a national competition among doctoral students who are working on their dissertations. The amount awarded for each fellowship is $3,000.
"The Eugene Garfield Dissertation Fellowship will be a tremendous benefit to my doctoral research. It is a recognition for my work and will provide me valuable resources for gaining new knowledge," said Kim.
Kim's research focuses on the role of data processing in knowledge discovery from data. His dissertation is titled, “The impact of author name disambiguation on knowledge discovery from big scholarly data.”
Abstract: By utilizing large-scale bibliometric data, scholars in diverse fields gleaned knowledge for use in scholarly evaluation, collaborator recommendations, and network-evolution modeling. A common challenge has been that author names in bibliometric data are not properly disambiguated: authors may share the same name (different authors are sometimes misrepresented to be a single author; merging of identities). In addition, one author may use name variations (an author may be represented as two or more different authors; splitting of identities). When faced with these authority-control challenges, a majority of scholars have processed bibliometric data using simple heuristics: if two author names share the same surname and given name initials, they are presumed to refer to the same author. Furthermore, without proper justification, those scholars have based their choice of data processing on the assumption that their findings are robust to authority-control errors.
My dissertation tests this assumption by measuring the impact of author name ambiguity on network properties. I accomplish this under varying conditions, including network size and time window using four large-scale bibliometric datasets that cover: biomedicine, computer science, physics, and one nation’s entire domestic publication output (Korea). For this, statistical properties of collaboration networks generated from algorithmically disambiguated data (i.e., close to clean data) are compared against those of the same networks but compromised by misidentified authors due to name ambiguity. My findings show that data processing can severely distort both our micro-level and macro-level understanding of a given network. This distortion can sometimes lead to false knowledge of network formation and evolution mechanisms such as preferential attachment generating power-law distribution of node degree. In addition, my dissertation explores whether compromised author names can be identified by their network-based characteristics, and provides practical guidance for scholars and decision makers.