Text mining analytics as a method of benchmarking interdisciplinary research collaboration

Abstract

This paper introduces the process of adopting and implementing modern text mining approaches of analysis within the Cluster of Excellence (CoE) Tailor-Made Fuels from Biomass (TMFB) at RWTH Aachen University and presents initial results of the analysis of research output by use of common clustering algorithms, namely Principal Component Analysis and k-means. As one main part of this paper the data driven approach is classified into benchmarking efforts, which are part of the research work of the so called Supplementary Cluster Activities. The SCA, supporting the cluster management, are initiated in order to promote interdisciplinary collaboration of CoE researchers with different disciplinary backgrounds. This cross-linking is aided by means of knowledge engineering and knowledge transfer strategies, such as the exploration of synergies and benchmarking of research results as well as progress. In this course an adoption of current benchmarking efforts to the specific cluster research framework conditions is described. At this, in case of differing data sources according to those used in widespread business organisational benchmarking, possible TMFB data sources are outlined and a selection for analysis is reasoned. While benchmarking is usually differentiated in internal and external benchmarking, in this case focus lies on internal analysis of publications in order to reflect research work. Benchmarking of publications is used and implemented to identify (best) methods, practices and processes of CoE to improve the research organization. Second major part and central question within the scope of this paper is in which way text mining respectively clustering algorithms are sensitive applicable to TMFB publications and are able to be used as benchmark for research clusters. Thus thematically priorities of TMFB researchers will be investigated in order to create an overview according to research topics, keywords and methods. In case of an outlook further steps, e.g. Dealing with generated results, data visualisation or further acquisition of data corpora, are formulated.